RAN Unleashed … Strategies for being the best (or the worst) cellular network (Part III).

I have been spending my holiday break this year (December 2021) updating my dataset on Western Europe Mobile Operators, comprising 58+ mobile operators in 16 major Western European markets, focusing on spectrum positions, market dynamics, technology diffusion (i.e., customer migration to 5G), advanced antenna strategies, (modeled) investment levels and last but not least answering the question: what makes a cellular network the best in a given market or the world. What are the critical ingredients for an award-winning mobile network?

An award-winning cellular network, the best network, also provides its customers with a superior experience, the best network experience possible in a given market.

I am fascinated by the many reasons and stories we tell ourselves (and others) why this or that cellular network is the best. The story may differ whether you are an operator, a network supplier, or an analyst covering the industry. I have had the privileged to lead a mobile network (T-Mobile Netherlands) that won the Umlaut best mobile network award in The Netherlands since 2016 (5 consecutive times) and even scored the highest amount of points in the world in 2019 and 2020/2021. So, I guess it would make me a sort of “authority” on winning best network awards? (=sarcasm).

In my opinion and experience, a cellular operator has a much better than fair chance at having the best mobile network, compared to its competition, with access to the most extensive active spectrum portfolio, across all relevant cellular bands, implemented on a better (or best) antenna technology (on average) situated on a superior network footprint (e.g., more sites).

For T-Mobile Netherlands, firstly, we have the largest spectrum portfolio (260 MHz) compared to KPN (205 MHz) and Vodafone (215 MHz). The spectrum advantage of T-Mobile, as shown above, is both in low-band (< 1800 MHz) as well as mid-band range (> 1500 MHz). Secondly, as we started out back in 1998, our cell site grid was based on 1800 MHz, requiring a denser cell site grid (thus, more sites required) than the networks based on 900 MHz of the two Dutch incumbent operators, KPN and Vodafone. Therefore, T-Mobile ended up with more cell sites than our competition. We maintained the site advantage even after the industry’s cell grid densification needs of UMTS at 2100 MHz (back in the early 2000s). Our two very successful mergers have also helped our site portfolio, back in 2007 acquiring and merging with Orange NL and in 2019 merging with Tele2 NL.

The number of sites (or cells) matter for coverage, capacity, and overall customer experience. Thirdly, T-Mobile was also first in deploying advanced antenna systems in the Dutch market (e.g., aggressive use of higher-order MiMo antennas) across many of our frequency bands and cell sites. Our antenna strategy has allowed for a high effective spectral efficiency (across our network). Thus, we could (and can) handle more bits per second in our network than our competition.

Moreover, over the last 3 years, T-Mobile has undergone (passive) site modernization that has improved coverage and quality for our customers. This last point is not surprising since the original network was built based on a single 1800 MHz frequency, and since 1998 we have added 7 additional bands (from 700 MHz to 2.5 GHz) that need to be considered in the passive site optimization. Of course, as site modernization is ongoing, an operator (like T-Mobile) also should consider the impact of future bands that may be required (e.g., 3.x GHz). Optimize subject to the past as well as the future spectrum outlook. Last but not least, we at T-Mobile have been blessed with a world-class engineering team that has been instrumental in squeezing out continuous improvements of our cellular network over the last 6 years.

So, suppose you have 25% less spectrum than a competitor. In that case, you either need to compensate by building 25% more cells (very costly & time-consuming), deploying better antennas with a 25% better effective spectral efficiency (limited, costly & relatively easy to copy/match), or a combination of both (expensive & time-consuming). The most challenging driver to copy for network superiority is the amount of spectrum. A competitor only compensates by building more sites, deploying better antenna technology, and over decades to try to equalize spectrum position is subsequent spectrum auctions (e.g., valid for Europe, not so for the USA where acquired spectrum usually is owned in perpetuity).

T-Mobile has consistently won the best mobile network award over the last 6 years (and 5 consecutive times) due to these 3 multiplying core dimensions (i.e., spectrum × antenna technology × sites) and our world-class leading engineering team.

THE MAGIC RECIPE FOR CELLULAR PERFORMANCE.

We can formalize the above network heuristics in the following key (very beautiful IMO) formula for cellular network capacity measured in throughput (bits per second);

It is actually that simple. Cellular capacity is made as simple as possible, dependent on three basic elements, but not more straightforward. Maybe, super clear, though only active spectrum counts. Any spectrum not deployed is an opportunity for a competitor to gain network leadership on you.

If an operator has a superior spectrum position and everything else is equal (i.e., antenna technology & the number of sites), that operator should be unbeatable in its market.

There are some caveats, though. In an overloaded (congested) cellular network, performance would decrease, and superior network performance would be unlikely to be ensured compared to competitors not experiencing such congestion. Furthermore, spectrum superiority must be across the depth of the market-relevant cellular frequencies (i.e., 600 MHz – 3.x GHz and higher). In other words, if a cellular operator “only” has to work with, for example, 100 MHz @ 3.5GHz, it is unlikely that this would guarantee a superior network performance across a market (country) compared to a much better balance spectrum portfolio.

The option space any operator has is to consider the following across the three key network quality dimensions;

Let us look at the hypothetical Western European country Mediana. Mediana, with a population of 25 million, has 3 mobile operators each have 8 cellular frequency bands, incumbent Winky has a total cellular bandwidth of 270 MHz, Dipsy has 220 MHz, and Po has 320 MHz (top their initial weaker spectrum position through acquisitions). Apart from having the most robust spectrum portfolio, Po also has more cell sites than any other in the market (10,000) and keeps winning the best network award. Winky, being the incumbent, is not happy about this situation. No new spectrum opportunities will become available in the next 10 years. Winky’s cellular network, based initially on 900MHz but densified over time, has about 20% fewer sites than Po. Po and Winky’s deployed state of antenna technology is comparable.

What can Winky do to gain network leadership? Winky has assessed that Po has ca. 20% stronger spectrum position than they, state of antenna technology is comparable, and they (Po) have ca. 20% more sites. Using the above formula, Winky estimates that Po’s have 44% more raw cellular network quality available compared to their own capability. Winky’s commenced a network modernization program that adds another 500 new sites and significantly improves their antenna technology. After this modernization program, Winky has decreased its site deficit to having 10% fewer sites than Po and almost 60% better antenna technology capability than Po. Overall, using the above network quality formula, Winky has changed their network position to a lead over Po with ca. 18%. In theory, it should have an excellent chance to capture the best network award.

Of course, Po could simply follow and deploy the same antenna technology as Winky and would easily overtake Winky’s position due to its superior spectrum position (that Winky cannot beat the next 10 to 15 years at least).

In economic terms, it may be tempting to conclude that Winky has avoided 625 Million Euro in spectrum fees by possessing 50 MHz less than Po (i.e., median spectrum fee in Mediana is 0.50 Euro per MHz per pop times the avoided 50 MHz times the population of Mediana 25 Million pops) and that for sure should allow Winky to make a lot of network (and market) investments to gain network leadership. By adding more sites, assuming it is possible to do where they are needed and invest in better antenna technology. However, do the math with realistic prices and costs incurred over a 10 to 15 year period (i.e., until the next spectrum opportunity). You may be more likely to find a higher total cost for Winky than the spectrum fee avoidance. Also, the strategy of Winky is easy to copy and overtake in the next modernization cycle of Po.

Is there any value for operators engaging in such the best network equivalent of a “nuclear arms” race? That interesting question is for another article. Though the answer (spoiler alert) is (maybe) not so black and white as one may think.

An operator can compensate for a weaker spectrum position by adding more cell sites and deploying better antenna technologies.

A superior spectrum portfolio is not an entitlement. Still, an opportunity to become the sustainable best network in a given market (for the duration that spectrum is available to the operator, e.g., 10 – 15 years in Europe at least).

WESTERN EUROPE SPECTRUM POSITIONS.

A cellular operator’s spectrum position is an important prerequisite for superior performance and customer experience. If an operator has the highest amount of spectrum (well balanced over low, mid, and high-frequency bands), it will have a powerful position to become the best network in that given market. Using Spectrum Monitor’s Global Mobile Frequency database (last updated May 2021), I analyzed the spectrum position of a total of 58 cellular operators in 16 Western European markets. The result is shown below as (a) Total spectrum position, (b) Low-band spectrum position covering spectrum below and including 1500 MHz (SDL band), and (c) Mid-band spectrum covering the spectrum above 1500 MHz (SDL band). For clarity, I include the 3.X GHz (C-band) as mid-band and do not include any mmWave (n257 band) positions (anyway would be high band, obviously).

4 operators are in a category by themselves with 400+ MHz of total cellular bandwidth in their spectrum portfolios; A1 (Austria), TDC (Denmark), Cosmote (Greece), and Swisscom (Switzerland). TDC and Swisscom have incredibly strong low-band and mid-band positions compared to their competition. Magenta in Austria has a 20 MHz advantage to A1 in low-band (very good) but trails A1 with 92 MHz in mid-band (not so good). Cosmote slightly follows behind on low-band compared to Vodafone (+10 MHz in their favor), and they head the Greek race with +50 MHz (over Vodafone) in mid-band. All 4 operators should be far ahead of their competitors in network quality. At least if they used their spectrum resources wisely in combination with good (or superior) antenna technologies and a sufficient cellular network footprint. In all else being equal, these 4 operators should be sustainable unbeatable based on their incredible strong spectrum positions. Within Western Europe, I would, over the next few years, expect to see all round best networks with very high best network benchmark scores in Denmark (TDC), Switzerland (Swisscom), Austria (A1), and Greece (Cosmote). Western European countries with relatively more minor surface areas (e.g., <100,000 square km) should outperform much larger countries.

In fact, 3 of the 4 top spectrum-holding operators also have the best cellular networks in their markets. The only exception is A1 in Austria, which lost to Magenta in the most recent Umlaut best network benchmark. Magenta has the best low-band position in the Austrian market, providing for above and beyond cellular indoor-quality coverage that the low-band provides.

There are so many more interesting insights in my collected data. Alas for another article at another time (e.g., topics like the economic value of being the best and winning awards, industry investment levels vs. performance, infrastructure strategies, incumbent vs. later stages operator dynamics, 3.X GHz and mmWave positions in WEU, etc…).

The MNO rank within a country will depend on the relative spectrum position between 1st and 2nd operator. If below 10% (i.e., dark red in chart below), I assess that it will be relative easy for number 2 to match or beat number 1 with improved antenna technology. As the relative strength of the spectrum position of number 1 relative to number 2 is increased, it will become increasingly difficult (assuming number 1 uses an optimal deployment strategy).

The Stars (e.g., #TDCNet / #Nuuday#Swisscom and #EE) have more than a 30% relative spectrum strength compared to the 2nd ranked MNO in a given market. They will have to severely mess up, not to take (or have!) the best cellular network position in their relevant markets. Moreover, network economically, the Stars should have a substantial better Capex position compared to their competitors (although 1 of the Stars seem a “bit” out-of-whack in their sustainable Capex spend, but may be due to fixed broadband focus as well?). As a “cherry on the pie” both Nuuday/TDCNet and Swisscom have some of the strongest spectral overhead positions (i.e., MHz per pop) in Western Europe (relative small populations to very strong spectrum portfolios), which is obviously should enable superior customer experience.

HOW AND HOW NOT TO WIN BEST NETWORK AWARDS.

Out of the 16 cellular operators having the best networks (i.e., rank 1), 12 (75%) also had the strongest (in market) spectrum positions. 3 Operators having the second-best spectrum position ended up taking the best network position, and 1 operator (WindTre, Italy) with the 3rd best spectrum position took the pole network position. The incumbent TIM (Italy) has the strongest spectrum position both in low- (+40 MHz vs. WindTre) and mid-band (+52 MHz vs. WindTre). Clearly, it is not a given that having a superior spectrum position also leads to a superior network position. Though 12 out of 16 operators leverage their spectrum superiority compared to their respective competitors.

For operators with the 2nd largest spectrum position, more variation is observed. 7 out of 16 operators end up with the 2nd position as best network (using Umlaut scoring). 3 ended up as best network, and the rest either in 3rd or 4th position. The reason is that often the difference between 2nd and 3rd spectrum rank position is not per see considerable and therefor, other effects, such as several sites, better antenna technologies, and/or better engineering team, are more likely to be decisive factors.

Nevertheless, the total spectrum is a strong predictor for having the best cellular network and winning the best network award (by Umlaut).

As I have collected quite a rich dataset for mobile operators in Western Europe, it may also be possible to model the expected ranking of operators in a given market. Maybe even reasonably predict an Umlaut score (Hakan, don’t worry, I am not quite there … yet!). This said, while the dataset comprises 58+ operators across 16 markets, more data would be required to increase the confidence in benchmark predictions (if that is what one would like to do). Particular to predict absolute benchmark scores (e.g., voice, data, and crowd) as compiled by Umlaut. Speed benchmarks, ala what Ookla’s provides, are (much) easier to predict with much less sophistication (IMO).

Here I will just show my little toy model using the following rank data (using Jupyter R);

The rank dataset set has 64 rows representing rank data and 5 columns containing (1) performance rank (perf_rank, the response), (2) total spectrum rank (spec_rank, predictor), (3) low-band spectrum rank (lo_spec_rank, predictor), (4) high-band spectrum rank (hi_spec_rank, predictor) and (5) Hz-per-customer rank (hz_cust_rank, predictor).

Concerning the predictor (or feature) Hz-per-customer, I am tracking all cellular operators’ so-called spectrum-overhead, which indicates how much Hz can be assigned to a customer (obviously an over-simplification but nevertheless an indicator). Rank 1 means that there is a significant overhead. That is, we have a lot of spectral capacity per customer. Rank 4 has the opposite meaning: the spectral overhead is small, and we have less spectral capacity per customer. It is good to remember that this particular feature is usually dynamic unless the spectrum situation changes for a given cellular operator (e.g., like traffic and customers may grow).

A (very) simple illustration of the “toy model” is shown below, choosing only low-band and high-band ranks as relevant predictors. Almost 60% of the network-benchmark rank can be explained by the low- and high-band ranks.

The model can, of course, be enriched by including more features, such as effective antenna-capability, Hz-per-Customer, Hz-per-Byte, Coverage KPI, Incident rates, Equipment Aging, Supplier, investment level (over last 2 – 3 years), etc… Given the ongoing debate of the importance of supplier to best networks (and their associated awards), I do not find a particularly strong correlation between RAN (incl. antenna) supplier, network performance, and benchmark rank. The total amount of deployed spectrum is a more important predictor. Of course, given the network performance formula above, if an antenna deployment delivers more effective spectral efficiency (or antenna “boost”) than competitors, it will increase the overall network quality for that operator. However, such an operator would still need to overcompensate the potential lack of spectrum compared to a spectrum-superior competitor.

END THOUGHTS.

Having the best cellular network in a market is something to be very proud of. Winning best network awards is obviously great for an operator and its employees. However, it should really mean that the customers of that best network operator also get the best cellular experience compared to any other operator in that market. A superior customer experience is key.

Firstly, the essential driver (enabler) for best network or network leadership is having a superior spectrum position. In low-band, mid-band, and longer-term also in high-band (e.g., mmWave spectrum). The second is having a good coverage footprint across your market. Compared to competitors, a superior spectrum portfolio could even be with fewer cell sites than a competitor with an inferior spectrum position (forced to densify earlier due to spectral capacity limitations as traffic increases). For a spectrum laggard, building more cell sites is costly (i.e., Capex, Opex, and Time) to attempt to improve or match a superior spectrum competitor. Thirdly, having superior antenna technology deployed is essential. It is also a relatively “easy” way to catch up with a superior competitor, at least in the case of relative minor spectrum position differences. Compared to buying additional spectrum (assuming such is available when you need it) or building out a substantial amount of new cell sites to equalize a cellular performance difference, investing into the best (or better or good-enough-to-win) antenna technology, particular for a spectrum laggard, seems to be the best strategy. Economically, relative to the other two options, and operationally, as time-to-catch-up can be relatively short.

After all, this has been said and done, a superior cellular spectrum portfolio is one of the best predictors for having the best network and even winning the best network award.

Economically, it could imply that a spectrum-superior operator, depending on the spectrum distance to the next-best spectrum position in a given market, may not need to invest in the same level of antenna technology as an inferior operator or could delay such investments to a more opportune moment. This could be important, particularly as advanced antenna development is still at its “toddler” state, and more innovative, powerful (and economical) solutions are expected over the next few years. Though, for operators with relatively minor spectrum differences, the battle will be via the advancement of antenna technology and further cell site sectorization (as opposed to building new sites).

ACKNOWLEDGEMENT.

I greatly acknowledge my wife, Eva Varadi, for her support, patience, and understanding during the creative process of writing this Blog. Also, many of my Deutsche Telekom AG and Industry colleagues, in general, have in countless ways contributed to my thinking and ideas leading to this little Blog. Again, I would like to draw attention to Petr Ledl and his super-competent team in Deutsche Telekom’s Group Research & Trials. Thank you so much for being a constant inspiration and always being available to talk antennas and cellular tech in general.

FURTHER READINGS.

Spectrum Monitoring, “Global Mobile Frequencies Database”, the last update on the database was May 2021. You have a limited amount of free inquiries before you will have to pay an affordable fee for access.

Umlaut, “Umlaut Benchmarking” is an important resources for mobile (and fixed) network benchmarks across the world. The umlaut benchmarking methodology is the de-facto industry standard today and applied in more than 120 countries measuring over 200 mobile networks worldwide. I have also made use of the associated Connect Testlab resouce; www.connect-testlab.com. Most network benchmark data goes back to at least 2017. The Umlaut benchmark is based on in-country drive test for voice and data as well as crowd sourced data. It is by a very big margin The cellular network benchmark to use for ranking cellular operators (imo).

Speedtest (Ookla), “Global Index”, most recent data is Q3, 2021. There are three Western European markets that I have not found any Umlaut (or P3 prior to 2020) benchmarks for; Denmark, France and Norway. For those markets I have (regrettably) had to use Ookla data which is clearly not as rich as Umlaut (at least for public domain data).

5G Economics – The Numbers (Appendix X).

100% COVERAGE.

100% 5G coverage is not going to happen with 30 – 300 GHz millimeter-wave frequencies alone.

The “NGMN 5G white paper” , which I will in the subsequent parts refer to as the 5G vision paper, require the 5G coverage to be 100%.

At 100% cellular coverage it becomes somewhat academic whether we talk about population coverage or geographical (area) coverage. The best way to make sure you cover 100% of population is covering 100% of the geography. Of course if you cover 100% of the geography, you are “reasonably” ensured to cover 100% of the population.

While it is theoretically possible to cover 100% (or very near to) of population without covering 100% of the geography, it might be instructive to think why 100% geographical coverage could be a useful target in 5G;

  1. Network-augmented driving and support for varous degrees of autonomous driving would require all roads to be covered (however small).
  2. Internet of Things (IoT) Sensors and Actuators are likely going to be of use also in rural areas (e.g., agriculture, forestation, security, waterways, railways, traffic lights, speed-detectors, villages..) and would require a network to connect to.
  3. Given many users personal area IoT networks (e.g., fitness & health monitors, location detection, smart-devices in general) ubiquitous becomes essential.
  4. Internet of flying things (e.g., drones) are also likely to benefit from 100% area and aerial coverage.

However, many countries remain lacking in comprehensive geographical coverage. Here is an overview of the situation in EU28 (as of 2015);

For EU28 countries, 14% of all house holds in 2015 still had no LTE coverage. This was approx.30+ million households or equivalent to 70+ million citizens without LTE coverage. The 14% might seem benign. However, it covers a Rural neglect of 64% of households not having LTE coverage. One of the core reasons for the lack of rural (population and household) coverage is mainly an economic one. Due to the relative low number of population covered per rural site and compounded by affordability issues for the rural population, overall rural sites tend to have low or no profitability. Network sharing can however improve the rural site profitability as site-related costs are shared.

From an area coverage perspective, the 64% of rural households in EU28 not having LTE coverage is likely to amount to a sizable lack of LTE coverage area. This rural proportion of areas and households are also very likely by far the least profitable to cover for any operator possibly even with very progressive network sharing arrangements.

Fixed broadband, Fiber to the Premises (FTTP) and DOCSIS3.0, lacks further behind that of mobile LTE-based broadband. Maybe not surprisingly from an business economic perspective, in rural areas fixed broadband is largely unavailable across EU28.

The chart below illustrates the variation in lack of broadband coverage across LTE, Fiber to the Premises (FTTP) and DOCSIS3.0 (i.e., Cable) from a total country perspective (i.e., rural areas included in average).

We observe that most countries have very far to go on fixed broadband provisioning (i.e., FTTP and DOCSIS3.0) and even on LTE coverage lacks complete coverage. The rural coverage view (not shown here) would be substantially worse than the above Total view.

The 5G ambition is to cover 100% of all population and households. Due to the demographics of how rural households (and populations) are spread, it is also likely that fairly large geographical areas would need to be covered in order to come true on the 100% ambition.

It would appear that bridging this lack of broadband coverage would be best served by a cellular-based technology. Given the fairly low population density in such areas relative higher average service quality (i.e., broadband) could be delivered as long as the cell range is optimized and sufficient spectrum at a relative low carrier frequency (< 1 GHz) would be available. It should be remembered that the super-high 5G 1 – 10 Gbps performance cannot be expected in rural areas. Due to the lower carrier frequency range need to provide economic rural coverage both advanced antenna systems and very large bandwidth (e.g., such as found in the mm-frequency range)  would not be available to those areas. Thus limiting the capacity and peak performance possible even with 5G.

I would suspect that irrespective of the 100% ambition, telecom providers would be challenged by the economics of cellular deployment and traffic distribution. Rural areas really sucks in profitability, even in fairly aggressive sharing scenarios. Although multi-party (more than 2) sharing might be a way to minimize the profitability burden on deep rural coverage.

The above chart shows the relationship between traffic distribution and sites. As a rule of thumb 50% of revenue is typically generated by 10% of all sites (i.e., in a normal legacy mobile network) and approx. 50% of (rural) sites share roughly 10% of the revenue. Note: in emerging markets the distribution is somewhat steeper as less comprehensive rural coverage typically exist. (Source: The ABC of Network Sharing – The Fundamentals.).

Irrespective of my relative pessimism of the wider coverage utility and economics of millimeter-wave (mm-wave) based coverage, there shall be no doubt that mm-wave coverage will be essential for smaller and smallest cell coverage where due to density of users or applications will require extreme (in comparison to today’s demand) data speeds and capacities. Millimeter-wave coverage-based architectures offer very attractive / advanced antenna solutions that further will allow for increased spectral efficiency and throughput. Also the possibility of using mm-wave point to multipoint connectivity as last mile replacement for fiber appears very attractive in rural and sub-urban clutters (and possible beyond if the cost of the electronics drop according the expeced huge increase in demand for such). This last point however is in my opinion independent of 5G as Facebook with their Terragraph development have shown (i.e., 60 GHz WiGig-based system). A great account for mm-wave wireless communications systems  can be found in T.S. Rappaport et al.’s book “Millimeter Wave Wireless Communications” which not only comprises the benefits of mm-wave systems but also provides an account for the challenges. It should be noted that this topic is still a very active (and interesting) research area that is relative far away from having reached maturity.

In order to provide 100% 5G coverage for the mass market of people & things, we need to engage the traditional cellular frequency bands from 600 MHz to 3 GHz.

1 – 10 Gbps PEAK DATA RATE PER USER.

Getting a Giga bit per second speed is going to require a lot of frequency bandwidth, highly advanced antenna systems and lots of additional cells. And that is likely going to lead to a (very) costly 5G deployment. Irrespective of the anticipated reduced unit cost or relative cost per Byte or bit-per-second.

At 1 Gbps it would take approx. 16 seconds to download a 2 GB SD movie. It would take less than a minute for the HD version (i.e., at 10 Gbps it just gets better;-). Say you have a 16GB smartphone, you loose maybe up to 20+% for the OS, leaving around 13GB for things to download. With 1Gbps it would take less than 2 minutes to fill up your smartphones storage (assuming you haven’t run out of credit on your data plan or reached your data ceiling before then … of course unless you happen to be a customer of T-Mobile US in which case you can binge on = you have no problems!).

The biggest share of broadband usage comes from video streaming which takes up 60% to 80% of all volumetric traffic pending country (i.e., LTE terminal penetration dependent). Providing higher speed to your customer than is required by the applied video streaming technology and smartphone or tablet display being used, seems somewhat futile to aim for. The Table below provides an overview of streaming standards, their optimal speeds and typical viewing distance for optimal experience;

Source: 5G Economics – An Introduction (Chapter 1).

So … 1Gbps could be cool … if we deliver 32K video to our customers end device, i.e., 750 – 1600 Mbps optimal data rate. Though it is hard to see customers benefiting from this performance boost given current smartphone or tablet display sizes. The screen size really have to be ridiculously large to truly benefit from this kind of resolution. Of course Star Trek-like full emersion (i.e., holodeck) scenarios would arguably require a lot (=understatement) bandwidth and even more (=beyond understatement) computing power … though such would scenario appears unlikely to be coming out of cellular devices (even in Star Trek).

1 Gbps fixed broadband plans have started to sell across Europe. Typically on Fiber networks although also on DOCSIS3.1 (10Gbps DS/1 Gbps US) networks as well in a few places. It will only be a matter of time before we see 10 Gbps fixed broadband plans being offered to consumers. Irrespective of compelling use cases might be lacking it might at least give you the bragging rights of having the biggest.

From European Commissions “Europe’s Digital Progress Report 2016”,  22 % of European homes subscribe to fast broadband access of at least 30 Mbps. An estimated 8% of European households subscribe to broadband plans of at least 100 Mbps. It is worth noticing that this is not a problem with coverage as according with the EC’s “Digital Progress Report” around 70% of all homes are covered with at least 30 Mbps and ca. 50% are covered with speeds exceeding 100 Mbps.

The chart below illustrates the broadband speed coverage in EU28;

Even if 1Gbps fixed broadband plans are being offered, still majority of European homes are at speeds below the 100 Mbps. Possible suggesting that affordability and household economics plays a role as well as the basic perceived need for speed might not (yet?) be much beyond 30 Mbps?

Most aggregation and core transport networks are designed, planned, built and operated on a assumption of dominantly customer demand of lower than 100 Mbps packages. As 1Gbps and 10 Gbps gets commercial traction, substantial upgrades are require in aggregation, core transport and last but not least possible also on an access level (to design shorter paths). It is highly likely distances between access, aggregation and core transport elements are too long to support these much higher data rates leading to very substantial redesigns and physical work to support this push to substantial higher throughputs.

Most telecommunications companies will require very substantial investments in their existing transport networks all the way from access to aggregation through the optical core switching networks, out into the world wide web of internet to support 1Gbps to 10 Gbps. Optical switching cards needs to be substantially upgraded, legacy IP/MPLS architectures might no longer work very well (i.e., scale & complexity issue).

Most analysts today believe that incumbent fixed & mobile broadband telecommunications companies with a reasonable modernized transport network are best positioned for 5G compared to mobile-only operators or fixed-mobile incumbents with an aging transport infrastructure.

What about the state of LTE speeds across Europe? OpenSignal recurrently reports on the State of LTE, the following summarizes LTE speeds in Mbps as of June 2017 for EU28 (with the exception of a few countries not included in the OpenSignal dataset);

The OpenSignal measurements are based on more than half a million devices, almost 20 billion measurements over the period of the 3 first month of 2017.

The 5G speed ambition is by todays standards 10 to 30+ times away from present 2016/2017 household fixed broadband demand or the reality of provided LTE speeds.

Let us look at cellular spectral efficiency to be expected from 5G. Using the well known framework;

In essence, I can provide very high data rates in bits per second by providing a lot of frequency bandwidth B, use the most spectrally efficient technologies maximizing η, and/or add as many cells N that my economics allow for.

In the following I rely largely on Jonathan Rodriquez great book on “Fundamentals of 5G Mobile Networks” as a source of inspiration.

The average spectral efficiency is expected to be coming out in the order of 10 Mbps/MHz/cell using advanced receiver architectures, multi-antenna, multi-cell transmission and corporation. So pretty much all the high tech goodies we have in the tool box is being put to use of squeezing out as many bits per spectral Hz available and in a sustainable matter. Under very ideal Signal to Noise Ratio conditions, massive antenna arrays of up to 64 antenna elements (i.e., an optimum) seems to indicate that 50+ Mbps/MHz/Cell might be feasible in peak.

So for a spectral efficiency of 10 Mbps/MHz/cell and a demanded 1 Gbps data rate we would need 100 MHz frequency bandwidth per cell (i.e., using the above formula). Under very ideal conditions and relative large antenna arrays this might lead to a spectral requirement of only 20 MHz at 50 Mbps/MHz/Cell. Obviously, for 10 Gbps data rate we would require 1,000 MHz frequency bandwidth (1 GHz!) per cell at an average spectral efficiency of 10 Mbps/MHz/cell.

The spectral efficiency assumed for 5G heavily depends on successful deployment of many-antenna segment arrays (e.g., Massive MiMo, beam-forming antennas, …). Such fairly complex antenna deployment scenarios work best at higher frequencies, typically above 2GHz. Also such antenna systems works better at TDD than FDD with some margin on spectral efficiency. These advanced antenna solutions works perfectly  in the millimeter wave range (i.e., ca. 30 – 300 GHz) where the antenna segments are much smaller and antennas can be made fairly (very) compact (note: resonance frequency of the antenna proportional to half the wavelength with is inverse proportional to the carrier frequency and thus higher frequencies need smaller material dimension to operate).

Below 2 GHz higher-order MiMo becomes increasingly impractical and the spectral efficiency regress to the limitation of a simple single-path antenna. Substantially lower than what can be achieved at much high frequencies with for example massive-MiMo.

So for the 1Gbps to 10 Gbps data rates to work out we have the following relative simple rationale;

  • High data rates require a lot of frequency bandwidth (>100 MHz to several GHz per channel).
  • Lots of frequency bandwidth are increasingly easier to find at high and very high carrier frequencies (i.e., why millimeter wave frequency band between 30 – 300 GHz is so appealing).
  • High and very high carrier frequencies results in small, smaller and smallest cells with very high bits per second per unit area (i.e., the area is very small!).
  • High and very high carrier frequency allows me to get the most out of higher order MiMo antennas (i.e., with lots of antenna elements),
  • Due to fairly limited cell range, I boost my overall capacity by adding many smallest cells (i.e., at the highest frequencies).

We need to watch out for the small cell densification which tends not to scale very well economically. The scaling becomes a particular problem when we need hundreds of thousands of such small cells as it is expected in most 5G deployment scenarios (i.e., particular driven by the x1000 traffic increase). The advanced antenna systems required (including the computation resources needed) to max out on spectral efficiency are likely going to be one of the major causes of breaking the economical scaling. Although there are many other CapEx and OpEx scaling factors to be concerned about for small cell deployment at scale.

Further, for mass market 5G coverage, as opposed to hot traffic zones or indoor solutions, lower carrier frequencies are needed. These will tend to be in the usual cellular range we know from our legacy cellular communications systems today (e.g., 600 MHz – 2.1 GHz). It should not be expected that 5G spectral efficiency will gain much above what is already possible with LTE and LTE-advanced at this legacy cellular frequency range. Sheer bandwidth accumulation (multi-frequency carrier aggregation) and increased site density is for the lower frequency range a more likely 5G path. Of course mass market 5G customers will benefit from faster reaction times (i.e., lower latencies), higher availability, more advanced & higher performing services arising from the very substantial changes expected in transport networks and data centers with the introduction of 5G.

Last but not least to this story … 80% and above of all mobile broadband customers usage, data as well as voice, happens in very few cells (e.g., 3!) … representing their Home and Work.

Source: Slideshare presentation by Dr. Kim “Capacity planning in mobile data networks experiencing exponential growth in demand.”

As most of the mobile cellular traffic happen at the home and at work (i.e., thus in most cases indoor) there are many ways to support such traffic without being concerned about the limitation of cell ranges.

The giga bit per second cellular service is NOT a service for the mass market, at least not in its macro-cellular form.

≤ 1 ms IN ROUND-TRIP DELAY.

A total round-trip delay of 1 or less millisecond is very much attuned to niche service. But a niche service that nevertheless could be very costly for all to implement.

I am not going to address this topic too much here. It has to a great extend been addressed almost to ad nauseam in 5G Economics – An Introduction (Chapter 1) and 5G Economics – The Tactile Internet (Chapter 2). I think this particular aspect of 5G is being over-hyped in comparison to how important it ultimately will turn out to be from a return on investment perspective.

Speed of light travels ca. 300 km per millisecond (ms) in vacuum and approx. 210 km per ms in fiber (some material dependency here). Lately engineers have gotten really excited about the speed of light not being fast enough and have made a lot of heavy thinking abou edge this and that (e.g., computing, cloud, cloudlets, CDNs,, etc…). This said it is certainly true that most modern data centers have not been build taking too much into account that speed of light might become insufficient. And should there really be a great business case of sub-millisecond total (i.e., including the application layer) roundtrip time scales edge computing resources would be required a lot closer to customers than what is the case today.

It is common to use delay, round-trip time or round-trip delay, or latency as meaning the same thing. Though it is always cool to make sure people really talk about the same thing by confirming that it is indeed a round-trip rather than single path. Also to be clear it is worthwhile to check that all people around the table talk about delay at the same place in the OSI stack or  network path or whatever reference point agreed to be used.

In the context of  the 5G vision paper it is emphasized that specified round-trip time is based on the application layer (i.e., OSI model) as reference point. It is certainly the most meaningful measure of user experience. This is defined as the End-2-End (E2E) Latency metric and measure the complete delay traversing the OSI stack from physical layer all the way up through network layer to the top application layer, down again, between source and destination including acknowledgement of a successful data packet delivery.

The 5G system shall provide 10 ms E2E latency in general and 1 ms E2E latency for use cases requiring extremely low latency.

The 5G vision paper states “Note these latency targets assume the application layer processing time is negligible to the delay introduced by transport and switching.” (Section 4.1.3 page 26 in “NGMN 5G White paper”).

In my opinion it is a very substantial mouthful to assume that the Application Layer (actually what is above the Network Layer) will not contribute significantly to the overall latency. Certainly for many applications residing outside the operators network borders, in the world wide web, we can expect a very substantial delay (i.e., even in comparison with 10 ms). Again this aspect was also addressed in my two first chapters.

Very substantial investments are likely needed to meet E2E delays envisioned in 5G. In fact the cost of improving latencies gets prohibitively more expensive as the target is lowered. The overall cost of design for 10 ms would be a lot less costly than designing for 1 ms or lower. The network design challenge if 1 millisecond or below is required, is that it might not matter that this is only a “service” needed in very special situations, overall the network would have to be designed for the strictest denominator.

Moreover, if remedies needs to be found to mitigate likely delays above the Network Layer, distance and insufficient speed of light might be the least of worries to get this ambition nailed (even at the 10 ms target). Of course if all applications are moved inside operator’s networked premises with simpler transport paths (and yes shorter effective distances) and distributed across a hierarchical cloud (edge, frontend, backend, etc..), the assumption of negligible delay in layers above the Network Layer might become much more likely. However, it does sound a lot like America Online walled garden fast forward to the past kind of paradigm.

So with 1 ms E2E delay … yeah yeah … “play it again Sam” … relevant applications clearly need to be inside network boundary and being optimized for processing speed or silly & simple (i.e., negligible delay above the Network Layer), no queuing delay (to the extend of being in-efficiency?), near-instantaneous transmission (i.e., negligible transmission delay) and distances likely below tenth of km (i.e., very short propagation delay).

When the speed of light is too slow there are few economic options to solve that challenge.

≥ 10,000 Gbps / Km2 DATA DENSITY.

The data density is maybe not the most sensible measure around. If taken too serious could lead to hyper-ultra dense smallest network deployments.

This has always been a fun one in my opinion. It can be a meaningful design metric or completely meaningless.

There is of course nothing particular challenging in getting a very high throughput density if an area is small enough. If I have a cellular range of few tens of meters, say 20 meters, then my cell area is smaller than 1/1000 of a km2. If I have 620 MHz bandwidth aggregated between 28 GHz and 39 GHz (i.e., both in the millimeter wave band) with a 10 Mbps/MHz/Cell, I could support 6,200 Gbps/km2. That’s almost 3 Petabyte in an hour or 10 years of 24/7 binge watching of HD videos. Note given my spectral efficiency is based on an average value, it is likely that I could achieve substantially more bandwidth density and in peaks closer to the 10,000 Gbps/km2 … easily.

Pretty Awesome Wow!

The basic; a Terabit equals 1024 Gigabits (but I tend to ignore that last 24 … sorry I am not).

With a traffic density of ca. 10,000 Gbps per km2, one would expect to have between 1,000 (@ 10 Gbps peak) to 10,000 (@ 1 Gbps peak) concurrent users per square km.

At 10 Mbps/MHz/Cell one would expect to have a 1,000 Cell-GHz/km2. Assume that we would have 1 GHz bandwidth (i.e., somewhere in the 30 – 300 GHz mm-wave range), one would need 1,000 cells per km2. On average with a cell range of about 20 meters (smaller to smallest … I guess what Nokia would call an Hyper-Ultra-Dense Network;-). Thus each cell would minimum have between 1 to 10 concurrent users.

Just as a reminder! 1 minutes at 1 Gbps corresponds to 7.5 GB. A bit more than what you need for a 80 minute HD (i.e., 720pp) full movie stream … in 1 minutes. So with your (almost) personal smallest cell what about the remaining 59 minutes? Seems somewhat wasteful at least until kingdom come (alas maybe sooner than that).

It would appear that the very high 5G data density target could result in very in-efficient networks from a utilization perspective.

≥ 1 MN / Km2 DEVICE DENSITY.

One million 5G devices per square kilometer appears to be far far out in a future where one would expect us to be talking about 7G or even higher Gs.

1 Million devices seems like a lot and certainly per km2. It is 1 device per square meter on average. A 20 meter cell-range smallest cell would contain ca. 1,200 devices.

To give this number perspective lets compare it with one of my favorite South-East Asian cities. The city with one of the highest population densities around, Manila (Philippines). Manila has more than 40 thousand people per square km. Thus in Manila this would mean that we would have about 24 devices per person or 100+ per household per km2. Overall, in Manila we would then expect approx. 40 million devices spread across the city (i.e., Manila has ca. 1.8 Million inhabitants over an area of 43 km2. Philippines has a population of approx. 100 Million).

Just for the curious, it is possible to find other more populated areas in the world. However, these highly dense areas tends to be over relative smaller surface areas, often much smaller than a square kilometer and with relative few people. For example Fadiouth Island in Dakar have a surface area of 0.15 km2 and 9,000 inhabitants making it one of the most pop densest areas in the world (i.e., 60,000 pop per km2).

I hope I made my case! A million devices per km2 is a big number.

Let us look at it from a forecasting perspective. Just to see whether we are possibly getting close to this 5G ambition number.

IHS forecasts 30.5 Billion installed devices by 2020, IDC is also believes it to be around 30 Billion by 2020. Machina Research is less bullish and projects 27 Billion by 2025 (IHS expects that number to be 75.4 Billion) but this forecast is from 2013. Irrespective, we are obviously in the league of very big numbers. By the way 5G IoT if at all considered is only a tiny fraction of the overall projected IoT numbers (e.g., Machine Research expects 10 Million 5G IoT connections by 2024 …that is extremely small numbers in comparison to the overall IoT projections).

A consensus number for 2020 appears to be 30±5 Billion IoT devices with lower numbers based on 2015 forecasts and higher numbers typically from 2016.

To break this number down to something that could be more meaningful than just being Big and impressive, let just establish a couple of worldish numbers that can help us with this;

  • 2020 population expected to be around 7.8 Billion compared to 2016 7.4 Billion.
  • Global pop per HH is ~3.5 (average number!) which might be marginally lower in 2020. Urban populations tend to have less pop per households ca. 3.0. Urban populations in so-called developed countries are having a pop per HH of ca. 2.4.
  • ca. 55% of world population lives in Urban areas. This will be higher by 2020.
  • Less than 20% of world population lives in developed countries (based on HDI). This is a 2016 estimate and will be higher by 2020.
  • World surface area is 510 Million km2 (including water).
  • of which ca. 150 million km2 is land area
  • of which ca. 75 million km2 is habitable.
  • of which 3% is an upper limit estimate of earth surface area covered by urban development, i.e., 15.3 Million km2.
  • of which approx. 1.7 Million km2 comprises developed regions urban areas.
  • ca. 37% of all land-based area is agricultural land.

Using 30 Billion IoT devices by 2020 is equivalent to;

  • ca. 4 IoT per world population.
  • ca. 14 IoT per world households.
  • ca. 200 IoT per km2 of all land-based surface area.
  • ca. 2,000 IoT per km2 of all urban developed surface area.

If we limit IoT’s in 2020 to developed countries, which wrongly or rightly exclude China, India and larger parts of Latin America, we get the following by 2020;

  • ca. 20 IoT per developed country population.
  • ca. 50 IoT per developed country households.
  • ca. 18,000 IoT per km2 developed country urbanized areas.

Given that it would make sense to include larger areas and population of both China, India and Latin America, the above developed country numbers are bound to be (a lot) lower per Pop, HH and km2. If we include agricultural land the number of IoTs will go down per km2.

So far far away from a Million IoT per km2.

What about parking spaces, for sure IoT will add up when we consider parking spaces!? … Right? Well in Europe you will find that most big cities will have between 50 to 200 (public) parking spaces per square kilometer (e.g., ca. 67 per km2 for Berlin and 160 per km2 in Greater Copenhagen). Aha not really making up to the Million IoT per km2 … what about cars?

In EU28 there are approx. 256 Million passenger cars (2015 data) over a population of ca. 510 Million pops (or ca. 213 million households). So a bit more than 1 passenger car per household on EU28 average. In Eu28 approx. 75+% lives in urban area which comprises ca. 150 thousand square kilometers (i.e., 3.8% of EU28’s 4 Million km2). So one would expect little more (if not a little less) than 1,300 passenger cars per km2. You may say … aha but it is not fair … you don’t include motor vehicles that are used for work … well that is an exercise for you (too convince yourself why that doesn’t really matter too much and with my royal rounding up numbers maybe is already accounted for). Also consider that many EU28 major cities with good public transportation are having significantly less cars per household or population than the average would allude to.

Surely, public street light will make it through? Nope! Typical bigger modern developed country city will have on average approx. 85 street lights per km2, although it varies from 0 to 1,000+. Light bulbs per residential household (from a 2012 study of the US) ranges from 50 to 80+. In developed countries we have roughly 1,000 households per km2 and thus we would expect between 50 thousand to 80+ thousand lightbulbs per km2. Shops and business would add some additions to this number.

With a cumulated annual growth rate of ca. 22% it would take 20 years (from 2020) to reach a Million IoT devices per km2 if we will have 20 thousand per km2 by 2020. With a 30% CAGR it would still take 15 years (from 2020) to reach a Million IoT per km2.

The current IoT projections of 30 Billion IoT devices in operation by 2020 does not appear to be unrealistic when broken down on a household or population level in developed areas (even less ambitious on a worldwide level). The 18,000 IoT per km2 of developed urban surface area by 2020 does appear somewhat ambitious. However, if we would include agricultural land the number would become possible a more reasonable.

If you include street crossings, traffic radars, city-based video monitoring (e.g., London has approx. 300 per km2, Hong Kong ca. 200 per km2), city-based traffic sensors, environmental sensors, etc.. you are going to get to sizable numbers.

However, 18,000 per km2 in urban areas appears somewhat of a challenge. Getting to 1 Million per km2 … hmmm … we will see around 2035 to 2040 (I have added an internet reminder for a check-in by 2035).

Maybe the 1 Million Devices per km2 ambition is not one of the most important 5G design criteria’s for the short term (i.e., next 10 – 20 years).

Oh and most IoT forecasts from the period 2015 – 2016 does not really include 5G IoT devices in particular. The chart below illustrates Machina Research IoT forecast for 2024 (from August 2015). In a more recent forecast from 2016, Machine Research predict that by 2024 there would be ca. 10 million 5G IoT connections or 0.04% of the total number of forecasted connections;

The winner is … IoTs using WiFi or other short range communications protocols. Obviously, the cynic in me (mea culpa) would say that a mm-wave based 5G connections can also be characterized as short range … so there might be a very interesting replacement market there for 5G IoT … maybe? 😉

Expectations to 5G-based IoT does not appear to be very impressive at least over the next 10 years and possible beyond.

The un-importance of 5G IoT should not be a great surprise given most 5G deployment scenarios are focused on millimeter-wave smallest 5G cell coverage which is not good for comprehensive coverage of  IoT devices not being limited to those very special 5G coverage situations being thought about today.

Only operators focusing on comprehensive 5G coverage re-purposing lower carrier frequency bands (i.e., 1 GHz and lower) can possible expect to gain a reasonable (as opposed to niche) 5G IoT business. T-Mobile US with their 600 MHz  5G strategy might very well be uniquely positions for taking a large share of future proof IoT business across USA. Though they are also pretty uniquely position for NB-IoT with their comprehensive 700MHz LTE coverage.

For 5G IoT to be meaningful (at scale) the conventional macro-cellular networks needs to be in play for 5G coverage .,, certainly 100% 5G coverage will be a requirement. Although, even with 5G there maybe 100s of Billion of non-5G IoT devices that require coverage and management.

≤ 500 km/h SERVICE SUPPORT.

Sure why not?  but why not faster than that? At hyperloop or commercial passenger airplane speeds for example?

Before we get all excited about Gbps speeds at 500 km/h, it should be clear that the 5G vision paper only proposed speeds between 10 Mbps up-to 50 Mbps (actually it is allowed to regress down to 50 kilo bits per second). With 200 Mbps for broadcast like services.

So in general, this is a pretty reasonable requirement. Maybe with the 200 Mbps for broadcasting services being somewhat head scratching unless the vehicle is one big 16K screen. Although the users proximity to such a screen does not guaranty an ideal 16K viewing experience to say the least.

What moves so fast?

The fastest train today is tracking at ca. 435 km/h (Shanghai Maglev, China).

Typical cruising airspeed for a long-distance commercial passenger aircraft is approx. 900 km/h. So we might not be able to provide the best 5G experience in commercial passenger aircrafts … unless we solve that with an in-plane communications system rather than trying to provide Gbps speed by external coverage means.

Why take a plane when you can jump on the local Hyperloop? The proposed Hyperloop should track at an average speed of around 970 km/h (faster or similar speeds as commercial passengers aircrafts), with a top speed of 1,200 km/h. So if you happen to be in between LA and San Francisco in 2020+ you might not be able to get the best 5G service possible … what a bummer! This is clearly an area where the vision did not look far enough.

Providing services to moving things at a relative fast speed does require a reasonable good coverage. Whether it being train track, hyperloop tunnel or ground to air coverage of commercial passenger aircraft, new coverage solutions would need to be deployed. Or alternative in-vehicular coverage solutions providing a perception of 5G experience might be an alternative that could turn out to be more economical.

The speed requirement is a very reasonable one particular for train coverage.

50% TOTAL NETWORK ENERGY REDUCTION.

If 5G development could come true on this ambition we talk about 10 Billion US Dollars (for the cellular industry). Equivalent to a percentage point on the margin.

There are two aspects of energy efficiency in a cellular based communication system.

  • User equipment that will benefit from longer intervals without charging and thus improve customers experience and overall save energy from less frequently charges.
  • Network infrastructure energy consumption savings will directly positively impact a telecom operators Ebitda.

Energy efficient Smartphones

The first aspect of user equipment is addressed by the 5G vision paper under “4.3 Device Requirements”  sub-section “4.3.3 Device Power Efficiency”; Battery life shall be significantly increased: at least 3 days for a smartphone, and up tp 15 years for a low-cost MTC device.” (note: MTC = Machine Type Communications).

Apple’s iPhone 7 battery life (on a full charge) is around 6 hours of constant use with 7 Plus beating that with ca. 3 hours (i.e., total 9 hours). So 3 days will go a long way.

From a recent 2016 survey from Ask Your Target Market on smartphone consumers requirements to battery lifetime and charging times;

  • 64% of smartphone owners said they are at least somewhat satisfied with their phone’s battery life.
  • 92% of smartphone owners said they consider battery life to be an important factor when considering a new smartphone purchase.
  • 66% said they would even pay a bit more for a cell phone that has a longer battery life.

Looking at the mobile smartphone & tablet non-voice consumption it is also clear why battery lifetime and not in-important the charging time matters;

Source: eMarketer, April 2016. While 2016 and 2017 are eMarketer forecasts (why dotted line and red circle!) these do appear well in line with other more recent measurements.

Non-voice smartphone & tablet based usage is expected by now to exceed 4 hours (240 minutes) per day on average for US Adults.

That longer battery life-times are needed among smartphone consumers is clear from sales figures and anticipated sales growth of smartphone power-banks (or battery chargers) boosting the life-time with several more hours.

It is however unclear whether the 3 extra days of a 5G smartphone battery life-time is supposed to be under active usage conditions or just in idle mode. Obviously in order to matter materially to the consumer one would expect this vision to apply to active usage (i.e., 4+ hours a day at 100s of Mbps – 1Gbps operations).

Energy efficient network infrastructure.

The 5G vision paper defines energy efficiency as number of bits that can be transmitted over the telecom infrastructure per Joule of Energy.

The total energy cost, i.e., operational expense (OpEx), of telecommunications network can be considerable. Despite our mobile access technologies having become more energy efficient with each generation, the total OpEx of energy attributed to the network infrastructure has increased over the last 10 years in general. The growth in telco infrastructure related energy consumption has been driven by the consumer demand for broadband services in mobile and fixed including incredible increase in data center computing and storage requirements.

In general power consumption OpEx share of total technology cost amounts to 8% to 15% (i.e., for Telcos without heavy reliance of diesel). The general assumption is that with regular modernization, energy efficiency gain in newer electronics can keep growth in energy consumption to a minimum compensating for increased broadband and computing demand.

Note: Technology Opex (including NT & IT) on average lays between 18% to 25% of total corporate Telco Opex. Out of the Technology Opex between 8% to 15% (max) can typically be attributed to telco infrastructure energy consumption. The access & aggregation contribution to the energy cost typically would towards 80% plus. Data centers are expected to increasingly contribute to the power consumption and cost as well. Deep diving into the access equipment power consumption, ca. 60% can be attributed to rectifiers and amplifiers, 15% by the DC power system & miscellaneous and another 25% by cooling.

5G vision paper is very bullish in their requirement to reduce the total energy and its associated cost; it is stated “5G should support a 1,000 times traffic increase in the next 10 years timeframe, with an energy consumption by the whole network of only half that typically consumed by today’s networks. This leads to the requirement of an energy efficiency of x2,000 in the next 10 years timeframe.” (sub-section “4.6.2 Energy Efficiency” NGMN 5G White Paper).

This requirement would mean that in a pure 5G world (i.e., all traffic on 5G), the power consumption arising from the cellular network would be 50% of what is consumed todayIn 2016 terms the Mobile-based Opex saving would be in the order of 5 Billion US$ to 10+ Billion US$ annually. This would be equivalent to 0.5% to 1.1% margin improvement globally (note: using GSMA 2016 Revenue & Growth data and Pyramid Research forecast). If energy price would increase over the next 10 years the saving / benefits would of course be proportionally larger.

As we have seen in the above, it is reasonable to expect a very considerable increase in cell density as the broadband traffic demand increases from peak bandwidth (i.e., 1 – 10 Gbps) and traffic density (i.e., 1 Tbps per km2) expectations.

Depending on the demanded traffic density, spectrum and carrier frequency available for 5G between 100 to 1,000 small cell sites per km2 could be required over the next 10 years. This cell site increase will be required in addition to existing macro-cellular network infrastructure.

Today (in 2017) an operator in EU28-sized country may have between ca. 3,500 to 35,000 cell sites with approx. 50% covering rural areas. Many analysts are expecting that for medium sized countries (e.g., with 3,500 – 10,000 macro cellular sites), operators would eventually have up-to 100,000 small cells under management in addition to their existing macro-cellular sites. Most of those 5G small cells and many of the 5G macro-sites we will have over the next 10 years, are also going to have advanced massive MiMo antenna systems with many active antenna elements per installed base antenna requiring substantial computing to gain maximum performance.

It appears with today’s knowledge extremely challenging (to put it mildly) to envision a 5G network consuming 50% of today’s total energy consumption.

It is highly likely that the 5G radio node electronics in a small cell environment (and maybe also in a macro cellular environment?) will consume less Joules per delivery bit (per second) due to technology advances and less transmitted power required (i.e., its a small or smallest cell). However, this power efficiency technology and network cellular architecture gain can very easily be destroyed by the massive additional demand of small, smaller and smallest cells combined with highly sophisticated antenna systems consuming additional energy for their compute operations to make such systems work. Furthermore, we will see operators increasingly providing sophisticated data center resources network operations as well as for the customers they serve. If the speed of light is insufficient for some services or country geographies, additional edge data centers will be introduced, also leading to an increased energy consumption not present in todays telecom networks. Increased computing and storage demand will also make the absolute efficiency requirement highly challenging.

Will 5G be able to deliver bits (per second) more efficiently … Yes!

Will 5G be able to reduce the overall power consumption of todays telecom networks with 50% … highly unlikely.

In my opinion the industry will have done a pretty good technology job if we can keep the existing energy cost at the level of today (or even allowing for unit price increases over the next 10 years).

The Total power reduction of our telecommunications networks will be one of the most important 5G development tasks as the industry cannot afford a new technology that results in waste amount of incremental absolute cost. Great relative cost doesn’t matter if it results in above and beyond total cost.

≥ 99.999% NETWORK AVAILABILITY & DATA CONNECTION RELIABILITY.

A network availability of 5Ns across all individual network elements and over time correspond to less than a second a day downtime anywhere in the network. Few telecom networks are designed for that today.

5 Nines (5N) is a great aspiration for services and network infrastructures. It also tends to be fairly costly and likely to raise the level of network complexity. Although in the 5G world of heterogeneous networks … well its is already complicated.

5N Network Availability.

From a network and/or service availability perspective it means that over the cause of the day, your service should not experience more than 0.86 seconds of downtime. Across a year the total downtime should not be more than 5 minutes and 16 seconds.

The way 5N Network Availability is define is “The network is available for the targeted communications in 99.999% of the locations  where the network is deployed and 99.999% of the time”. (from “4.4.4 Resilience and High Availability”, NGMN 5G White Paper).

Thus in a 100,000 cell network only 1 cell is allowed experience a downtime and for no longer than less than a second a day.

It should be noted that there are not many networks today that come even close to this kind of requirement. Certainly in countries with frequent long power outages and limited ancillary backup (i.e., battery and/or diesel) this could be a very costly design requirement. Networks relying on weather-sensitive microwave radios for backhaul or for mm-wave frequencies 5G coverage would be required to design in a very substantial amount of redundancy to keep such high geographical & time availability requirements

In general designing a cellular access network for this kind of 5N availability could be fairly to very costly (i.e., Capex could easily run up in several percentage points of Revenue).

One way out from a design perspective is to rely on hierarchical coverage. Thus, for example if a small cell environment is un-available (=down!) the macro-cellular network (or overlay network) continues the service although at a lower service level (i.e., lower or much lower speed compared to the primary service). As also suggested in the vision paper making use of self-healing network features and other real-time measures are expected to further increase the network infrastructure availability. This is also what one may define as Network Resilience.

Nevertheless, the “NGMN 5G White Paper” allows for operators to define the level of network availability appropriate from their own perspective (and budgets I assume).

5N Data Packet Transmission Reliability.

The 5G vision paper, defines Reliability as “… amount of sent data packets successfully delivered to a given destination, within the time constraint required by the targeted service, divided by the total number of sent data packets.”. (“4.4.5 Reliability” in “NGMN 5G White Paper”).

It should be noted that the 5N specification in particular addresses specific use cases or services of which such a reliability is required, e.g., mission critical communications and ultra-low latency service. The 5G allows for a very wide range of reliable data connection. Whether the 5N Reliability requirement will lead to substantial investments or can be managed within the overall 5G design and architectural framework, might depend on the amount of traffic requiring 5Ns.

The 5N data packet transmission reliability target would impose stricter network design. Whether this requirement would result in substantial incremental investment and cost is likely dependent on the current state of existing network infrastructure and its fundamental design.

 

Profitability of the Mobile Business Model … The Rise! & Inevitable Fall?

A Mature & Emerging Market Profitability Analysis … From Past, through Present & to the Future.

  • I dedicate this Blog to David Haszeldine whom has been (and will remain) a true partner when it comes to discussing, thinking and challenging cost structures, corporate excesses and optimizing the Telco profitability.
  • Opex growth & declining revenue growth is the biggest exposure to margin decline & profitability risk for emerging growth markets as well as mature mobile markets.
  • 48 Major Mobile Market’s Revenue & Opex Growth have been analyzed over the period 2007 to 2013 (for some countries from 2003 to 2013). The results have been provided in an easy to compare overview chart.
  • For 23 out of the 48 Mobile Markets, Opex have grown faster than Revenue and poses a substantial risk to Telco profitability in the near & long-term unless Opex will be better managed and controlled.
  • Mobile Profitability Risk is a substantial Emerging Growth Market Problem where cost has grown much faster than the corresponding Revenues.
  • 11 Major Emerging Growth Markets have had an Opex compounded annual growth rate between 2007 to 2013 that was higher than the Revenue Growth substantially squeezing margin and straining EBITDA.
  • On average the compounded annual growth rate of Opex grew 2.2% faster than corresponding Revenue over the period 2007 to 2013. Between 2012 to 2013 Opex grew (on average) 3.7% faster than Revenue.
  • A Market Profit Sustainability Risk Index (based on Bayesian inference) is proposed as a way to provide an overview of mobile markets profitability directions based on their Revenue and Opex growth rates.
  • Statistical Analysis on available data shows that a Mobile Markets Opex level is driven by (1) Population, (2) Customers, (3) Penetration and (4) ARPU. The GDP & Surface Area have only minor and indirect influence on the various markets Opex levels.
  • A profitability framework for understanding individual operators profit dynamics is proposed.
  • It is shown that Profitability can be written as withbeing the margin, with ou and ru being the user dependent OpEx and Revenue (i.e., AOPU and ARPU), of the fixed OpEx divided by the Total Subscriber Market andis the subscriber market share.
  • The proposed operator profitability framework provides a high degree of descriptive power and understanding of individual operators margin dynamics as a function of subscriber market share as well as other important economical drivers.

I have long & frequently been pondering over the mobile industry’s profitability.In particular, I have spend a lot of my time researching the structure & dynamics of profitability and mapping out factors that contributes in both negative & positive ways? My interest is the underlying cost structures and business models that drives the profitability in both good and bad ways. I have met Executives who felt a similar passion for strategizing, optimizing and managing their companies Telco cost structures and thereby profit and I have also met Executives who mainly cared for the Revenue.

Obviously, both Revenue and Cost are important to optimize. This said it is wise to keep in mind the following Cost- structure & Revenue Heuristics;

  • Cost is an almost Certainty once made & Revenues are by nature Uncertain.
  • Cost left Unmanaged will by default Increase over time.
  • Revenue is more likely to Decrease over time than increase.
  • Majority of Cost exist on a different & longer time-scale than Revenue.

In the following I will use EBITDA, which stands for Earnings Before Interest, Taxes, Depreciation and Amortization, as a measure of profitability and EBITDA to Revenue Ratio as a measure of my profit margin or just margin. It should be clear that EBITDA is a proxy of profitability and as such have shortfalls in specific Accounting and P&L Scenarios. Also according with GAAP (General Accepted Accounting Principles) and under IFRS (International Financial Reporting Standards) EBITDA is not a standardized accepted accounting measure. Nevertheless, both EBITDA and EBITDA Margin are widely accepted and used in the mobile industry as a proxy for operational performance and profitability. I am going to assume that for most purposes & examples discussed in this Blog, EBITDA & the corresponding Margin remains sufficiently good measures profitability.

While I am touching upon mobile revenues as an issue for profitability, I am not going to provide much thoughts on how to boost revenues or add new incremental revenues that might compensate from loss of mobile legacy service revenues (i.e., voice, messaging and access). My revenue focus in particular addresses revenue growth on a more generalized level compared to the mobile cost being incurred operating such services in particular and a mobile business in general. For an in-depth and beautiful treatment of mobile revenues past, present and future, I would like to refer to Chetan Sharma’s 2012 paper “Operator’s Dilemma (and Opportunity): The 4th Wave” (note: you can download the paper by following the link in the html article) on mobile revenue dynamics from (1) Voice (1st Revenue or Service Wave), (2) Messaging (2nd Revenue or Service Wave) to todays (3) Access (3rd Revenue Wave) and the commence to what Chetan Sharma defines as the 4th Wave of Revenues (note: think of waves as S-curves describing initial growth spurt, slow down phase, stagnation and eventually decline) which really describes a collection of revenue or service waves (i.e., S-curves) representing a portfolio of Digital Services, such as (a) Connected Home, (b) Connected Car,  (c) Health, (d) Payment, (e) Commerce, (f) Advertising, (g) Cloud Services (h) Enterprise solutions, (i) Identity, Profile & Analysis etc..  I feel confident that adding any Digital Service enabled by Internet-of-Things (IoT) and M2M would be important inclusions to the Digital Services Wave. Given the competition (i.e., Facebook, Google, Amazon, Ebay, etc..) that mobile operators will face entering the 4th Wave of Digital Services Space, in combination with having only national or limited international scale, will make this area a tough challenge to return direct profit on. The inherent limited international or national-only scale appears to be one of the biggest barrier to turn many of the proposed Digital Services, particular with those with strong Social Media Touch Points, into meaningful business opportunities for mobile operators.

This said, I do believe (strongly) that Telecom Operators have very good opportunities for winning Digital Services Battles in areas where their physical infrastructure (including Spectrum & IT Architecture) is an asset and essential for delivering secure, private and reliable services. Local regulation and privacy laws may indeed turn out to be a blessing for Telecom Operators and other national-oriented businesses. The current privacy trend and general consumer suspicion of American-based Global Digital Services / Social Media Enterprises may create new revenue opportunities for national-focused mobile operators as well as for other national-oriented digital businesses. In particular if Telco Operators work together creating Digital Services working across operator’s networks, platforms and beyond (e.g., payment, health, private search, …) rather than walled-garden digital services, they might become very credible alternatives to multi-national offerings. It is highly likely that consumers would be more willing to trust national mobile operator entities with her or his personal data & money (in fact they already do that in many areas) than a multinational social-media corporation. In addition to the above Digital Services, I do expect that Mobile/Telecom Operators and Entertainment Networks (e.g., satellite, cable, IP-based) will increasingly firm up partnerships as well as acquire & merge their businesses & business models. In all effect this is already happening.

For emerging growth markets without extensive and reliable fixed broadband infrastructures, high-quality (& likely higher cost compared to today’s networks!) mobile broadband infrastructures would be essential to drive additional Digital Services and respective revenues as well as for new entertainment business models (other than existing Satellite TV). Anyway, Chetan captures these Digital Services (or 4th Wave) revenue streams very nicely and I recommend very much to read his articles in general (i.e., including “Mobile 4th Wave: The Evolution of the Next Trillion Dollars” which is the 2nd “4th Wave” article).

Back to mobile profitability and how to ensure that the mobile business model doesn’t breakdown as revenue growth starts to slow down and decline while the growth of mobile cost overtakes the revenue growth.

A good friend of mine, who also is a great and successful CFO, stated that Profitability is rarely a problem to achieve (in the short term)”; “I turn down my market invest (i.e., OpEx) and my Profitability (as measured in terms of EBITDA) goes up. All I have done is getting my business profitable in the short term without having created any sustainable value or profit by this. Just engineered my bonus.”

Our aim must be to ensure sustainable and stable profitability. This can only be done by understanding, careful managing and engineering our basic Telco cost structures.

While most Telco’s tend to plan several years ahead for Capital Expenditures (CapEx) and often with a high degree of sophistication, the same Telco’s mainly focus on one (1!) year ahead for OpEx. Further effort channeled into OpEx is frequently highly simplistic and at times in-consistent with the planned CapEx. Obviously, in the growth phase of the business cycle one may take the easy way out on OpEx and focus more on the required CapEx to grow the business. However, as the time-scales for committed OpEx “lives” on a much longer period than Revenue (particular Prepaid Revenue or even CapEx for that matter), any shortfall in Revenue and Profitability will be much more difficult to mitigate by OpEx measures that takes time to become effective. In markets with little or no market investment the penalty can be even harsher as there is no or little OpEx cushion that can be used to soften a disappointing direction in profitability.

How come a telecom business in Asia, or other emerging growth markets around the world, can maintain, by European standards, such incredible high EBITDA Margins. Margin’s that run into 50s or even higher. Is this “just” a matter of different lower-cost & low GDP economies? Does the higher margins simply reflect a different stage in the business cycle (i.e., growth versus super-saturation)?, Should Mature Market really care too much about Emerging Growth Markets? in the sense of whether Mature Markets can learn anything from Emerging Growth Markets and maybe even vice versa? (i.e., certainly mature markets have made many mistakes, particular when shifting gears from growth to what should be sustainability).

Before all those questions have much of a meaning, it might be instructive to look at the differences between a Mature Market and an Emerging Growth Market. I obviously would not have started this Blog, unless I believe that there are important lessons to be had by understanding what is going on in both types of markets. I also should make it clear that I am only using the term Emerging Growth Markets as most of the markets I study is typically defined as such by economists and consultants. However from a mobile technology perspective few of those markets we tend to call Emerging Growth Markets can really be called emerging any longer and growth has slowed down a lot in most of those markets. This said, from a mobile broadband perspective most of the markets defined in this analysis as Emerging Growth Markets are pretty much dead on that definition.

Whether the emerging markets really should be looking forward to mobile broadband data growth might depend a lot on whether you are the consumer or the provider of services.

For most Mature Markets the introduction of 3G and mobile broadband data heralded a massive slow-down and in some cases even decline in revenue. This imposed severe strains on Mobile Margins and their EBITDAs. Today most mature markets mobile operators are facing a negative revenue growth rate and is “forced” continuously keep a razor focus on OpEx, Mitigating the revenue decline keeping Margin and EBITDA reasonably in check.

Emerging Markets should as early as possible focus on their operational expenses and Optimize with a Vengeance.

Well well let ‘s get back to the comparison and see what we can learn!

It doesn’t take to long to make a list of some of the key and maybe at times obvious differentiators (not intended to be exhaustive) between Mature and Emerging Markets are;

  • Side Note: it should be clear that by today many of the markets we used to call emerging growth markets are from mobile telephony penetration & business development certainly not emerging any longer and as growing as they were 5 or 10 years ago. This said from a 3G/4G mobile broadband data penetration perspective it might still be fair to characterize those markets as emerging and growing. Though as mature markets have seen that journey is not per se a financial growth story.

Looking at the above table we can assess that Firstly: the straightforward (and possible naïve) explanation of relative profitability differences between Mature and Emerging Markets, might be that emerging markets cost structures are much more favorable compared to what we find in mature market economies. Basically the difference between Low and High GDP economies. However, we should not allow ourselves to be too naïve here as lessons learned from low GDP economies are that some cost structure elements (e.g., real estate, fuel, electricity, etc..) are as costly (some times more so) than what we find back in mature high/higher GDP markets. Secondly: many emerging growth market’s economies are substantially more populous & dense than what we find in mature markets (i.e., although it is hard to beat Netherlands or the Ruhr Area in Germany). Maybe the higher population count & population density leads to better scale than can be achieved in mature markets. However, while maybe true for the urban population, emerging markets tend to have substantially higher ratio of their population living in rural areas compared to what we find in mature markets.  Thirdly: maybe the go-to-market approach in emerging markets is different from mature markets (e.g., subsidies, quality including network coverage, marketing,…) offering substantially lower mobile quality overall compared to what is the practice in mature markets. Providing poor mobile network quality certainly have been a recurring theme in the Philippines mobile industry despite the Telco Industry in Philippines enjoys Margins that most mature markets operators can only dream of. It is pretty clear that for 3G-UMTS based mobile broadband, 900 MHz does not have sufficient bandwidth to support the anticipated mobile broadband uptake in emerging markets (e.g., particular as 900MHz is occupied by 2G-GSM as well). IF emerging markets mobile operators will want to offer mobile data at reasonable quality levels (i.e., and the IF is intentional), sustain anticipated customer demand and growth they are likely to require network densification (i.e., extra CapEx and OpEx) at 2100 MHz. Alternative they might choose to wait for APT 700 MHz and drive an affordable low-cost LTE device ecosystem albeit this is some years ahead.

More than likely some of the answers of why emerging markets have a much better margins (at the moment at least) will have to do with cost-structure differences combined with possibly better scale and different go-to-market requirements more than compensating the low revenue per user.

Let us have a look at the usual suspects towards the differences between mature & emerging markets. The EBITDA can be derived as Revenue minus the Operational Expenses (i.e., OpEx) and the corresponding margin is Ebitda divided by the Revenue (ignoring special accounting effects that here);

EBITDA (E) = Revenue (R) – OpEx (O) and Margin (M) = EBITDA / Revenue.

The EBITDA & Margin tells us in absolute and relative terms how much of our Revenue we keep after all our Operational expenses (i.e., OpEx) have been paid (i.e., beside tax, interests, depreciation & amortization charges).

We can write Revenue as a the product of ARPU (Average Number of Users) times Number of Users N and thus the EBITDA can also be written as;

. We see that even if ARPU is low (or very) low, an Emerging Market with lot of users might match the Revenue of a Mature Market with higher ARPU and worse population scale (i.e., lower amount of users). Pretty simple!

But what about the Margin? , in order for an Emerging Market to have substantially better Margin than corresponding Mature Market at the same revenue level, it is clear that the Emerging Market’s OpEx (O) needs to be lower than that of a Mature markets. We also observe that if the Emerging Market Revenue is lower than the Mature Market, the corresponding Opex needs to be even lower than if the Revenues were identical. One would expect that lower GDP countries would have lower Opex (or Cost in general) combined with better population scale is really what makes for a great emerging market mobile Margins! … Or is it ?

A Small but essential de-tour into Cost Structure.

Some of the answers towards the differences in margin between mature and emerging markets obviously lay in the OpEx part or in the Cost-structure differences. Let’s take a look at a mature market’s cost structure (i.e., as you will find in Western & Eastern Europe) which pretty much looks like this;

With the following OpEx or cost-structure elements;

  • Usage-related OpEx:  typically take up between 10% to 35% of of the total OpEx with an average of ca. 25%. On average this OpEx contribution is approximately 17% of the revenue in mature European markets. Trend wise it is declining. Usage-based OpEx is dominated by interconnect & roaming voice traffic and to a less degree of data interconnect and peering. In a scenario where there is little circuit switched voice left (i.e., the ultimate LTE scenario) this cost element will diminish substantially from the operators cost structure. It should be noted that this also to some extend is being influenced by regulatory forces.
  • Market Invest: can be decomposed into Subscriber Acquisition Cost (SAC), i.e., “bribing” the customers to leave your competitor for yourself, Subscriber Retention Cost (SRC), i.e., “bribing” your existing (valuable) customers to not let them be “bribed” by you’re a competitor and leave you (i.e., churn), and lastly Other Marketing spend for advertisement, promotional and so forth. This cost-structure element contribution to OpEx can vary greatly depending on the market composition. In Europe’s mature markets it will vary from 10% to 31% with a mean value of ca. 23% of the total OpEx. On average it will be around 14% of the Revenue. It should be noted that as the mobile penetration increases and enter into heavy saturation (i.e., >100%), SAC tends to reduce and SRC will increase. Further in markets that are very prepaid heavy SAC and SRC will naturally be fairly minor cost structure element (i.e., 10% of OpEx or lower and only a couple of % of Revenue). Profit and Margin can rapidly be influenced by changes in the market invest. SAC and SRC cost-structure elements will in general be small in emerging growth markets (compared to corresponding mature markets).
  • Terminal-equipment related OpEx: is the cost associated by procuring terminals equipment (i.e, handsets, smartphones, data cards, etc.). In the past (prior to 2008) it was fairly common that OpEx from procuring and revenues from selling terminals were close to a zero-sum game. In other words the cost made for the operator of procuring terminals was pretty much covered by re-selling them to their customer base. This cost structure element is another  heavy weight and vary from 10% to 20% of the OpEx with an average in mature European markets of 17%. Terminal-related cost on average amounts to ca. 11% of the Revenue (in mature markets). Most operators in emerging growth markets don’t massively procure, re-sell and subsidies handsets, as is the case in many mature markets. Typically handsets and devices in emerging markets will be supplied by a substantial 2nd hand gray and black market readily available.
  • Personnel Cost: amounts to between 6% to 15% of the Total OpEx with a best-practice share of around the 10%. The ones who believe that this ratio is lower in emerging markets might re-think their impression. In my experience emerging growth markets (including the ones in Eastern & Central Europe) have a lower unit personnel cost but also tends to have much larger organizations. This leads to many emerging growth markets operators having a personnel cost share that is closer to the 15% than to the 10% or lower. On average personnel cost should be below 10% of revenue with best practice between 5% and 8% of the Revenue.
  • Technology Cost (Network & IT): includes all technology related OpEx for both Network and Information Technology. Personnel-related technology OpEx (prior to capitalization ) is accounted for in the above Personnel Cost Category and would typically be around 30% of the personnel cost pending on outsourcing level and organizational structure. Emerging markets in Central & Eastern Europe historical have had higher technology related personnel cost than mature markets. In general this is attributed to high-quality relative low-cost technology staff leading to less advantages in outsourcing technology functions. As Technology OpEx is the most frequent “victim” of efficiency initiatives, lets just have a look at how the anatomy of the Technology Cost Structure looks like:

  • Technology Cost (Network & IT) – continued: Although, above Chart (i.e., taken from my 2012 Keynote at the Broadband MEA 2012, Dubai “Ultra-efficient network factory: Network sharing and other means to leapfrog operator efficiencies”) emphasizes a Mature Market View, emerging markets cost distribution does not differ that much from the above with a few exceptions. In Emerging Growth Markets with poor electrification rates diesel generators and the associated diesel full will strain the Energy Cost substantially. As the biggest exposure to poor electrical grid (in emerging markets) in general tend to be in Rural and Sub-Urban areas it is a particular OpEx concern as the emerging market operators expands towards Rural Areas to capture the additional subscriber potential present there. Further diesel fuel has on average increased with 10% annually (i..e, over the least 10 years) and as such is a very substantial Margin and Profitability risk if a very large part of the cellular / mobile network requires diesel generators and respective fuel. Obviously, “Rental & Leasing” as well as “Service & Maintenance” & “Personnel Cost” would be positively impacted (i.e., reduced) by Network Sharing initiatives. Best practices Network Sharing can bring around 35% OpEx savings on relevant cost structures. For more details on benefits and disadvantages (often forgotten in the heat of the moment) see my Blog “The ABC of Network Sharing – The Fundamentals”. In my experience one of the greatest opportunities in Emerging Growth Markets for increased efficiency are in the Services part covering Maintenance & Repair (which obviously also incudes field maintenance and spare part services).
  • Other Cost: typically covers the rest of OpEx not captured by the above specific items. It can also be viewed as overhead cost. It is also often used to “hide” cost that might be painful for the organization (i.e., in terms of authorization or consequences of mistakes). In general you will find a very large amount of smaller to medium cost items here rather than larger ones. Best practices should keep this below 10% of total OpEx and ca. 5% of Revenues. Much above this either means mis-categorization, ad-hoc projects, or something else that needs further clarification.

So how does this help us compare a Mature Mobile Market with an Emerging Growth Market?

As already mentioned in the description of the above cost structure categories particular Market Invest and Terminal-equipment Cost are items that tend to be substantially lower for emerging market operators or entirely absent from their cost structures.

Lets assume our average mobile operator in an average mature mobile market (in Western Europe) have a Margin of 36%. In its existing (OpEx) cost structure they spend 15% of Revenue on Market Invest of which ca. 53% goes to subscriber acquisition (i.e., SAC cost category), 40% on subscriber retention (SRC) and another 7% for other marketing expenses. Further, this operator has been subsidizing their handset portfolio (i.e., Terminal Cost) which make up another 10% of the Revenue.

Our Average Operator comes up with the disruptive strategy to remove all SAC and SRC from their cost structure and stop procuring terminal equipment. Assuming (and that is a very big one in a typical western European mature market) that revenue remains at the same level, how would this average operator fare?

Removing SAC and SRC, which was 14% of the Revenue will improve the Margin adding another 14 percentage points. Removing terminal procurement from its cost structure leads to an additional Margin jump of 10 percentage points. The final result is a Margin of 60% which is fairly close to some of the highest margins we find in emerging growth markets. Obviously, completely annihilating Market Invest might not be the most market efficient move unless it is a market-wide initiative.

Albeit the example might be perceived as a wee bit academic, it serves to illustrate that some of the larger margin differences we observe between mobile operators in mature and emerging growth markets can be largely explain by differences in the basic cost structure, i..e, the lack of substantial subscriber acquisition and retention costs as well as not procuring terminals does offer advantages to the emerging market business model.

However, it also means that many operators in emerging markets have little OpEx flexibility, in the sense of faster OpEx reduction opportunities once mobile margin reduces due to for example slowing revenue growth. This typical becomes a challenge as mobile penetration starts reaching saturation and as ARPU reduces due to diminishing return on incremental customer acquisition.

There is not much substantial OpEx flexibility (i..e, market invest & terminal procurement) in Emerging Growth Markets mobile accounts. This adds to the challenge of avoiding profitability squeeze and margin exposure by quickly scaling back OpEx.

This is to some extend different from mature markets that historically had quiet a few low hanging fruits to address before OpEx efficiency and reduction became a real challenge. Though ultimately it does become a challenge.

Back to Profitability with a Vengeance.

So it is all pretty simple! … leave out Market Invest and Terminal Procurement … Then add that we typically have to do with Lower GDP countries which conventional wisdom would expect also to have lower Opex (or Cost in general) combined with better population scale .. isn’t that really what makes for a great emerging growth market Mobile Margin?

Hmmm … Albeit Compelling ! ? … For the ones (of us) who would think that the cost would scale nicely with GDP and therefor a Low GDP Country would have a relative Lower Cost Base, well …

  • In the Chart above the Y-axis is depicted with logarithmic scaling in order to provide a better impression of the data points across the different economies. It should be noted that throughout the years 2007 to 2013 (note: 2013 data is shown above)  there is no correlation between a countries mobile Opex, as estimated by Revenue – EBITDA, and the GDP.

Well … GDP really doesn’t provide the best explanation (to say the least)! … So what does then?

I have carried out multi-linear regression analysis on the available data from the “Bank of America Merrill Lynch (BoAML) Global Wireless Matrix Q1, 2014” datasets between the years 2007 to 2013. The multi-linear regression approach is based on year-by-year analysis of the data with many different subsets & combination of data chosen including adding random data.

I find that the best description (R-square 0.73, F-Ratio of 30 and p-value(s) <0.0001) of the 48 country’s data on Opex. The amount of data points used in the multi-regression is at least 48 for each parameter and that for each of the 7 years analyzed. The result of the (preliminary) analysis is given by the following statistically significant parameters explaining the Mobile Market OpEx:

  1. Population – The higher the size of the population, proportional less Mobile Market Opex is spend (i.e., scale advantage).
  2. Penetration – The higher the mobile penetration, proportionally less Mobile Market Opex is being spend (i.e., scale advantage and the incremental penetration at an already high penetration would have less value thus less Opex should be spend).
  3. Users (i..e., as measured by subscriptions) – The more Users the higher the Mobile Market Opex (note: prepaid ratio has not been found to add statistical significance).
  4. ARPU (Average Revenue Per User) – The higher the ARPU, the higher will the Mobile Market Opex be.

If I leave out ARPU, GDP does enter as a possible descriptive candidate although the overall quality of the regression analysis suffers. However, it appears that the GDP and ARPU cannot co-exist in the analysis. When Mobile Market ARPU data are included, GDP becomes non-significant. Furthermore, a countries Surface Area, which I previously believed would have a sizable impact on a Mobile Market’s OpEx, also does not enter as a significant descriptive parameter in this analysis. In general the Technology related OpEx is between 15% to 25% (maximum) of the Total OpEx and out that possible 40% to 60% would be related to sites that would be needed to cover a given surface area. This might no be significant enough in comparison to the other parameters or simply not a significant factor in the overall country related mobile OpEx.

I had also expected 3G-UMTS to have had a significant contribution to the Opex. However this was not very clear from the analysis either. Although in the some of the earlier years (2005 – 2007), 3G does enter albeit not with a lot of weight. In Western Europe most incremental OpEx related to 3G has been absorb in the existing cost structure and very little (if any) incremental OpEx would be visible particular after 2007. This might not be the case in most Emerging Markets unless they can rely on UMTS deployments at 900 MHz (i.e., the traditional GSM band). Also the UMTS 900 solution would only last until capacity demand require the operators to deploy UMTS 2100 (or let their customers suffer with less mobile data quality and keep the OpEx at existing levels). In rural areas (already covered by GSM at 900 MHz) the 900 MHz UMTS deployment option may mitigate incremental OpEx of new site deployment and further encourage rural active network sharing to allow for lower cost deployment and providing rural populations with mobile data and internet access.

The Population Size of a Country, the Mobile Penetration and the number of Users and their ARPU (note last two basically multiplies up to the revenue) are most clearly driving a mobile markets Opex.

Philippines versus Germany – Revenue, Cost & Profitability.

Philippines in 2013 is estimated to have a population of ca. 100 Million compared to Germany’s ca. 80 Million. The Urban population in Germany is 75% taking up ca. 17% of the German surface area (ca. 61,000 km2 or a bit more than Croatia). Comparison this to Philippines 50% urbanization that takes up up only 3% (ca. 9,000 km2 or equivalent to the surface area of Cyprus). Germany surface area is about 20% larger than Philippines (although the geographies are widely .. wildly may be a better word … different, with the Philippines archipelago comprising 7,107 islands of which ca. 2,000 are inhabited, making the German geography slightly boring in comparison).

In principle if all I care about is to cover and offer services to the urban population (supposedly the ones with the money?) I only need to cover 9 – 10 thousand square kilometer in the Philippines to capture ca. 50 Million potential mobile users (or 5,000 pop per km2), while I would need to cover about 6 times that amount of surface area to capture 60 million urban users in Germany (or 1,000 pop per km2). Even when taking capacity and quality into account, my Philippine cellular network should be a lot smaller and more efficient than my German mobile network. If everything would be equal, I basically would need 6 times more sites in Germany compared to Philippines. Particular if I don’t care too much about good quality but just want to provide best effort services (that would never work in Germany by the way). Philippines would win any day over Germany in terms of OpEx and obviously also in terms of capital investments or CapEx. It does help the German Network Economics that the ARPU level in Germany is between 4 times (in 2003) to 6 times (in 2013) higher than in Philippines. Do note that the two major Germany mobile operators covers almost 100% of the population as well as most of the German surface area and that with a superior quality of voice as well as mobile broadband data. This does not true hold true for Philippines.

In 2003 a mobile consumer in Philippines would spend on average almost 8 US$ per month for mobile services. This was ca. 4x lower than a German customer for that year. The 2003 ARPU of the Philippines roughly corresponded to 10% of the GDP per Capita versus 1.2% of the German equivalent. Over the 10 Years from 2003 to 2013, ARPU dropped 60% in Philippine and by 2013 corresponded to ca. 1.5% of GDP per Capita (i.e., a lot more affordable propositions). The German 2013 ARPU to GDP per Capita ratio was 0.5% and its ARPU was ca. 40% lower than in 2003.

The Philippines ARPU decline and Opex increase over the 10 year period led to a Margin drop from 64% to 45% (19% drop!) and their Margin is still highly likely to fall further in the near to medium-term. Despite the Margin drop Philippines still made a PHP26 Billion more EBITDA in 2013 than compared to 2003 (ca. 45% more or equivalent compounded annual growth rate of 3.8%).

in 2003

  • Germany had ca. 3x more mobile subscribers compared to Philippines.
  • German Mobile Revenue was 14x higher than Philippines.
  • German EBITDA was 9x higher than that of Philippines.
  • German OpEx was 23x higher than that of Philippines Mobile Industry.
  • Mobile Margin of the Philippines was 64% versus 42% of Germany.
  • Germany’s GPD per Capita (in US$) was 35 times larger than that of Philippines.
  • Germany’s mobile ARPU was 4 times higher than that of Philippines.

in 2013 (+ 10 Years)

  • Philippines & Germany have almost the same amount of mobile subscriptions.
  • Germany Mobile Revenue was 6x higher than Philippines.
  • German EBITDA was only 5x higher than that of Philippines.
  • German OpEx was 6x higher than Mobile OpEx in Philippines (and German OpEx was at level with 2003).
  • Mobile Margin of the Philippines dropped 19% to 45% compared to 42% of Germany (essential similar to 2003).
  • In local currencies, Philippines increased their EBITDA with ca. 45%, Germany remain constant.
  • Both Philippines and Germany has lost 11% in absolute EBITDA between the 10 Year periods maximum and 2013.
  • Germany’s GDP per Capita (in US$) was 14 times larger than that of the Philippines.
  • Germany’s ARPU was 6 times higher than that of Philippines.

In the Philippines, mobile revenues have grown with 7.4% per anno (between 2003 and 2013) while the corresponding mobile OpEx grew with 12% and thus eroding margin massively over the period as increasingly more mobile customers were addressed. In Philippines, the 2013 OpEx level was 3 times that of 2003 (despite one major network consolidation and being an essential duopoly after the consolidation). In Philippines over this period the annual growth rate of mobile users were 17% (versus Germany’s 6%). In absolute terms the number of users in Germany and Philippines were almost the same in 2013, ca. 115 Million versus 109 Million. In Germany over the same period Financial growth was hardly present although more than 50 Million subscriptions were added.

When OpEx grows faster than Revenue, Profitability will suffer today & even more so tomorrow.

Mobile capital investments (i.e., CapEx) over the period 2003 to 2013 was for Germany 5 times higher than that of Philippines (i.e., remember that Germany also needs at least 5 – 6 times more sites to cover the Urban population) and tracks at a 13% Capex to Revenue ratio versus Philippines 20%.

The stories of Mobile Philippines and of Mobile Germany are not unique. Likewise examples can be found in Emerging Growth Markets as well as Mature Markets.

Can Mature Markets learn or even match (keep on dreaming?) from Emerging Markets in terms of efficiency? Assuming such markets really are efficient of course!

As logic (true or false) would dictate given the relative low ARPUs in emerging growth markets and their correspondingly high margins, one should think that such emerging markets are forced to run their business much more efficient than in Mature Markets. While compelling to believe this, the economical data would indicate that most emerging growth markets have been riding the subscriber & revenue growth band wagon without too much thoughts to the OpEx part … and Frankly why should you care about OpEx when your business generates margins much excess of 40%? Well … it is (much) easier to manage & control OpEx year by year than to abruptly “one day” having to cut cost in panic mode when growth slows down the really ugly way and OpEx keeps increasing without a care in the world. Many mature market operators have been in this situation in the past (e.g., 2004 – 2008) and still today works hard to keep their margins stable and profitability from declining.

Most Companies will report both Revenues and EBITDA on quarterly and annual basis as both are key financial & operational indicators for growth. They tend not report Opex but as seen from above that’s really not a problem to estimate when you have Revenue and EBITDA (i.e., OpEx = Revenue – EBITDA).

Thus, had you left the European Telco scene (assuming you were there in the first place) for the last 10 years and then came back you might have concluded that not much have happened in your absence … at least from a profitability perspective. Germany was in 2013 almost at its Ebitda margin level of 2003. Of course as the ones who did not take a long holiday knows those last 10 years were far from blissful financial & operational harmony in the mature markets where one efficiency program after the other struggled to manage, control and reduce Operators Operational Expenses.

However, over that 10-year period Germany added 50+ Million mobile subscriptions and invested more than 37 Billion US$ into the mobile networks from T-Deutschland, Vodafone, E-plus and Telefonica-O2. The mobile country margin over the 10-year period has been ca. 43% and the Capex to Revenue ratio ca. 13%. By 2013 the total amount of mobile subscription was in the order of 115 Million out of a population of 81 Million (i.e., 54 Million of the German population is between 15 and 64 years of age). The observant numerologist would have realized that there are many more subscriptions than population … this is not surprising as it reflects that many subscribers are having multiple different SIM cards (as opposed to cloned SIMs) or subscription types based on their device portfolio and a host of other reasons.

All Wunderbar! … or? .. well not really … Take a look at the revenue and profitability over the 10 year period and you will find that no (or very very little) revenue and incremental profitability has been gained over the period from 2003 to 2013. AND we did add 80+% more subscriptions to the base!

Here is the Germany Mobile development over the period;

Apart from adding subscribers, having modernized the mobile networks at least twice over the period (i.e, CapEx with little OpEx impact) and introduced LTE into the German market (i.e., with little additional revenue to show for it) not much additional value has been added. It is however no small treat what has happen in Germany (and in many other mature markets for that matter). Not only did Germany almost double the mobile customers (in terms of subscriptions), over the period 3G Nodes-B’s were over-layed across the existing 2G network. Many additional sites were added in Germany as the fundamental 2G cellular grid was primarily based on 900 MHz and to accommodate the higher UMTS frequency (i.e., 2100 MHz) more new locations were added to provide a superior 3G coverage (and capacity/quality). Still Germany managed all this without increasing the Mobile Country OpEx across the period (apart from some minor swings). This has been achieved by a tremendous attention to OpEx efficiency with every part of the Industry having razor sharp attention to cost reduction and operating at increasingly efficiency.

Philippines story is a Fabulous Story of Growth (as summarized above) … and of Profitability & Margin Decline.

Philippines today is in effect a duopoly with PLDT having approx. 2/3 of the mobile market and Globe the remaining 1/3. During the period the Philippine Market saw Sun Cellular being acquired and merged by PLDT. Further, 3G was deployed and mobile data launched in major urban areas. SMS revenues remained the largest share of non-voice revenue to the two remaining mobile operators PLDT and Globe. Over the period 2003 to 2013, the mobile subscriber base (in terms of subscriptions) grew with 16% per anno and the ARPU fell accordingly with 10% per anno (all measured in local currency). All-in-all safe guarding a “healthy” revenue increase over the period from ca. 93 Billion PHP in 2003 to 190 Billion PHP in 2013 (i.e., a 65% increase over the period corresponding to a 5% annual growth rate).

However, the Philippine market could not maintain their relative profitability & initial efficiency as the mobile market grew.

So we observe (at least) two effects (1) Reduction in ARPU as market is growing & (2) Increasing Opex cost to sustain the growth in the market. As more customers are added to a mobile network the return on thus customers increasingly diminishes as network needs to be massively extended capturing the full market potential versus “just” the major urban potential.

Mobile Philippines did become less economical efficient as its scale increases and ARPU dropped (i.e., by almost 70%). This is not an unusual finding across Emerging Growth Markets.

As I have described in my previous Blog “SMS – Assimilation is inevitable, Resistance is Futile!”, Philippines mobile market has an extreme exposure to SMS Revenues which amounts to more than 35% of Total Revenues. Particular as mobile data and smartphones penetrate the Philippine markets. As described in my previous Blog, SMS Services enjoy the highest profitability across the whole range of mobile services we offer the mobile customer including voice. As SMS is being cannibalized by IP-based messaging, the revenue will decline dramatically and the mobile data revenue is not likely to catch up with this decline. Furthermore, profitability will suffer as the the most profitable service (i.e., SMS) is replaced by mobile data that by nature has a different profitability impact compared to simple SMS services.

Philippines do not only have a substantial Margin & EBITDA risk from un-managed OpEx but also from SMS revenue cannibalization (a la KPN in the Netherlands and then some).

Let us compare the ARPU & Opex development for Philippines (above Chart) with that of Germany over the same period 2003 to 2013 (please note that the scale of Opex is very narrow)

Mobile Germany managed their Cost Structure despite 40+% decrease in ARPU and as another 60% in mobile penetration was added to the mobile business. Again similar trend will be found in most Mature Markets in Western Europe.

One may argue (and not being too wrong) that Germany (and most mature mobile markets) in 2003 already had most of its OpEx bearing organization, processes, logistics and infrastructure in place to continue acquiring subscribers (i.e., as measured in subscriptions). Therefor it have been much easier for the mature market operators to maintain their OpEx as they continued to grow. Also true that many emerging mobile markets did not have the same (high) deployment and quality criteria, as in western mature markets, in their initial network and service deployment (i.e., certainly true for the Philippines as is evident from the many Regulatory warnings both PLDT and Globe received over the years) providing basic voice coverage in populated areas but little service in sub-urban and rural areas.

Most of the initial emerging market networks has been based on coarse (by mature market standards) GSM 900 MHz (or CDMA 850 MHz) grids with relative little available capacity and indoor coverage in comparison to population and clutter types (i.e., geographical topologies characterized by their cellular radio interference patterns). The challenge is, as an operator wants to capture more customers, it will need to build out / extend its mobile network in the areas those potential or prospective new customers live and work in. From a cost perspective sub-urban and rural areas in emerging markets are not per se lower cost areas despite such areas in general being lower revenue areas than their urban equivalents. Thus, as more customers are added (i.e.,  increased mobile penetration) proportionally more cost are generated than revenue being capture and the relative margin will decline. … and this is how the Ugly-cost (or profitability tail) is created.

  • I just cannot write about profitability and cost structure without throwing the Ugly-(cost)-Tail on the page.I strongly encourage all mobile operators to make their own Ugly-Tail analysis. You will find more details of how to remedy this Ugliness from your cost structure in “The ABC of Network Sharing – The Fundamentals”.

In Western Europe’s mature mobile markets we find that more than 50% of our mobile cellular sites captures no more than 10% of the Revenues (but we do tend to cover almost all surface area several times unless the mobile operators have managed to see the logic of rural network sharing and consolidated those rural & sub-urban networks). Given emerging mobile markets have “gone less over board” in terms of lowest revenue un-profitable network deployments in rural areas you will find that the number of sites carrying 10% of less of the revenue is around 40%. It should be remembered that the rural populations in emerging growth markets tend to be a lot larger than in of that in mature markets and as such revenue is in principle spread out more than what would be the case in mature markets.

Population & Mobile Statistics and Opex Trends.

The following provides a 2013 Summary of Mobile Penetration, 3G Penetration (measured in subscriptions), Urban Population and the corresponding share of surface area under urban settlement. Further to guide the eye the 100% line has been inserted (red solid line), a red dotted line that represents the share of the population that is between 15 and 64 years of age (i.e., who are more likely to afford a mobile service) and a dashed red line providing the average across all the 43 countries analyzed in this Blog.

  • Sources: United Nations, Department of Economic & Social Affairs, Population Division.  The UN data is somewhat outdated though for most data points across emerging and mature markets changes have been minor. Mobile Penetration is based on Pyramid Research and Bank of America Merrill Lynch Global Wireless Matrix Q1, 2014. Index Mundi is the source for the Country Age structure and data for %tage of population between 15 and 64 years of age and shown as a red dotted line which swings between 53.2% (Nigeria) to 78.2% (Singapore), with an average of 66.5% (red dashed line).

There is a couple of points (out of many) that can be made on the above data;

  1. There are no real emerging markets any longer in the sense of providing basic mobile telephone services such as voice and messaging.
  2. For mobile broadband data via 3G-UMTS (or LTE for that matter), what we tend to characterize as emerging markets are truly emerging or in some case nascent (e.g., Algeria, Iraq, India, Pakistan, etc..). 
  3. All mature markets have mobile penetration rates way above 100% with exception of Canada, i.e., 80% (i.e., though getting to 100% in Canada might be a real challenge due to a very dispersed remaining 20+% of the population).
  4. Most emerging markets are by now covering all urban areas and corresponding urban population. Many have also reach 100% mobile penetration rates.
  5. Most Emerging Markets are lagging Western Mature Markets in 3G penetration. Even providing urban population & urban areas with high bandwidth mobile data is behind that of mature markets.

Size & density does matter … in all kind of ways when it comes to the economics of mobile networks and the business itself.

In Australia I only need to cover ca. 40 thousand km2 (i.e., 0.5% of the total surface area and a bit less than the size of Denmark) to have captured almost 90% of the Australian population (e.g., Australia’s total size is 180+ times that of Denmark excluding Greenland). I frequently hear my Australian friends telling me how Australia covers almost 100% of the population (and I am sure that they cover more area than is equivalent to Denmark too) … but without being (too) disrespectful that record is not for Guinness Book of Records anytime soon. in US (e.g., 20% more surface area than Australia) I need to cover in almost 800 thousand km2 (8.2% of surface area or equivalent  to a bit more than Turkey) to capture more than 80% of the population. In Thailand I can only capture 35% of the population by covering ca. 5% of the surface area or a little less than 30 thousand km2 (approx. the equivalent of Belgium). The remaining of 65% of the Thai population is rural-based and spread across a much larger surface area requiring extensive mobile network to provide coverage to and capture additional market share outside the urban population.

So in Thailand I might need a bit less cell sites to cover 35% of my population (i.e., 22M) than in Australia to cover almost 90% of the population (i.e., ca. 21M). That’s pretty cool economics for Australia which is also reflected in a very low profitability risk score. For Thailand (and other countries with similar urban demographics) it is tough luck if they want to reach out and get the remaining 65% of their population. The geographical dispersion of the population outside urban areas is very wide and increasing geographical area is required to be covered in order to catch this population group. UMTS at 900 MHz will help to deploy economical mobile broadband, as will LTE in the APT 700 MHz band (being it either FDD Band 28 or TDD Band 44) as the terminal portfolio becomes affordable for rural and sub-urban populations in emerging growth markets.

In Western Europe on average I can capture 77% of my population (i..e, the urban pop) covering 14.2% of the surface area (i.e., average over markets in this analysis), This is all very agreeable and almost all Western European countries cover their surface areas to at least 80% and in most cases beyond that (i.e., it’s just less & easier land to cover though not per see less costly). In most cases rural coverage is encourage (or required) by the mature market license regime and not always a choice of the mobile operators.

Before we look in depth to the growth (incl. positive as well as negative growth), lets first have a peek at what has happened to the mobile revenue in terms of ARPU and Number of Mobile User and the corresponding mobile penetration over the period 2007 to 2013.

  • Source: Bank of America Merrill Lynch Global Wireless Matrix Q1, 2014 and Pyramid Research Data data were used to calculated the growth of ARPU as compounded annual growth rate between 2007 to 2013 and the annual growth rate between 2012 and 2013. Since 2007 the mobile ARPUs have been in decline and to make matters worse the decline has even accelerated rather than slowed down as markets mobile penetration saturated.

  • Source: Mobile Penetrations taken from Bank of America Merrill Lynch Global Wireless Matrix Q1, 2014 and Pyramid Research Data data .Index Mundi is the source for the Country Age structure and data for %tage of population between 15 and 64 years of age and shown as a red dotted line which swings between 53.2% (Nigeria) to 78.2% (Singapore), with an average of 66.5% (red dashed line). It s interesting to observe that most emerging growth markets are now where the mature markets were in 2007 in terms of mobile penetration.

Apart from a very few markets, ARPU has been in a steady decline since 2007. Further in many countries the ARPU decline has even accelerated rather than slowed down. From most mature markets the conclusion that we can draw is that there are no evidence that mobile broadband data (via 3G-UMTS or LTE) has had any positive effect on ARPU. Although some of the ARPU decline over the period in mature markets (particular European Union countries) can be attributed to regulatory actions. In general as soon a country mobile penetration reaches 100% (in all effect reaches the part of the population 15-64 years of age) ARPU tends to decline faster rather than slowing down. Of course one may correctly argue that this is not a big issue as long as the ARPU times the Users (i.e., total revenue) remain growing healthily. However, as we will see that is yet another challenge for the mobile industry as also the total revenue in mature markets also are in decline on a year by year basis. Given the market, revenue & cost structures of emerging growth markets, it is not unlikely that they will face similar challenges to their mobile revenues (and thus profitability). This could have a much more dramatic effect on their overall mobile economics & business models than what has been experienced in the mature markets which have had a lot more “cushion” on the P&Ls to defend and even grow (albeit weakly) their profitability. It is instructive to see that the most emerging growth markets mobile penetrations have reached the levels of Mature Markets in 2007. Combined with the introduction and uptake of mobile broadband data this marks a more troublesome business model phase than what these markets have experienced in the past.Some of the emerging growth market have yet to introduce 3G-UMTS, and some to leapfrog mobile broadband by launching LTE. Both events, based on lessons learned from mature markets, heralds a more difficult business model period of managing cost structures while defending revenues from decline and satisfy customers appetite for mobile broadband internet that cannot be supported by such countries fixed telecommunications infrastructures.

For us to understand more profoundly where our mobile profitability is heading it is obviously a good idea to understand how our Revenue and OpEx is trending. In this Section I am only concerned about the Mobile Market in Country and not the individual mobile operators in the country. For that latter (i.e., Operator Profitability) you will find a really cool and exiting analytic framework in the Section after this. I am also not interested (in this article) in modeling the mobile business bottom up (been there & done that … but that is an entirely different story line). However, I am interested and I am hunting for some higher level understanding and a more holistic approach that will allow me to probabilistically (by way of Bayesian analysis & ultimately inference) to predict in which direction a given market is heading when it comes to Revenue, OpEx and of course the resulting EBITDA and Margin. The analysis I am presenting in this Section is preliminary and only includes compounded annual growth rates as well as the Year-by-Year growth rates of Revenue and OpEx. Further developments will include specific market & regulatory developments as well to further improve on the Bayesian approach. Given the wealth of data accumulated over the years from the Bank of America Merrill Lynch (BoAML) Global Wireless Matrix datasets it is fairly easy to construct & train statistical models as well as testing those consistent with best practices.

The Chart below comprises 48 countries Revenue & OpEx growth rates as derived from the “Bank of America Merrill Lynch (BoAML) Global Wireless Matrix Q1, 2014” dataset (note: BoAML data available in this analysis goes back to 2003). Out of the 48 Countries, 23 countries have an Opex compounded annual growth rate higher than the corresponding Revenue growth rate. Thus, it is clear that those 23 countries are having a higher risk of reduced margin and strained profitability due to over-proportionate growth of OpEx. Out of the 23 countries with high or very high profitability risk, 11 countries have been characterized in macro-economical terms as emerging growth markets (i.e.,  China, India, Indonesia, Philippines, Egypt, Morocco, Nigeria, Russia, Turkey, Chile, Mexico) the remaining 12 countries can be characterized as mature markets in macro-economical terms (i.e., New Zealand, Singapore, Austria, Belgium, France, Greece, Spain, Canada, South Korea, Malaysia, Taiwan, Israel). Furthermore, 26 countries had a higher Opex growth between 2012 and 2013 than their revenues and is likely to be trending towards dangerous territory in terms of Profitability Risk.

  • Source: Bank of America Merrill Lynch Global Wireless Matrix Q1, 2014. Revenue depicted here is Service Revenues and the OPEX has been calculated as Service REVENUE minus EBITDA. The Compounded Annual Growth Rate (CAGR) is calculated with X being Revenue and Opex. Y-axis scale is from -25% to +25% (i.e., similar to the scale chosen in the Year- by-Year growth rate shown in the Chart below).

With few exceptions one does not need to read the countries names on the Chart above to immediately see where we have the Mature Markets with little or negative growth and where what we typically call emerging growth markets are located.

As the above Chart clearly illustrate the mobile industry across different types of markets have an increasing challenge to deliver profitable growth and if the trend continues to keep their profitability period!

Opex grows faster than Mobile Operator’s can capture Revenue … That’s a problem!

In order gauge whether the growth dynamics of the last 7 years is something to be concerned about (it is! … it most definitely is! but humor me!) … it is worthwhile to take a look at the year by year growth rate trends (i.e. as CAGR only measures the starting point and the end point and “doesn’t really care” about what happens in the in-between years).

  • Source: Bank of America Merrill Lynch Global Wireless Matrix Q1, 2014. Revenue depicted here is Service Revenues and the OPEX has been calculated as Service REVENUE minus EBITDA. Year on Year growth is calculated and is depicted in the Chart above. Y-axis scale is from -25% to +25%. Note that the Y-scales in the Year-on-Year Growth Chart and the above 7-Year CAGR Growth Chart are the same and thus directly comparable.

From the Year on Year Growth dynamics compared to the compounded 7-year annual growth rate, we find that Mature Markets Mobile Revenues decline has accelerated. However, in most cases the Mature Market OpEx is declining as well and the Control & Management of the cost structure has improved markedly over the last 7 years. Despite the cost structure management most Mature Markets Revenue have been declining faster than the OpEx. As a result Profitability Squeeze remains a substantial risk in Mature Markets in general.

In almost all Emerging Growth Markets the 2013 to 2012 revenue growth rate has declined in comparison with the compounded annual growth rate. Not surprising as most of those markets are heading towards 100% mobile penetration (as measured in subscriptions). OpEx growth remains a dire concern for most of the emerging growth markets and will continue to squeeze emerging markets profitability and respective margins. There is no indication (in the dataset analyzed) that OpEx is really under control in Emerging Growth Markets, at least to the same degree as what is observed in the Mature Markets (i.e., particular Western Europe). What further adds to the emerging markets profitability risk is that mobile data networks (i.e., 3G-UMTS, HSPA+,..) and corresponding mobile data uptakes are just in its infancy in most of the Emerging Growth Markets in this analysis. The networks required to sustain demand (at a reasonable quality) are more extensive than what was required to provide okay-voice and SMS. Most of the emerging growth markets have no significant fixed (broadband data) infrastructure and in addition poor media distribution infrastructure which can relieve the mobile data networks being built. Huge rural populations with little available ARPU potential but a huge appetite to get connected to internet and media will further stress the mobile business models cost structure and sustainable profitability.

This argument is best illustrated by comparing the household digital ecosystem evolution (or revolution) in Western Europe with the projected evolution of Emerging Growth Markets.

 

  • Above Chart illustrates the likely evolution in Home and Personal Digital Infrastructure Ecosystem of an emerging market’s Household (HH). Particular note that the amount of TV Displays are very low and much of the media distribution is expected to happen over cellular and wireless networks. An additional challenge is that the fixed broadband infrastructure is widely lagging in many emerging markets (in particular in sub-urban and rural areas) increasing the requirements of the mobile network in those markets. It is compelling to believe that we will witness a completely different use case scenarios of digital media consumption than experienced in the Western Mature Markets. The emerging market is not likely to have the same degree of mobile/cellular data off-load as experienced in mature markets and as such will strain mobile networks air-interface, backhaul and backbone substantially more than is the case in mature markets. Source: Dr. Kim K Larsen Broadband MEA 2013 keynote on “Growth Pains: How networks will supply data capacity for 2020

  • Same as above but projection for Western Europe. In comparison with Emerging Markets a Mature Market Household  (HH) has many more TV as wells as a substantially higher fixed broadband penetration offering high-bandwidth digital media distribution as well as off-load optionality for mobile devices via WiFi. Source: Dr. Kim K Larsen Broadband MEA 2013 keynote on “Growth Pains: How networks will supply data capacity for 2020

Mobile Market Profit Sustainability Risk Index

The comprehensive dataset from Bank of America Merrill Lynch Global Wireless Matrix allows us to estimate what I have chosen to call a Market Profit Sustainability Risk Index. This Index provides a measure for the direction (i.e., growth rates) of Revenue & Opex and thus for the Profitability.

The Chart below is the preliminary result of such an analysis limited to the BoAML Global Wireless Matrix Quarter 1 of 2014. I am currently extending the Bayesian Analysis to include additional data rather than relying only on growth rates of Revenue & Opex, e.g., (1) market consolidation should improve the cost structure of the mobile business, (2) introducing 3G usually introduces a negative jump in the mobile operator cost structure, (3) mobile revenue growth rate reduces as mobile penetration increases, (4) regulatory actions & forces will reduce revenues and might have both positive and negative effects on the relevant cost structure, etc.…

So here it is! Preliminary but nevertheless directionally reasonable based on Revenue & Opex growth rates, the Market Profit Sustainability Risk Index over for 48 Mature & Emerging Growth Markets worldwide:

The above Market Profit Sustainability Risk Index is using the following risk profiles

  1. Very High Risk (index –5): (i.e., for margin decline): (i) Compounded Annual Growth Rate (CAGR) between 2007 and 2013 of Opex was higher than equivalent for Revenue AND (ii) Year-on-Year (YoY) Growth Rate 2012 to 2013 of Opex higher than that of Revenue AND (iii) Opex Year-on-Year 2012 to 2013 Growth Rate is higher than the Opex CAGR over the period 2007 to 2013.
  2. High Risk (index –3): Same as above Very High Risk with condition (iii) removed OR YoY Revenue Growth 2012 to 2013 lower than the corresponding Opex Growth.
  3. Medium Risk (index –2): CAGR of Revenue lower than CAGR of Opex but last year (i.e., 2012 t0 2013) growth rate of Revenue higher than that of Opex.
  4. Low Risk (index 1): (i) CAGR of Revenue higher than CAGR of Opex AND (ii) YoY Revenue Growth higher than Opex Growth but lower than the inflation of the previous year.
  5. Very Low Risk (index 3): Same as above Low Risk with YoY Revenue Growth Rate required to be higher than the Opex Growth with at least the previous year’s inflation rate.

The Outlook for Mature Markets are fairly positive as most of those Market have engaged in structural cost control and management for the last 7 to 8 years. Emerging Growth Markets Profit Sustainability Risk Index are cause for concern. As the mobile markets are saturating it usually results in lower ARPU and higher cost to reach the remaining parts of the population (often “encouraged” by regulation). Most Emerging Growth markets have started to introduce mobile data, which is likely to result in higher cost-structure pressure & with traditional revenue streams under pressure (if history of Mature Markets are to repeat itself in emerging growth markets). The Emerging Growth Markets have had little incentive (in the past) to focus on cost structure control and management, due to the exceedingly high margins that they historically could present with their legacy mobile services (i.e., Voice & SMS) and relative light networks (as always in comparison to Mature Markets).

Cautionary note is appropriate. All the above are based on the Mobile Market across the world. There are causes and effects that can move a market from having a high risk profile to a lower. Even if I feel that the dataset supports the categorization it remains preliminary as more effects should be included in the current risk model to add even more confidence in its predictive power. Furthermore, the analysis is probabilistic in nature and as such does not claim to carve in stone the future. All the Index claims to do is to indicate a probable direction of the profitability (as well as Revenue & OpEx). There are several ways that Operators and Regulatory Authorities might influence the direction of the profitability changing Risk Exposure (in the Wrong as well as in the Right Direction)

Furthermore, it would be wrong to apply the Market Profit Sustainability Risk Index to individual mobile operators in the relevant markets analyzed here. The profitability dynamics of individual mobile operators are a wee bit more complicated, albeit some guidelines and predictive trends for their profitability dynamics in terms of Revenue and Opex can be defined. This will all be revealed in the following Section.

Operator Profitability – the Profitability Math.

We have seen that the Margin M an be written as

with E, R and O being EBITDA, REVENUE and OPEX respectively.

However, much more interesting is that it can also be written as a function of subscriber share

being valid forwith being the margin and the subscriber market share can be found between 0% to 100%. The rest will follow in more details below, suffice to say that as the subscriber market share increases the Margin (or relative profitability) increases as well although not linearly (if anyone would have expected that ).

Before we get down and dirty on the math lets discuss Operator Profitability from a higher level and in terms of such an operators subscriber market share (i.e., typically measured in subscriptions rather than individual users).

In the following I will show some Individual Operator examples of EBITDA Margin dynamics from Mature Markets limited to Western Europe. Obviously the analysis and approach is not limited emerging markets and can (have been) directly extended to Emerging Growth Markets or any mobile market for that matter. Again BoAML Global Matrix provides a very rich data set for applying the approach described in this Blog.

It has been well established (i.e., by un-accountable and/or un-countable Consultants & Advisors) that an Operator’s Margin correlates reasonably well with its Subscriber Market Share as the Chart below illustrates very well. In addition the Chart below also includes the T-Mobile Netherlands profitability journey from 2002 to 2006 up to the point where Deutsche Telekom looked into acquiring Orange Netherlands. An event that took place in the Summer of 2007.

I do love the above Chart (i.e., must be the physicist in me?) as it shows that such a richness in business dynamics all boiled down to two main driver, i.e., Margin & Subscriber Market Shared.

So how can an Operator strategize to improve its profitability?

Let us take an Example

Here is how we can think about it in terms of Subscriber Market Share and EBITDA as depicted by the above Chart. In simple terms an Operator have a combination of two choices (Bullet 1 in above Chart) Improve its profitability through Opex reductions and making its operation more efficient without much additional growth (i.e., also resulting in little subscriber acquisition cost), it can improve its ARPU profile by increasing its revenue per subscriber (smiling a bit cynical here while writing this) again without adding much in additional market share. The first part of Bullet 1 has been pretty much business as usual in Western Europe since 2004 at least (unfortunately very few examples of the 2nd part of Bullet 1) and (Bullet 2 in above Chart) The above “Margin vs. Subscriber Market Share”  Chart indicates that if you can acquire the customers of another company (i.e., via Acquisition & Merger) it should be possible to quantum leap your market share while increasing the efficiencies of the operation by scale effects. In the above Example Chart our Hero has ca. 15% Customer Market Share and the Hero’s target ca. 10%. Thus after an acquisition our Hero would expect to get ca. 25% (if they play it well enough). Similarly we would expect a boost in profitability and hope for at least 38% if our Hero has 18% margin and our Target has 20%. Maybe even better as the scale should improve this further. Obviously, this kind of “math” assumes that our Hero and Target can work in isolation from the rest of the market and that no competitive forces would be at play to disrupt the well thought through plan (or that nothing otherwise disruptive happens in parallel with the merger of the two businesses). Of course such a venture comes with a price tag (i.e., the acquisition price) that needs to be factored into the overall economics of acquiring customers. As said most (Western) Operators are in a perpetual state of managing & controlling cost to maintain their Margin, protect and/or improve their EBITDA.

So one thing is theory! Let us see how the Dutch Mobile Markets Profitability Dynamics evolved over the 10 year period from 2003 to 2013;

From both KPN’s acquisition of Telfort as well as the acquisition & merger of Orange by T-Mobile above Margin vs. Subscriber Market Share Chart, we see that in general, the Market Share logic works. On the other hand the management of the integration of the business would have been fairly unlucky for that to be right. When it comes to the EBITDA logic it does look a little less obvious. KPN clearly got unlucky (if un-luck has something to do with it?) as their margin decline with a small uplift albeit still lower than where they started pre-acquisition. KPN should have expected a margin lift to 50+%. That did not happen to KPN – Telfort. T-Mobile did fare better although we do observe a margin uplift to around 30% that can be attributed to Opex synergies resulting from the integration of the two businesses. However, it has taken many Opex efficiency rounds to get the Margin up to 38% that was the original target for the T-Mobile – Orange transaction.

In the past it was customary to take lots of operators from many countries, plot their margin versus subscriber markets share, draw a straight line through the data points and conclude that the margin potential is directly related to the Subscriber Market Share. This idea is depicted by the Left Side Chart and the Straight line “Best” Fit to data.

Lets just terminate that idea … it is wrong and does not reflect the right margin dynamics as a function of the subscriber markets share. Furthermore, the margin dynamics is not a straight-line function of the subscriber market share but rather asymptotic falling off towards minus infinity, i.e., when the company have no subscribers and no revenue but non-zero cost. We also observed a diminishing return on additional market share in the sense that as more market share is gained smaller and smaller incremental margins are gained. The magenta dashed line in the Left Chart below illustrates how one should expect the Margin to behave as a function of Subscriber market share.

 

The Right Chart above shows has broken down the data points in country by country. It is obvious that different countries have different margin versus market share behavior and that drawing a curve through all of those might be a bit naïve.

So how can we understand this behavior? Let us start with making a very simple formula a lot more complex :–)

We can write the Marginas the ratio of Earning before Interest Tax Depreciation & Amortization (EBITDA)and Revenue R:, EBITDA is defined as Revenue minus Opex. Both Opex and Revenue I can de-compose into a fixed and a variable part: O = Of + AOPU x U and R = Rf + ARPU x U with AOPU being the Average Opex per User, ARPU the Average (blended) Revenue per User and U the number of users. For the moment I will be ignoring the fixed part of the revenue and write R = ARPU x U. Further, the number of users can be written as with being the market share and M being the market size. So we can now write the margin as

with and .

being valid for

The Margin is not a linear function of the Subscriber Market Share (if anybody would have expected that) but relates to the Inverse of Market Share.

Still the Margin becomes larger as the market share grows with maximum achievable margin of as the market share equals 1 (i.e., Monopoly). We observe that even in a Monopoly there is a limit to how profitable such a business can be. It should be noted that this is not a constant but a function of how operationally efficient a given operator is as well as its market conditions. Furthermore, as the market share reduces towards zero .

Fixed Opex (of) per total subscriber market: This cost element is in principle related to cost structure that is independent on the amount of customers that a given mobile operator have. For example a big country with a relative low population (or mobile penetration) will have higher fixed cost per total amount of subscribers than a smaller country with a larger population (or mobile penetration). Fixed cost is difficult to change as it depends on the network and be country specific in nature. For an individual Operator the fixed cost (per total market subscribers) will be influenced by;

  • Coverage strategy, i.e., to what extend the country’s surface area will be covered, network sharing, national roaming vs. rural coverage, leased bandwidth, etc..
  • Spectrum portfolio, i.e, lower frequencies are more economical than higher frequencies for surface area coverage but will in general have less bandwidth available (i.e., driving up the number of sites in capacity limited scenarios). Only real exception to bandwidth limitations of low frequency spectrum would be the APT700 band (though would “force” an operator to deploy LTE which might not be timed right given specifics of the market).
  • General economical trends, lease/rental cost, inflation, salary levels, etc..

Average Variable Opex per User (ou): This cost structure element capture cost that is directly related to the subscriber, such as

  • Market Invest (i.e., Subscriber Acquisition Cost SAC, Subscriber Retention Cost SRC), handset subsidies, usage-related cost, etc..
  • Any other variable cost directly associated with the customer (e.g., customer facing functions in the operator organization).

This behavior is exactly what we observe in the presented Margin vs. Subscriber Market Share data and also explains why the data needs to be treated on a country by country basis. It is worthwhile to note that after the higher the market share the less incremental margin gain should be expected for additional market share.

The above presented profitability framework can be used to test whether a given mobile operator is market & operationally efficient compared to its peers.

The overall Margin dynamics is shown above Chart for the various settings of fixed and variable Opex as well as a given operators ARPU. We see that as the fixed Opex (in relation to the total subscriber market) increasing it will get more difficult to get EBITDA positive and increasingly more market share is required to reach a reasonable profitability targets. The following maps a 3 player market according with the profitability logic derived here:

What we first notice is that operators in the initial phase of what you might define as the “Market-share Capture Phase” are extremely sensitive to setbacks. A small loss of subscriber market share (i.e. 2%) can tumble the operator back into the abyss (i.e, 15% Margin setback) and wreck havoc to the business model. The profitability logic also illustrates that once an operator has reached Market-share maturity adding new subscribers is less valuable than to keep them. Even big market share addition will only result in little additional profitability (i.e., the law of diminishing returns).

The derived Profitability framework can be used also to illustrate what happens to the Margin in a market-wise steady situation (i.e., only minor changes to an operators market share) or what the Market Share needs to be to keep a given Margin or how cost needs to be controlled in the event that ARPU drops and we want to keep our margin and cannot grow market share (or any other market, profitability or cost-structure exercise for that matter);

  • Above chart illustrates Margin as a function of ARPU & Cost (fixed & variable) Development at a fixed market share here chosen to be 33%. The starting point is an ARPU ru of EUR25.8 per month, a variable cost per user ou assumed to be EUR15 and a fixed cost per total mobile user market (of) of EUR0.5. The first scenario (a Orange Solid Line) with an end of period margin of 32.7% assumes that ARPU reduces with 2% per anno, that the variable cost can be controlled and likewise will reduce with 2% pa. Variable cost is here assumed to increase with 3% on an annual basis. During the 10 year period it is assumed that the Operators market share remains at 33%. The second scenario (b Red Dashed Line) is essential same as (a) with the only difference that the variable cost remains at the initial level of EUR15 and will not change over time. This scenario ends at 21.1% after 10 Years. In principle it shows that Mobile Operators will not have a choice on reducing their variable cost as ARPU declines (again the trade-off between certainty of cost and risk/uncertainty of revenue). In fact the most successful mature mobile operators are spending a lot of efforts to manage & control their cost to keep their margin even if ARPU & Revenues decline.

  • The above chart illustrates what market share is required to keep the margin at 36% when ARPU reduces with 2% pa, fixed cost increases with 3% pa and the variable cost either (a Orange Solid Line) can be reduced with 2% in line with the ARPU decline or (b Red Solid Line) remains fixed at the initial level. In scenario (a) the mobile operator would need to grow its market share to 52% to main its margin at 36%. This will obviously be very challenging as this would be on the expense of other operators in this market (here assume to be 3). Scenario (b) is extremely dramatic and in my opinion mission impossible as it requires a complete 100% market dominance.

  • Above Chart illustrates how we need to manage & control my variable cost compared to the –2% decline pa in order to keep the Margin constant at 36% assuming that the Operator Subscriber Market Share remains at 33% over the period. The Orange Solid Line in the Chart shows the –2% variable cost decline pa and the Red Dashed Line the variable cost requirement to keep the margin at 36%.

The following illustrates the Profitability Framework as described above applied to a few Western European Markets. As this only serves as an illustration I have chosen to show older data (i..e, 2006). It is however very easy to apply the methodology to any country and the BoAML Global Wireless Matrix with its richness in data can serve as an excellent source for such analysis. Needless to say the methodology can be extended to assess an operators profitability sensitivity to market share and market dynamics in general.

The Charts below shows the Equalized Market Share which simply means the fair market share of operators, i.e., if I have 3 operators the fair or equalized market share would 1/3 (33.3%), in case of 4 operators it should be 25% and so forth, I am also depicting what I call the Max Margin Potential this is simply the Margin potential at 100% Market Share at a given set of ARPU (ru), AOPU (ou) and Fixed Cost (of) Level in relation to the total market.

  • Netherlands Chart: Equalized Market Share assumes Orange has been consolidated with T-Mobile Netherlands. The analysis would indicate that no more than ca. 40% Margin should be expected in The Netherlands for any of the 4 Mobile Operators. Note that for T-Mobile and Orange small increases in market share should in theory lead to larger margins, while KPN’s margin would be pretty much un-affected by additional market share.

  • Germany Chart: Shows Vodafone to slightly higher and T-Mobile Deutschland slight lower in Margin than the idealized Margin versus Subscriber Market share. At the time T-Mobile had almost exclusive leased lines and outsourced their site infrastructure while Vodafone had almost exclusively Microwaves and owned its own site infrastructure. The two new comers to the German market (E-Plus and Telefonica-O2) is trailing on the left side of the Equalized Market Share. At this point in time should Telefonica and E-Plus have merged one would have expected them eventually (post-integration) to exceed a margin of 40%. Such a scenario would lead to an almost equilibrium market situation with remaining 3 operators having similar market shares and margins.

 

 

 

 

 

Acknowledgement

I greatly acknowledge my wife Eva Varadi for her support, patience and understanding during the creative process of creating this Blog. I certainly have not always been very present during the analysis and writing.

Time Value of Money, Real Options, Uncertainty & Risk in Technology Investment Decisions

“We have met the Enemy … and he is us”

is how the Kauffman Foundations starts their extensive report on investments in Venture Capital Funds and their abysmal poor performance over the last 20 years. Only 20 out of 200 Venture Funds generated returns that beat the public-market equivalent with more than 3%. 10 of those were Funds created prior to 1995. Clearly there is something rotten in the state of valuation, value creation and management. Is this state of affairs limited only to portfolio management (i..e, one might have hoped a better diversified VC portfolio) is this poor track record on investment decisions (even diversified portfolios) generic to any investment decision made in any business? I let smarter people answer this question. Though there is little doubt in my mind that the quote “We have met the Enemy … and he is us” could apply to most corporations and the VC results might not be that far away from any corporation’s internal investment portfolio. Most business models and business cases will be subject to wishful thinking and a whole artillery of other biases that will tend to overemphasize the positives and under-estimate (or ignore) the negatives.The avoidance of scenario thinking and reference class forecasting will tend to bias investments towards the upper boundaries and beyond of the achievable and ignore more attractive propositions that could be more valuable than the idea that is being pursued.

As I was going through my archive I stumbled over an old paper I wrote back in 2006 when I worked for T-Mobile International and Deutsche Telekom (a companion presentation due on Slideshare). At the time I was heavily engaged with Finance and Strategy in transforming Technology Investment Decision Making into a more economical responsible framework than had been the case previously. My paper was a call for more sophisticated approaches to technology investments decisions in the telecom sector as opposed to what was “standard practice” at the time and in my opinion pretty much still i.

Many who are involved in techno-economical & financial analysis as well as the decision makers acting upon recommendations from their analysts are in danger of basing their decisions on flawed economical analysis or simply have no appreciation of uncertainty and risk involved. A frequent mistake made in decision making of investment options is ignoring one of the most central themes of finance & economics, the Time-Value-of-Money. An investment decision taken was insensitive to the timing of the money flow. Furthermore, investment decisions based on Naïve TCO are good examples of such insensitivity bias and can lead to highly in-efficient decision making. Naïve here implies that time and timing does not matter in the analysis and subsequent decision.

Time-Value-of-Money:

I like to get my money today rather than tomorrow, but I don’t mind paying tomorrow rather than today”.

Time and timing matters when it comes to cash. Any investment decision that does not consider timing of expenses and/or income has a substantially higher likelihood of being an economical in-efficient decision. Costing the shareholders and investors (a lot of) money. As a side note Time-Value-of-Money assumes that you can actually do something with the cash today that is more valuable than waiting for it at a point in the future. Now that might work well for Homo Economicus but maybe not so for the majority of the human race (incl. Homo Financius).

Thus, if I am insensitive to timing of payments it does not matter for example whether I have to pay €110 Million more for a system the first year compared to deferring that increment to the 5th year

Clearly wrong!

In the above illustration outgoing cash flow (CF) example the naïve TCO (i..e, total cost of ownership) is similar for both CFs. I use the word naïve here to represent a non-discounted valuation framework. Both Blue and Orange CFs represent a naïve TCO value of €200 Million. So a decision maker (or an analyst) not considering time-value-of-money would be indifferent to one or the other cash flow scenario. Would the decision maker consider time-value-of-money (or in the above very obvious case see the timing of cash out) clear it would be in favor of Blue. Further front-loaded investment decisions are scary endeavors, particular for unproven technologies or business decisions with a high degree of future unknowns, as the exposure to risks and losses are so much higher than a more carefully designed cash-out/investment trajectory following the reduction of risk or increased growth. When only presented with the (naïve) TCO rather than the cash flows, it might even be that some scenarios might be unfavorable from a naïve TCO framework but favorable when time-value-of-money is considered. The following illustrates this;

The Orange CF above amounts to a naïve TCO of €180 Million versus to the Blue’s TCO of €200 Million. Clearly if all the decision maker is presented with is the two (naïve) TCOs, he can only choose the Orange scenario and “save” €20 Million. However, when time-value-of-money is considered the decision should clearly be for the Blue scenario that in terms of discounted cash flows yields €18 Million in its favor despite the TCO of €20 Million in favor of Orange. Obviously, the Blue scenario has many other advantages as opposed to Orange.

 

When does it make sense to invest in the future?

 

Frequently we are faced with  technology investment decisions that require spending incremental cash now for a feature or functionality that we might only need at some point in the future. We believe that the cash-out today is more efficient (i.e., better value) than introducing the feature/functionality at the time when we believe it might really be needed..

 

Example of the value of optionality: Assuming that you have two investment options and you need to provide management with which of those two are more favorable.

 

Product X with investment I1: provides support for 2 functionalities you need today and 1 that might be needed in the future (i.e., 3 Functionalities in total).

Product Y with investment I2: provides support for the 2 functionalities you need today and 3 functionalities that you might need in the future (i.e., 5 Functionalities in total).

 

I1 < I2 and = I2I1 > 0

 

If, in the future, we need more than 1 additional functionality it clearly make sense to ask whether it is better upfront to invest in Product Y, rather than X and then later Y (when needed). Particular when Product X would have to be de-commissioned when introducing Product Y, it is quite possible that investing in Product Y upfront is more favorable. 

 

From a naïve TCO perspective it clearly better to invest in Y than X + Y. The “naïve” analyst would claim that this saves us at least I1 (if he is really clever de-installation cost and write-offs might be included as well as saving or avoidance cost) by investing in Y upfront.

 

Of course if it should turn out that we do not need all the extra functionality that Product Y provides (within the useful life of Product X) then we have clearly made a mistake and over-invested byand would have been better off sticking to Product X (i.e., the reference is now between investing in Product Y versus Product X upfront).

 

Once we call upon an option, make an investment decision, other possibilities and alternatives are banished to the “land of lost opportunities”.

 

Considering time-value-of-money (i.e., discounted cash-flows) the math would still come out more favorable for Y than X+Y, though the incremental penalty would be lower as the future investment in Product Y would be later and the investment would be discounted back to Present Value.

 

So we should always upfront invest in the future?

 

Categorically no we should not!

 

Above we have identified 2 outcomes (though there are others as well);

Outcome 1: Product Y is not needed within lifetime T of Product X.

Outcome 2: Product Y is needed within lifetime T of Product X.

 

In our example, for Outcome 1 the NPV difference between Product X and Product Y is -10 Million US$. If we invest into Product Y and do not need all its functionality within the lifetime of Product X we would have “wasted” 10 Million US$ (i.e., opportunity cost) that could have been avoided by sticking to Product X.

 

The value of Outcome 2 is a bit more complicated as it depends on when Product Y is required within the lifetime of Product X. Let’s assume that Product X useful lifetime is 7 years, i.e., after which period we would need to replace Product X anyway requiring a modernization investment. We assume that for the first 2 years (i.e., yr 2 and yr 3) there is no need for the additional functionality that Product Y offers (or it would be obvious to deploy up-front at least within this examples economics). From Year 4 to Year 7 there is an increased likelihood of the functionalities of Product X to be required.

 

In Outcome 2 the blended NPV is 3.0 Million US$ positive to deploy Product X instead of Product Y and then later Product X (i.e., the X+Y scenario) when it is required. After the 7th year we would have to re-invest in a new product and the obviously looking beyond this timeline makes little sense in our simplified investment example.

 

Finally if we assess that there is a 40% chance that the Product Y will not be required within the life-time of Product X, we have the overall effective NPV of our options would be negative (i.e., 40%(-10) + 3 = –1 Million). Thus we conclude it is better to defer the investment in Product Y than to invest in it upfront. In other words it is economical more valuable to deploy Product X within this examples assumptions.

 

I could make an even stronger case for deferring investing in Product Y: (1) if I can re-use Product X when I introduce Product Y, (2) if I believe that the price of Product Y will be much lower in the future (i..e, due to maturity and competition), or (3) that there is a relative high likelihood that the Product Y might become obsolete before the additional functionalities are required (e.g., new superior products at lower cost compared to Product Y). The last point is often found when investing into the very first product releases (i.e., substantial immaturity) or highly innovative products just being introduced. Moreover, there might be lower-cost lower-tech options that could provide the same functionality when required that would make investing upfront in higher-tech higher-cost options un-economical. For example, a product that provide a single targeted functionality at the point in time it is needed, might be more economical than investing in a product supporting 5 functionalities (of which 3 is not required) long before it is really required.

 

Many business cases are narrowly focusing on proving a particular point of view. Typically maximum 2 scenarios are compared directly, the old way and the proposed way. No surprise! The new proposed way of doing things will be more favorable than the old (why else do the analysis;-). While such analysis cannot be claimed to be wrong, it poses the danger of ignoring more valuable options available (but ignored by the analyst). The value of optionality and timing is ignored in most business cases.

 

For many technology investment decisions time is more a friend than an enemy. Deferring investing into a promise of future functionality is frequently the better value-optimizing strategy.

 

Rules of my thumb:

  • If a functionality is likely to be required beyond 36 months, the better decision is to defer the investment to later.
  • Innovative products with no immediate use are better introduced later rather than sooner as improvement cycles and competition are going to make such more economical to introduce later (and we avoid obsolescence risk).
  • Right timing is better than being the first (e.g., as Apple has proven a couple of times).

Decision makers are frequently betting on a future event (i..e, knowingly or unknowingly) will happen and that making an incremental investment decision today is more valuable than deferring the decision to later. Basically we deal with an Option or a Choice. When we deal with a non-financial Option we will call such a Real Option. Analyzing Real Options can be complex. Many factors needs to be considered in order to form a reasonable judgment of whether investing today in a functionality that only later might be required makes sense or not;

  1. When will the functionality be required (i.e., the earliest, most-likely and the latest).
  2. Given the timing of when it is required, what is the likelihood that something cheaper and better will be available (i.e., price-erosion, product competition, product development, etc..).
  3. Solutions obsolescence risks.

As there are various uncertain elements involved in whether or not to invest in a Real Option the analysis cannot be treated as a normal deterministic discounted cash flow. The probabilistic nature of the decision analysis needs to be correctly reflected in the analysis.

 

Most business models & cases are deterministic despite the probabilistic (i.e., uncertain and risky) nature they aim to address.

 

Most business models & cases are 1-dimensional in the sense of only considering what the analyst tries to prove and not per se alternative options.

 

My 2006 paper deals with such decisions and how to analyze them systematically and provide a richer and hopefully better framework for decision making subject to uncertainty (i.e., a fairly high proportion of investment decisions within technology).

Enjoy !

ABSTRACT

The typical business case analysis, based on discounted cash flows (DCF) and net-present valuation (NPV), inherently assumes that the future is known and that regardless of future events the business will follow the strategy laid down in the present. It is obvious that the future is not deterministic but highly probabilistic, and that, depending on events, a company’s strategy will be adopted to achieve maximum value out of its operation. It is important for a company to manage its investment portfolio actively and understand which strategic options generate the highest return on investment. In every technology decision our industry is faced with various embedded options, which needs to be considered together with the ever-prevalent uncertainty and risk of the real world. It is often overlooked that uncertainty creates a wealth of opportunities if the risk can be managed by mitigation and hedging. An important result concerning options is that the higher the uncertainty of the underlying asset, the more valuable could the related option become. This paper will provide the background for conventional project valuation, such as DCF and NPV. Moreover, it will be shown how a deterministic (i.e., conventional) business case easily can be made probabilistic, and what additional information can be gained with simulating the private as well as market-related uncertainties. Finally, real options analysis (ROA) will be presented as a natural extension of the conventional net-present value analysis. This paper will provide several examples of options in technology, such as radio access site-rollout strategies, product development options, and platform architectural choices.

INTRODUCTION

In technology, as well as in mainstream finance, business decisions are more often than not based on discounted cash flow (DCF) calculations using net-present value (NPV) as decision rationale for initiating substantial investments. Irrespective of the complexity and multitudes of assumptions made in business modeling the decision is represented by one single figure, the net present value. The NPV basically takes the future cash flows and discount these back to the present, assuming a so-called “risk –adjusted” discount rate. In most conventional analysis the “risk-adjusted” rate is chosen rather arbitrarily (e.g., 10%-25%) and is assumed to represent all project uncertainties and risks.The risk-adjusted rate should always as a good practice be compared with the weighted average cost of capital (WACC) and benchmarked against what Capital Asset Pricing Model (CAPM) would yield. Though in general the base rate will be set by your finance department and not per se something the analyst needs to worry too much about. Suffice to say that I am not a believer that all risk can be accounted for in the discount rate and that including risks/uncertainty into the cash flow model is essential.

 

It is naïve to believe that the applied discount rate can account for all risk a project may face.

 

In many respects the conventional valuation can be seen as supporting a one-dimensional decision process. DCF and NPV methodologies are commonly accepted in our industry and the finance community [1]. However, there is a lack of understanding of how uncertainty and risk, which is part of our business, impacts the methodology in use. The bulk of business cases and plans are deterministic by design. It would be far more appropriate to work with probabilistic business models reflecting uncertainty and risk. A probabilistic business model, in the hands of the true practitioner, provides considerable insight useful for steering strategic investment initiatives. It is essential that a proper balance is found between model complexity and result transparency. With available tools, such as Palisade Corporation’s @RISK Microsoft Excel add-in software [2], it is very easy to convert a conventional business case into a probabilistic model. The Analyst would need to converse with subject-matter experts in order to provide a reasonable representation of relevant uncertainties, statistical distributions, and their ranges in the probabilistic business model [3].

 

In this paper the word Uncertainty will be used as representing the stochastic (i.e., random) nature of the environment. Uncertainty as concept represents events and external factors, which cannot be directly controlled. The word Volatility will be used interchangeably with uncertainty. With Risk is meant the exposure to uncertainty, e.g., uncertain cash-flows resulting in out-of-money and catastrophic business failure. The total risk is determined by the collection of uncertain events and Management’s ability to deal with these uncertainties through mitigation and “luck”. Moreover, the words Option and Choice will also be used interchangeably throughout this paper.

 

Luck is something that never should be underestimated.

 

While working on the T-Mobile NL business case for the implementation of Wireless Application Protocol (WAP) for circuit switched data (CSD), a case was presented showing a 10% chance of losing money (over a 3 year period). The business case also showed an expected NPV of €10 Million, as well as a 10% chance of making more than €20 Million over a 3 year period. The spread in the NPV, due to identified uncertainties, were graphically visualized.

 

Management, however, requested only to be presented with the “normal” business case NPV as this “was what they could make a decision upon”. It is worthwhile to understand that the presenters made the mistake to make the presentation to Management too probabilistic and mathematical which in retrospect was a wrong approach [4]. Furthermore, as WAP was seen as something strategically important for long-term business survival, moving towards mobile data, it is not conceivable that Management would have turned down WAP even if the business case had been negative.

In retrospect, the WAP business case would have been more useful if it had pointed out the value of the embedded options inherent in the project;

  1. Defer/delay until market conditions became more certain.
  2. Defer/delay until GPRS became available.
  3. Outsource service with option to in-source or terminate depending on market conditions and service uptake.
  4. Defer/delay until technology becomes more mature, etc..

Financial “wisdom” states that business decisions should be made which targets the creation of value [5]. It is widely accepted that given a positive NPV, monetary value will be created for the company therefore projects with positive NPV should be implemented. Most companies’ investment means are limited. Innovative companies often are in a situation with more funding demand than available. It is therefore reasonable that projects targeting superior NPVs should be chosen. Considering the importance and weight businesses associate with the conventional analysis using DCF and NPV it worthwhile summarizing the key assumptions underlying decisions made using NPV: 

  • As a Decision is made, future cash flow streams are assumed fixed. There is no flexibility as soon as a decision has been made, and the project will be “passively” managed.
  • Cash-flow uncertainty is not considered, other than working with a risk-adjusted discount rate. The discount rate is often arbitrarily chosen (between 9%-25%) reflecting the analyst’s subjective perception of risk (and uncertainty) with the logic being the higher the discount rate the higher the anticipated risk (note: the applied rate should be reasonably consistent with Weighted Average Cost of Capital  and Capital Asset Pricing Model (CAPM)).
  • All risks are completely accounted for in the discount rate (i.e., which is naïve)
  • The discount rate remains constant over the life-time of the project (i.e., which is naïve).
  • There is no consideration of the value of flexibility, choices and different options.
  • Strategic value is rarely incorporated into the analysis. It is well known that many important benefits are difficult (but not impossible) to value in a quantifiable sense, such as intangible assets or strategic positions. If a strategy cannot be valued or quantified it should not be pursued.
  • Different project outcomes and the associated expected NPVs are rarely considered.
  • Cash-flows and investments are discounted with a single discount rate assuming that market risk and private (company) risk is identical. Correct accounting should use the risk-free rate for private risk and cash-flows subject to market risks should make use of market risk-adjusted discount rate.

In the following several valuation methodologies will be introduced, which build upon and extend the conventional discounted cash flow and net-present value analysis, providing more powerful means for decision and strategic thinking.

 

TRADITIONAL VALUATION

The net-present value is defined as the difference between the values assigned to a given asset, the cash-flows, and the cost and capital expenditures of operating the asset. The traditional valuation approach is based on the net-present value (NPV) formulation [6]

T is the period during which the valuation is considered, Ct is the future cash flow at time t, rram is the risk-adjusted discount rate applied to market-related risk, It is the investment cost at time t, and rrap is the risk-adjusted-discount rate applied to private-related risk. In most analysis it is customary to assume the same discount rate for private as well as market risk as it simplifies the valuation analysis. The “effective” discount rate r* is often arbitrarily chosen. The I0 is the initial investment at time t=0, and Ct* = Ct – It (for t>0) is the difference between future cash flows and investment costs. The approximation (i.e., ≈ sign) only holds in the limit where the rate rrap is close to rram. The private risk-adjusted rate is expected to be lower than the market risk-adjusted rate. Therefore, any future investments and operating costs will weight more than the future cash flows. Eventually value will be destroyed unless value growth can be achieved. It is therefore important to manage incurred cost, and at the same time explore growth aggressively (at minimum cost) over the project period. Assuming a risk-adjusted or effective rate for both market and private risk investment, cost and cash-flows could lead to a even serious over-estimation of a given project’s value. In general, the private risk-adjusted rate rrap would be between the risk-free rate and the market risk-adjusted discount rate rram.

 

EXAMPLE 1: An initial network investment of 20 mio euro needs to be committed to provide a new service for the customer base. It is assumed that sustenance investment per year amounts to 2% of the initial investment and that operations & maintenance is 20% of the accumulated investment (50% in initial year). Other network cost, such as transmission (assumes centralized platform solution) increases with 10% per year due to increased traffic with an initial cost of 150 thousand. The total network investment and cost structure should be discounted according with the risk-free rate (assumed to be 5%). Market assumptions: s-curve consistent growth assumed with a saturation of 5 Million service users after approximately 3 years. It has been assumed that the user pays 0.8 euro per month for the service and that the service price decreases with 10% per year. Cost of acquisition assumed to be 1 euro per customer, increasing with 5% per year. Other market dependent cost assumed initially to be 400 thousand and to increase with 10% per year. It is assumed that the project is terminated after 5 years and that the terminal value amounts to 0 euro. PV stands for present value and FV for future value. The PV has been discounted back to year 0. It can be seen from the table that the project breaks-even after 3 years. The first analysis presents the NPV results (over a 5 year period) when differentiating between private (private risk-adjusted rate) and market (market risk-adjusted rate) risk taking, a positive NPV of 26M is found. This should be compared with the standard approach assuming an effective rate of 12.5%, which (not surprisingly) results in a positive NPV of 46M. The difference between the two approaches amounts to about 19M.

.

Example above compares the approach of using an effective discount rate r* with an analysis that differentiates between private rrap and market risk rram in the NPV calculation. The example illustrates a project valuation example of introducing a new service. The introduction results in network investments and costs in order to provide and operate the service.  Future cash-flows arise from growth of customer base (i.e., service users), and is offset by market related costs. All network investments and costs are assumed to be subject to private risk and should be discounted with the risk-free rate. The market-related cost and revenues are subject to market risk and the risk-adjusted rate should be used [7]. Alternatively, all investment, costs and revenues can be treated with an effective discount rate. As seen from the example, the difference between the two valuation approaches can be substantial:

  • NPV = €26M for differentiated market and private risk, and
  • NPV = €46M using an effective discount rate (e.g., difference of €20M assuming the following discount rates rram = 20%, rrap =5%, r* = 12.5%). Obviously, as rram –> r* and rrap –> r* , the difference in the two valuation approaches will tend to zero. 

 

UNCERTAINTY, RISK & VALUATION

The traditional valuation methodology presented in the previous section makes no attempt to incorporate uncertainties and risk other than the effective discount-rate r* or risk-adjusted rates rram/rap. It is inherent in the analysis that cash-flows, as well as the future investments and cost structure, are assumed to be certain. The first level of incorporating uncertainty into the investment analysis would be to define market scenarios with an estimated (subjective) chance of occurring. A good introduction to uncertainty and risk modeling is provided in the well-written book by D. Vose [8], S.O. Sugiyama’s training notes [3] and S. Beninga’s “Financial Modeling” [7].

 

The Business Analyst working on the service introduction, presented in Example 1, assesses that there are 3 main NPV outcomes for the business model; NPV1= 45, NPV2= 20 and NPV3= -30.  The outcomes have been based on 3 different market assumptions related to customer uptake: 1. Optimistic, 2. Base and 3. Pessimistic. The NPVs are associated with the following chances of occurrence: P1 = 25%, P2 = 50% and P3 = 25%.

 

What would the expected net-present value be given the above scenarios?

 

The expected NPV (ENPV) would be ENPV=P1×NPV1+ P2×NPV2+ P3×NPV3=25%×45+50%×20+25%×(-30) =14. Example 2 (below) illustrates the process of obtaining the expected NPV.

Example 2: illustrates how to calculate the expected NPV (ENPV) when 3 NPV outcomes have been identified resulting from 3 different customer uptake scenarios. The expected NPV calculation assumes that we do not have any flexibility to avoid any of the 3 outcomes. The circular node represents a chance node yielding the expected outcome given the weighted NPVs.

 

In general the expected NPV can be written as

,where N is number of possible NPV outcomes, NPVi is the net present value of the ith outcome and Pi is the chance that the ith outcome will occur.  By including scenarios in the valuation analysis, the uncertainty of the real world is being captured. The risk of overestimating or underestimating a project valuation is thereby minimized. Typically, the estimation of P, which is the chance or probability, for a particular outcome is based on subjective “feeling” of the Business Analyst, who obviously still need to build a credible story around his choices of likelihood for the scenarios in questions. Clearly this is not a very satisfactory situation as all kind of heuristic biases are likely to influence the choice of a given scenarios likelihood. Still it is clearly more realistic than a purely deterministic approach with only one locked-in happening.

 

Example 3 shows various market outcomes used to study the uncertainty of market conditions upon the net-present value of Example 1and the project valuation subject these uncertainties. The curve represented by the thick solid line and open squares is the base market scenario used in Example 1, while the other curves represent variations to the base case.  Various uncertainties of the customer growth have been explored. An s-curve (logistic function) approach has been used to model the customer uptake of for the studied service: , t is time period, Smax is the maximum expected number of customer, be determines the slope in the growth phase, and (1/a) is the years to reach the mid-point of the S-curve. The function models the possible decline in customer base, with c being the rate of decline in the market share, and td the period when the decline sets in. Smax has been varied between 2.5 and 6.25 Million customers, with an average of 5.0 Million, b was chosen to be 50 (arbitrarily), (1/a) was varied between 1/3 and 2 (year), with a mean of 0.5 (year). In modeling the market decline, the rate of decline c was varied between 0% and 25% years, with a chosen mean value of 10%, and the td was varied between 0 and 3 years with a mean of 2 years before market decline starts. In all cases a so-called pert distribution was used to model the parameter variance. Instead of running a limited number of scenarios as shown in Example 2 (3 outcomes), a Monte Carlo (MC) simulation is carried out sampling several thousands of possible outcomes.

 

As already discussed a valuation analysis often involves many uncertain variables and assumptions. In the above Example 3 different NPV scenarios had been identified, which resulted from studying the customer uptake. Typically, the identified uncertain input variables in a simplified scenario-sensitivity approach would each have at least three possible values; minimum (x), base-line or most-likely (y), and maximum (z). For every uncertain input variable the Analyst has identified a variation, i.e., 3 possible variations. For an analysis with 2 uncertain input variables, each with variation, it is not difficult to show that the outcome is 9 different scenario-combinations, for 3 uncertain input variables the result is 72 scenario-combinations, 4 uncertain input variables results in 479 different scenario permutations, and so forth. In complex models containing 10 or more uncertain input variables, the number of combinations would have exceeded 30 Million permutations [9]. Clearly, if 1 or 2 uncertain input variables have been identified in a model the above presented scenario-sensitivity approach is practical. However, the range of possibilities quickly becomes very large and the simple analysis breaks down. In these situations the Business Analyst should turn to Monte Carlo [10] simulations, where a great number of outcomes and combinations can be sampled in a probabilistic manner and enables proper statistical analysis. Before the Analyst can perform an actual Monte Carlo simulation, a probability density function (pdf) needs to be assigned to each identified uncertain input variable and any correlation between model variables needs to be addressed. It should be emphasized that with the help of subject-matter experts, an experienced Analyst in most cases can identify the proper pdf to use for each uncertain input variable. A tool such as Palisade Corporation’s @RISK toolbox [2] for MS Excel visualizes, supports and greatly simplifies the process of including uncertainty into a deterministic model, and efficiently performs Monte Carlo simulations in Microsoft Excel.

 

Rather than guessing a given scenarios likelihood, it is preferable to transform the deterministic scenarios into one probabilistic scenario. Substituting important scalars (or drivers) with best practice probability distributions and introduce logical switches that mimic choices or options inherent in different driver outcomes. Statistical sampling across simulated outcomes will provide an effective (or blended) real option value.

 

In Example 1a standard deterministic valuation analysis was performed for a new service and the corresponding network investments. The inherent assumption was that all future cash-flows as well as cost-structures were known. The analysis yielded a 5-year NPV of 26 mio (using the market-private discount rates). This can be regarded as a pure deterministic outcome. The Business Analyst is requested by Management to study the impact on the project valuation incorporating uncertainties into the business model. Thus, the deterministic business model should be translated into a probabilistic model. It is quickly identified that the market assumptions, the customer intake, is an area which needs more analysis. Example 3shows various possible market outcomes. The reference market model is represented by the thick-solid line and open squares. The market outcome is linked to the business model (cash-flows, cost and net-present value). The deterministic model in Example 1 has now been transformed into a probabilistic model including market uncertainty.

Example 4: shows the impact of uncertainty in the marketing forecast of customer growth on the Net Present Value (extending Example 1). A Monte Carlo (MC) simulation was carried out subject to the variations of the market conditions (framed box with MC in right side) described above (Example 2) and the NPV results were sampled. As can be seen in the figure above an expected mean NPV of 22M was found with a standard deviation of 16M. Further, analysis reveals a 10% probability of loss (i.e., NPV £ 0 euro) and an opportunity of up to 46M. Charts below (Example 4b and 4c) show the NPV probability density function and integral (probability), respectively. 

Example 4b                                                                        Example 4c

Example 4 above summarizes the result of carrying out a Monte Carlo (MC) simulation, using @RISK [2], determining the risks and opportunities of the proposed service and therefore obtaining a better foundation for decision making. In the previous examples the net-present value was represented as a single number; €26M in Example 1 and an expect NPV of €14M in Example 2. In Example 4, the NPV is far richer (see the probability charts of NPV at the bottom of the page) – first note that the mean NPV of €22M agree well with Example 1. Moreover, the Monte Carlo analysis shows the project down-side, that there is a 10% chance of ending up with a poor investment, resulting in value destruction. The opportunity or upside is a chance (i.e., 5%) of gaining more than €46M within a 5-year time-horizon. The project risk profile is represented with the NPV standard deviation, i.e. the project volatility, of €16M. It is Management’s responsibility to weight the risk, downside as well as upside, and ensure that proper mitigation will be considered to reduce the impact of the project downside and potential value destruction.

 

The presented valuation methodologies so far do not consider flexibility in decision making. Once an investment decision has been taken investment management is assumed to be passive. Thus, should a project turn out to destroy value, which is inevitable if revenue growth becomes limited compared to the operating cost, Management is assumed not to terminate or abandon this project. In reality active Investment Management and Management Decision Making does consider options and their economical and strategic value. In the following a detailed discussion on the valuation of options and the impact on decision making are presented. The Real options analysis (ROA) will be introduced as a natural extension of probabilistic cash flow and net present value analysis. It should be emphasized that ROA is based on some advanced mathematical, as well as statistical concepts, which will not be addressed in this work.

However, it is possible to get started on ROA with proper re-arrangement of the conventional valuation analysis, as well as incorporating uncertainty where ever appropriate. In the following the goal is to get the reader introduced to thinking about the value of options.

 

REAL OPTIONS & VALUATION

An investment option can be seen as a decision flexibility, which depending upon uncertain conditions, might be realized. It should be emphasized, that as with a financial option, it is at the investor’s discretion to realize an option. Any cost or investment for the option itself can be viewed as the premium a company has to pay in order to obtain the option. For example, a company could be looking at an initial technology investment, with the option later on to expand should market conditions be favorable for value growth. Exercising the option, or making the decision to expand the capacity, results in a commitment of additional cost and capital investments – the “strike price” – into realizing the plan/option. Once the option to expand has been exercised, the expected revenue stream becomes the additional value subject to private and market risks. In every technology decision a decision-maker is faced with various options and would need to consider the ever-prevalent uncertainty and risk of real-world decisions.

 

In the following example, a multinational company is valuing a new service with the idea to commercially launch in all its operations. The cash-flows, associated with the service, are regarded as highly uncertain, and involve significant upfront development cost and investments in infrastructure to support the service. The company studying the service is faced with several options for the initial investment as well as future development of the service. Firstly, the company needs to make the decision to launch the service in all countries in which it is based, or to start-up in one or a few countries to test the service idea before committing to a full international deployment, investing in transport and service capacity. The company also needs to evaluate the architectural options in terms of platform centralization versus de-centralization, platform supplier harmonization or commit to a more-than-one-supplier strategy. In the following, options will be discussed in relation to the service deployment as well as the platform deployment, which supports the new service. In the first instance the Marketing strategy defines a base-line scenario in which the service is launched in all its operations at the same time. The base-line architectural choice is represented by a centralized platform scenario placed in one country, providing the service and initial capacity to the whole group.

.

Platform centralization provides for an efficient investment and resourcing; instead of several national platform implementation projects only one country focuses its resources. However, the operating costs might be higher due to need for international leased transmission connectivity to the centralized platform. Due to the uncertainty in the assumed cash-flows, arising from market uncertainties, the following strategy has been identified; The service will be launched initially in a limited number of operations (one or two) with the option to expand should the service be successful (option 1), or should the service fail to generate revenue and growth potential an option to abandon the service after 2 years (option 2). The valuation of the identified options should be assessed in comparison with the base-line scenario of launching the service in all operations. It is clear that the expansion option (option 1) leads to a range of options in terms of platform expansion strategies depending on the traffic volume and cost of the leased international transmission (carrying the traffic) to the centralized platform.

 

For example, if the cost of transmission exceeds the cost of operating the service platform locally an option to locally deploy the service platform is created. From this example it can be seen that by breaking up the investment decisions into strategic options the company has ensured that it can abandon should the service fail to generate the expected revenue or cash-flows, reducing loses and destruction of wealth. However, more importantly the company, while protecting itself from the downside, has left open the option to expand at the cost of the initial investment. It is evident that as the new service has been launched and cash-flows start being generated (or lack of appropriate cash-flows) the company gains more certainty and better grounds for making decisions on which strategic options should be exercised.

 

In the previous example, an investment and its associated valuation could be related to the choices which come naturally out of the collection of uncertainties and the resulting risk. In the literature (e.g., [11], [12]) it has been shown that conventional cash-flow analysis, which omits option valuation, tends to under-estimate the project value [13]. The additional project value results from identifying inherent options and valuing these options separately as strategic choices that can be made in a given time-horizon relevant to the project. The consideration of the value of options in the physical world closely relates to financial options theory and treatment of financial securities [14]. The financial options analysis relates to the valuation of derivatives [15] depending on financial assets, whereas the analysis described above identifying options related to physical or real assets, such as investment in tangible projects, is defined as real options analysis (ROA). Real options analysis is a fairly new development in project valuation (see [16], [17], [18], [19], [20], and [21]), and has been adopted to gain a better understanding of the value of flexibility of choice.

 

One of the most important ideas about options in general and real options in particular, is that uncertainty widens the range of potential outcomes. By proper mitigation and contingency strategy the downside of uncertainty can be significantly reduced, leaving the upside potential. Uncertainty, often feared by Management, can be very valuable, provided the right level of mitigation is exercised. In our industry most committed investments involve a high degree of uncertainty, in particular concerning market forces and revenue expectations, but also technology-related uncertainty and risk is not negligible. The value of an option, or strategic choice, arises from the uncertainty and related risk that real-world projects will be facing during their life-time. The uncertain world, as well as project complexity, results in a portfolio of options, or choice-path, a company can choose from. It has been shown that such options can add significant value to a project – however, presently options are often ignored or valued incorrectly [1121]. In projects, which are inherently uncertain, the Analyst would look for project-valuable options such as, for example:

  1. Defer/Delay – wait and see strategy (call option)
  2. Future growth/ Expand/Extend – resource and capacity expansion (call option)
  3. Replacement – technology obsolescence/end-of-life issues (call option)
  4. Introduction of new technology, service and/or product (call option)
  5. Contraction – capacity decommissioning (put option)
  6. Terminate/abandon – poor cash-flow contribution or market obsolescence (put option)
  7. Switching options – dynamic/real-time decision flexibility (call/put option)
  8. Compound options – phased and sequential investment (call/put option)

It is instructive to consider a number of examples of options/flexibilities which are representative for the mobile telecommunications industry. Real options or options on physical assets can be divided in to two basic types – calls and puts. A call option gives, the holder of the option, the right to buy an asset, and a put option provides the holder with the right to sell the underlying asset.

 

First, the call option will be illustrated with a few examples: One of the most important options open to management is the option to Defer or Delay (1) a project. This is a call option, right to buy, on the value of the project. The defer/delay option will be addressed at length later in this paper. The choice to Expand (2) is an option to invest in additional capacity and increase the offered output if conditions are favorable. This is defined as a call option, i.e., the right to buy or invest, on the value of the additional capacity that could enable extra customers, minutes-of-use, and of course additional revenue. The exercise price of the call option is the investment and additional cost of providing the additional capacity discounted to the time of the option exercise. A good example is the expansion of a mobile switching infrastructure to accommodate an increase in the customer base. Another example of expansion could be moving from platform centralization to de-centralization as traffic grows and the cost of centralization becomes higher than the cost of decentralizing a platform. For example, the cost of transporting traffic to a centralized platform location could, depending on cost-structure and traffic volume, become un-economical. Moreover, Management is often faced with the option to extend the life of an asset by re-investing in renewal – this choice is a so-called Replacement Option (3). This is a call option, the right to re-invest, on the assets future value. An example could be the renewal of the GSM base-transceiver stations (BTS), which would extend the life and adding additional revenue streams in the form of options to offer new services and products not possible on the older equipment. Furthermore, there might be additional value in reducing operational cost of old equipment, which typically would have higher running cost, than with new equipment. Terminate/Abandonment (5) in a project is an option to either sell or terminate a project. It is a so-called put option, i.e., it gives the holder the right to sell, on the projects value. The strike price would be the termination value of the project reduced by any closing-down costs.  This option mitigates the impact of a poor investment outcome and increases the valuation of the project. A concrete example could be the option to terminate poorly revenue generating services or products, or abandon a technology where the operating costs results in value destruction. The growth in cash-flows cannot compensate the operating costs. Contraction choices  (6) are options to reduce the scale of a project’s operation. This is a put option, right to “sell”, on the value of the lost capacity. The exercise price is the present value of future cost and investments saved as seen at the time of exercising the option. In reality most real investment projects can be broken up in several phases and therefore also will consist of several options and the proper investment and decision strategy will depend on the combination these options. Phased or sequential investment strategies often include Compounded Options (8), which are a series of options arising sequentially.

 

The radio access network site-rollout investment strategy is a good example of how compounded options analysis could be applied. The site rollout process can be broken out in (at least) 4 phases: 1. Site identification, 2. Site acquisition, 3. Site preparation (site build/civil work), and finally 4. Equipment installation, commissioning and network integration. Phase 2 depends on phase 1, phase 3 depends on phase 2, and phase 4 depends on phase 3 – a sequence of investment decisions depending on the previous decision, thus the anatomy of the real options is that of Compound Options (8) . Assuming that a given site location has been identified and acquired (call option on the site lease), which is typically the time-consuming and difficult part of the overall rollout process; the option to prepare the site emerges (Phase 3). This option, also a call option, could depend on the market expectations and the competitions strategy, local regulations and site-lease contract clauses. The flexibility arises from deferring/delaying the decision to commit investment to site preparation. The decision or option time-horizon for this deferral/delay option is typically set by the lease contract and its conditions. If the option expires the lease costs have been lost, but the value arises from not investing in a project that would result in negative cash-flow.  As market conditions for the rollout technology becomes more certain, higher confidence in revenue prospects, a decision to move to site preparation (Phase 3) can be made. In terms of investment management after Phase 3 has been completed there is little reason not to pursue Phase 4 and install and integrate the equipment enabling service coverage around the site location. If at the point of Phase 3 the technology or supplier choice still remains uncertain it might be a valuable option to await (deferral/delay option) a decision on supplier and/or technology to be deployed. In the site-rollout example described other options can be identified, such as abandon/terminate option on the lease contract (i.e., a put option). After Phase 4 has been completed there might come a day where an option to replace the existing equipment with new and more efficient / economical equipment arises.  It might even be interesting to consider the option value of terminating the site altogether and de-install the equipment. This could happen when operating costs exceeds the cash-flow. It should be noted that the termination option is quite dramatic with respect to site-rollout as this decision would disrupt network coverage and could aggress existing customers. However, the option to replace the older technology and maybe un-economical services with a new and more economical technology-service option might prove valuable. Most options are driven by various sources of uncertainty. In the site-rollout example, uncertainty might be found with respect to site-lease cost, time-to-secure-site, inflation (impacting the site-build cost), competition, site supply and demand, market uncertainties, and so forth

 

Going back to Example 1 and Example 4, the platform subject-matter expert (often different from the Analyst) has identified that if the customer base exceeds 4 Million customers and expansion of €10M will be needed. Thus, the previous examples underestimate the potential investments in platform expansion due to customer growth. Given that the base-line market scenario does identify that that this would be the case in the 2nd year of the project the €10M is included in the deterministic conventional business case for the new service. The result of including the €10M in the 2nd year of Example 1 is that the NPV drops from €26M to €8.7M (∆NPV minus €17.6M). Obviously, the conventional Analyst would stop here and still be satisfied that this seems to be a good and solid business case. The approach of Example 4 is applied to the new situation, subject to the same market uncertainty given in Example 3. From the Monte Carlo simulation, it is found that the NPV mean-value only is €4.7M. However, the downside is that the probability of loss (i.e., an NPV less than 0) now is 38%. It is important to realize that in both examples is the assumption that there is no choice or flexibility concerning the €10M investment; the investment will be committed in year two. However, the project has an option – the option to expand provided that the customer base exceeds 4 Million customers. Time wise it is a flexible option in the sense that if the project expected lifetime is 5 years, any time within this time-horizon is there a possibility that the customer base exceeds the critical mass for platform expansion.

Example 5: Shows the NPV valuation outcome when an option to expand is included in the model of Example 4. The €10M  is added if and only if the customer base exceeds 4 Million.

In the above Example 5  the probabilistic model has been changed to add €10M if and only if the customer base exceeds 4 Million. Basically, the option of expansion is being simulated. Treating the expansion as an option is clearly valuable for the business case, as the NPV mean-value has increased from €4.7M to €7.6M. In principle the option value could be taken to €2.9M. It is worthwhile noticing that the probability of loss (from 38% to 25%) has also been reduced by allowing for the option not to expand the platform if the customer base target is not achieved. It should be noted that although the example does illustrate the idea of options and flexibility it is not completely in line with a proper real options analysis.

Example 6 Shows the different valuation outcomes depending on whether the €10M platform expansion (when customer base exceeds 4 Million) is considered as un-avoidable (i.e., the “Deterministic No Option” and “Probabilistic No Option”) or as an option or choice to do so (“Probabilistic with Option”). It should be noted that the additional €3M in difference between “Probabilistic No Option” and “Probabilistic With Option” can be regarded as an effective option value, but it does not necessarily agree with a proper real-option valuation analysis of the option to expand. Another difference in the two probabilistic models is that in the model with option to expand an expansion can happen any year if customer base exceeds 4 Million, while the No option model only considers the expansion in year 2 where according with the marketing forecast the base exceeds the 4 Million. Note that Example 6 is different in assumptions than Example 1 and Example 4 as these do not include the additional €10M.

 

Example 6 above summarizes the three different approaches of valuation analysis; deterministic (essential 1-dimensional), probabilistic with options, and probabilistic including value options.

The investment analysis of real options as presented in this paper is not a revolution but rather an evolution of the conventional cash-flow and NPV analysis. The approach to valuation is first to understand and proper model the base-line case. After the conventional analysis has been carried out, the analyst, together with subject-matter experts, should determine areas of uncertainty by identifying the most relevant uncertain input parameters and their variation-ranges. As described in the previous section the deterministic business model is being transformed into a probabilistic model. The valuation range, or NPV probability distribution, is obtained by Monte Carlo simulations and the opportunity and risk profile is analyzed. The NPV opportunity-risk profile will identify the need for mitigation strategies, which in itself result in studying the various options inherent in the project. The next step in the valuation analysis is to value the identified project or real options. The qualitatively importance of considering real options in investment decisions has been provided in this paper. It has been shown that conventional investment analysis, represented by net-present value and discounted cash-flow analysis, gives only one side of the valuation analysis. As uncertainty is the “farther” of opportunity and risk it needs to be considered in the valuation process. Are identified options always valuable? The answer to that question is no – if we have certainty about an option movement is not in our favor then the option would be valuable. Think for example of considering a growth option at the onset of severe recession.

 

The real options analysis is often presented as being difficult and too mathematical; in particular

due to the involvement of the partial differential equations (PDE) that describes the underlying uncertainty (continuous-time stochastic processes, involvement of Markov processes, diffusion processes, and so forth). Studying PDEs are the basis for the ground-breaking work of the Black-Scholes-Merton [22] [23] on option pricing, which provided the financial community with an analytical expression for valuing financial options. However, “heavy” mathematical analysis is not really needed for getting started on real option.

 

Real options are a way of thinking, identifying valuable options in a project or potential investment that could create even more value by considering as an option instead of a deterministic given.

 

Furthermore, Cox et al [24] proposed a simplified algebraic approach, which involves so-called binominal trees representing price, cash-flow, or value movements in time. The binomial approach is very easy to understand and implement, resembling standard decision tree analysis, and visually easy to generate, as well as algebraically straightforward to solve.

 

SUMMARY

Real options are everywhere where uncertainty governs investment decisions. It should be clear that uncertainty can be turned into a great advantage for value growth providing proper contingencies are taken for reducing the downside of uncertainty – mitigating risk.  Very few investment decisions are static, as conventional discounted cash-flow analysis otherwise might indicate, but are ever changing due to changes in market conditions (global as well as local), technologies, cultural trends, etc. In order to continue to create wealth and value for the company value growth is needed and should force a dynamic investment management process that continuously looks at the existing as well as future valuable options available for the industry. It is compelling to say that a company’s value should be related to its real-options portfolio, and its track record in mitigating risk, and achieving the uncertain up-side of opportunities.

 

ACKNOWEDGEMENT

I am indebted to Sam Sugiyama (President & Founder of EC Risk USA & Europe) for taking time out from a very busy schedule and having a detailed look at the content of our paper. His insights and hard questions have greatly enriched this work. Moreover, I would also like to thank Maurice Ketel (Manager Network Economics), Jim Burke (who in 2006 was Head of T-Mobile Technology Office) and Norbert Matthes (who in 2007 was Head of Network Economics T-Mobile Deutschland) for their interest and very valuable comments and suggestions.

___________________________

APPENDIX – MATHEMATICS OF VALUE.

Firstly we note that the Future Value FV (of money) can be defined as the present Value PV (of money) times a relative increase given by an effective rate r* (i.e., that represents the change of money value between time periods), reflecting value increase or of course decrease over a cause of time t;

  

So the Present Value given we know the Future Value would be

For a sequence (or series) of future money flow we can write the present value as 

If r* is positive time-value-of-money follows naturally, i.e., money received in the future is worth less than today. It is a fundamental assumption that you can create more value with your money today than waiting to get them in the future (i.e., not per se right for majority of human beings but maybe for Homo Economicus).

First the sequence of future money value (discounted to the present) has the structure of a geometric series: , with yk+1 = g*yk (i.e., g* representing the change in y between two periods k and k+1).

Define and note that, thus in this framework we have that (note: I am doing all kind of “naughty” simplifications to not get too much trouble with the math).

The following relation is easy to realize:

, subtract the two equations from each other and the result is

. In the limit where n goes toward infinity (¥), providing that, it can be seen that .

It is often forgotten that this only is correct if and only if or in other words, if the discount rate (to present value) is higher than the future value growth rate.

You might often hear you finance folks (or M&A jockeys) talk about Terminal Value (they might also call it continuation value or horizon value … for many years I called it Termination Value … though that’s of course slightly out of synch with Homo Financius not to be mistaken for Homo Economicus :-).

with TV representing the Terminal Value and

NPV representing the net present value as calculated over a well-defined time span T.

 

I always found the Terminal Value fascinating as the size (matters?) or relative magnitude can be very substantial and frequently far greater than the NPV in terms of “value contribution” to the present value. Of course we do assume that our business model will survive to “Kingdom Come”. Appears to be a slightly optimistic assumptions (n’est pas mes amis? :-). We also assume that everything in the future is defined by the last year of cash-flow, the cash flow growth rate and our discount rate (hmmm don’t say that Homo Financius isn’t optimistic). Mathematically this is all okay (if ), economically maybe not so. I have had many and intense debates with past finance colleagues about the validity of Terminal Value. However, to date it remains a fairly standard practice to joggle up the enterprise value of a business model with a “bit” of Terminal Value.

Using the above (i.e., including our somewhat “naughty” simplifications)

It is easy to see why TV can be a very substantial contribution to the total value of a business model. The denominator (r-g) tends to be a lot smaller than 1 (i.e., note that always we have g<r) and though “blows” up the TV contribution to the present value (even when g is chosen to be zero).

Let’s evaluate the impact on uncertainty of the interest rate x, first re-write the NPV formula:

, yk is the cash-flow of time k (for the moment it remains unspecified), from

error/uncertainty propagation it is known that the standard deviation can be written as, where z=f(x,y,z,…) is a multi-variate function. Identifying the terms in the NPV formula is easy: z = Vn and

In the first approximation assume that x is the uncertain parameter, while yk is certain (i.e., ∆yk=0), then the following holds for the NPV standard deviation:

,

in the special case where yk is constant for all k’s,. It can be shown (similar analysis as above) that

with .

In the limit where n goes toward infinity, applying l’Hospital’s rule showing that , the following holds for propagating uncertainty/errors in the NPV formula:

Let’s take a numerical example, y=1, the interest rate x = 10% and the uncertainty/error is assumed to be no more than ∆x=3% (7%£ x £13%), assume that n®¥ (infinite time-horizon). Using the formula derived above NPV¥=11 and ∆NPV¥=±3.30 or a 30% error on estimated NPV. If the assumed cash-flows (i.e., yk) also uncertain the error will even be greater than 30%. The above analysis becomes more complex when yk is non-constant over time k and as yk to should be regarded as uncertain. The use of for example Microsoft Excel becomes rather useful to gain further insight (although the math is pretty fun too).


[1] This is likely due to the widespread use of MS Excel and financial pocket calculators allowing for easy NPV calculations, without the necessity for the user to understand the underlying mathematics, treating the formula as “black” box calculation. Note a common mistake using MS Excel NPV function is to include initial investment (t=0) in the formula – this is wrong the NPV formula starts with t=1. Thus, initial investment would be discounted which would lead to an overestimation of value.

[2] http://www.palisade-europe.com/. For purchases contact Palisade Sales & Training, The Blue House 30, Calvin Street, London E1 6NW, United Kingdom, Tel. +442074269955, Fax +442073751229.

[3] Sugiyama, S.O., “Risk Assessment Training using The Decision Tools Suite 4.5 – A step-by-step Approach” and “Introduction to Advanced Applications for Decision Tools Suite – Training Notebook – A step-by-step Approach”, Palisade Corporation. The Training Course as well as the training material itself can be highly recommended.

[4] Most people in general not schooled in probability theory, statistics and mathematical analysis. Great care should be taken to present matters in an intuitive rather than mathematical fashion.

[5] Hill, A., “Corporate Finance”, Financial Times Pitman Publishing, London, 1998.

[6]This result comes straight from geometric series calculus. Remember a geometric series is defined as, where is constant. For the NPV geometric series it can easily be shown that, r being the interest rate. A very important property is that the series converge if, which is the case for the NPV formula when the interest rate r>1. The convergent series sums to a finite value of for k starting at 1 and summed up to ¥ (infinite).

[7] Benninga, S., “Financial Modeling”, The MIT Press, Cambridge Massachusetts (2000), pp.27 – 52. Chapter 2 describes procedures for calculating cost of capital. This book is the true practitioners guide to financial modeling in MS Excel.

[8] Vose, D., “Risk Analysis A Quantitative Guide”, (2nd edition), Wiley, New York, 2000. A very competent book on risk modeling with a lot of examples and insight into competent/correct use of probability distribution functions.

[9] The number of scenario combinations are calculated as follows: an uncertain input variable can be characterized by the following possibility setwith length s, in case of k uncertain input variables the number of combinations can be calculated as , where is the COMBIN function of Microsoft Excel.

[10] A Monte Carlo simulation refers to the traditional method of sampling random (stochastic) variables in modeling. Samples are chosen completely randomly across the range of the distribution. For highly skewed or long-tailed distributions a large numbers of samples are needed for convergence. The @Risk product from Palisade Corporation (see http://www.palisade.com) supplies the perfect tool-box (Excel add-in) for converting a deterministic business model (or any other model) into a probabilistic one.

[11] Luehrman, T.A., “Investment Opportunities as Real Options: Getting Started with the Numbers”, Harvard Business Review, (July – August 1998), p.p. 3-15.

[12] Luehrman, T.A., “Strategy as a Portfolio of Real Options”, Harvard Business Review, (September-October 1998), p.p. 89-99.

[13] Providing that the business assumptions where not inflated to make the case positive in the first place.

[14] Hull, J.C., “Options, Futures, and Other Derivatives”, 5th Edition, Prentice Hall, New Jersey, 2003. This is a wonderful book, which provides the basic and advanced material for understanding options.

[15] A derivative is a financial instrument whose price depends on, or is derived from, the price of another asset.

[16] Boer, F.P., “The Valuation of Technology Business and Financial Issues in R&D”, Wiley, New York, 1999.

[17]  Amram, M., and Kulatilaka, N., “Real Options Managing Strategic Investment in an Uncertain World”, Harvard Business School Press, Boston, 1999. Non-mathematical, provides a lot of good insight into real options and qualitative analysis.

[18] Copeland, T., and V. Antikarov, “Real Options: A Practitioners Guide”, Texere, New York, 2001. While the book provides a lot of insight into the area of practical implementation of Real Options, great care should be taken with the examples in this book. Most of the examples are full of numerical mistakes. Working out the examples and correcting the mistakes provides a great mean of obtaining practical experience.

[19] Munn, J.C., “Real Options Analysis”, Wiley, New York, 2002.

[20] Amram. M., “Value Sweep Mapping Corporate Growth Opportunities”, Harvard Business School Press, Boston, 2002.

[21] Boer, F.P., “The Real Options Solution Finding Total Value in a High-Risk World”, Wiley, New York, 2002.

[22]] Black, F., and Scholes, M., “The Pricing of Options and Corporate Liabilities”, Journal of Political Economy, 81 (May/June 1973), pp. 637-659.

[23] Merton, J.C., “Theory of Rational Option Pricing”, Bell Journal of Economics and Management Science, 4 (Spring 1973), 141-183.

[24] Cox, J.C., Ross, S.A., and Rubinstein, M., “Option Pricing: A Simplified Approach”, Journal of Financial Economics, 7 (October 1979) pp. 229-63.

Mobile Data Consumption, the Average Truth? the Average Lie?

“Figures often beguile me” leading to the statement that “There are three kinds of lies: lies, damned lies, and statistics.” (Mark Twain, 1906).

We are so used to averages … Read any blog or newspaper article trying to capture a complex issue and its more than likely that you are being told a story of averages … Adding to Mark Twain’s quote on Lies, in our data intense world ” The Average is often enough the road to an un-intentional Lie” .. or just about “The Average Lie” .

Imagine this! Having (at the same time) your feet in the oven at 80C and you head in the freezer at -6C … You would be perfectly OK! On average! as your average temperature would equal 80C + (-6C) divided by 2 which is 37C, i.e., the normal and recommended body temperature for an adult human being. However both your feet and your head is likely to suffer from such an experiment (and therefore really should not be tried out … or left to Finns used to Sauna and Icy water … though even the Finns seldom enjoyed this simultaneously).

Try this! Add together the age of the members your household and divide by the number of members. This would give you the average age of your household … does the average age you calculated have any meaning? … if you have young children or grandparents living with you, I think that there is a fairly high chance that the answers to that question is NO! …  The average age of my family”s household is 28 years. However, this number is a meaningless average representation of my household. It is 20 times higher than my sons age and about 40% lower than my own age.

Most numbers, most conclusions, most stories, most (average) analysis are based on an average representation of one or another Reality …. and as such can easily lead to Reality Distortion.

When we are presented with averages (or mean values as it is also called in statistics), we tend to substitute Average with Normal and believe that the story represents most of us (i.e., statistically this means about 68% of us all). More often than not we sit back with the funny feeling that if what we just read is “normal” then maybe we are not.

On mobile data consumption (I ll come back to Smartphone data consumption a bit later) … There is one (non-average) truth about mobile data consumption that has widely (and correctly) been communicated …

Very few mobile customers (10%) consumes the very most of the mobile data traffic (90%).

(see for example: http://www.nytimes.com/2012/01/06/technology/top-1-of-mobile-users-use-half-of-worlds-wireless-bandwidth.html/).

Lets just assume that a mobile operator make claim to an average 200MB monthly consumption (source: http://gigaom.com/broadband/despite-critics-cisco-stands-by-its-data-deluge/). Lets assume that 10% of customer base generating 90% of the traffic. It follows that the high usage segment has an average  volumetric usage of 1,800MB and the low usage segment an average volumetric usage of only 22MB.  In other words 10% of the customer base have 80+ times higher consumption than the remaining 90%. The initial average consumption (taken across the whole customer base) of 200MB communicated is actually 9 times higher than the average consumption of 90% of the customer base. It follows (with some use case exceptions) that the 10% high usage segment spends a lot more Network Resources and Time. The time the high usage segment spend actively with their device are likely to be a lot higher than the 90% low usage segment.

The 200MB is hardly normal! It is one of many averages that can be calculated. Obviously 200MB is a lot more “sexy” than to state that 90% of the customer base consumes typically 22MB.

Created using PiktoChart http://app.piktochart.com.

Do Care about Measurement and Data Processing!

What further complicates consumptive values being quoted is how the underlying data have been measured, processed and calculated!

  1. Is the averaging done over the whole customer base?,
  2. Is the averaging done over active customers?, or
  3. A subset of active customers (i.e., 2G vs 3G, 3G vs HSPA+ vs LTE vs WiFi, smartphone  vs basic phone, iPad vs iPhone vs Laptop, prepaid vs postpaid, etc..) or
  4. A smaller subset based on particular sample criteria (i.e., iOS, Android, iPad, iPhone, Galaxy, price plan, etc..) or availability (mobile Apps installed, customer approval, etc..).  or …

Without knowing the basis of a given average number any bright analysis or cool conclusion might be little more than Conjecture or Clever Spin.

On Smartphone Usage

One the most recent publicized studies on Smartphone usage comes from O2/Telefonica UK (Source: http://mediacentre.o2.co.uk/Press-Releases/Making-calls-has-become-fifth-most-frequent-use-for-a-Smartphone-for-newly-networked-generation-of-users-390.aspx). The O2 data provides an overview of average daily Smartphone usage across 10 use case categories.

The O2’s Smartphone statistics have been broken down in detail by one of our industry”s brightest Tomi Ahonen (A Must Read http://www.communities-dominate.blogs.com/ though it is drowning in his Nokia/Mr. Elop “Howler Letters”). Tomi points out the Smartphone’s disruptive replacement potential of many legacy consumer products (e.g., think: watch, alarm clock, camera,  etc..).

The O2 Smartphone data is intuitive and exactly what one would expect! Boring really! Possible with the exception of Tomi’s story telling (see above reference)! The data was so boring that The Telegraph (source: http://www.telegraph.co.uk/technology/mobile-phones/9365085/Smartphones-hardly-used-for-calls.html) had to conclude that “Smartphones Hardly Used for Calls”. Relative to other uses of course not really an untruth.

Though The Telegraph did miss 9or did not care) the fact that both Calls and SMS appeared to be what one would expect (and why would a Smartphone generate more Voice and SMS than Normal? … hmmmm). Obviously, the Smartphone is used for a lot of other stuff than calling and SMSing! The data tells us that an average Smartphone user (whatever that means) spend ca. 42 minutes on web browsing and social networking while “only” 22 minutes on Calls and SMS (i.e., actually 9 minutes of SMS sounds more like a teenager than a high-end smartphone user … but never mind that!). There are lots of other stuff going on with that Smartphone. In fact out of the total daily usage of 128 minutes only 17% of the time (i.e., 22 minutes) is used for Plain Old Mobile Telephony Services (The POMTS). We do however find that both voice minutes and legacy messaging consumption are declining faster in the Smartphone segment than for Basic Phones (which are declining rapidly as well) as OTT Mobile Apps alternatives substitute POMTS (see inserted chart from http://www.slideshare.net/KimKyllesbechLarsen/de-risking-the-broadband-business-model-kkl2411201108x).

I have no doubt that the O2 data represents an averaging across a given Smartphone sample, the question is how does this data help us to understand the Real Smartphone User and his behavior.

So how did O2 measure this data?

(1) To be reliable and reasonable, data collection should be done by an App residing in the O2 customer’s smartphone. An alternative (2) would be deep packet inspection (dpi) but this would only capture network usage which can (and in most cases will be) very different from the time the customer actively uses his Smartphone. (3) Obviously the data could also be collected by old fashion Questionnaires being filled in. This would be notoriously unreliable and I cannot imagine this being the source.

Thus, I am making the reasonable guess that the Smartphone Data Collection is mobile App based.

“Thousand and 1 Questions”: Does the data collected represents a normal O2 Smartphone user? or a particular segment that don’t mind having a Software Sniffer (i.e., The Sniffer) on the used device reporting his behavior? Is “The Sniffer” a standard already installed (and activated?) App on all Smartphone devices?, only on a certain segment? or is it downloadable? (i..e, which would require a certain effort from the customer), is the collection done for both prepaid & contract customers, both old and new smartphones (i.e., usage patterns depends on OS version/type, device capabilities such as air interface speed DL & UL, CPU, memory management, etc..) … is WiFi included or excluded?, what about Apps running in the background (are these included), etc…

I should point out that it is always much easier to poke at somebody else data analysis than it often is to collect, analyse and present such data. Though, depending on the answer to the above “1,000 + 1” questions the O2 data either becomes a fair representation of an O2 Smartphone customer or “just” an interesting data point for one of their segments.

If the average Smartphone cellular (i.e., no WiFi blend) monthly consumption in UK is ca. 450MB (+/-50MB) and if the consumer had on average cellular speed of 0.5Mbps (i.e., likely conservative with exception of streaming services which could be lower), one would expect that Time spend consuming Network Resources would be no more than 120 minutes per month or 5 minutes per day (@ R99 384kbps this would be ca. 6 min per day). If I would chose a more sophisticated QoS distribution, the Network Consumption Time would anyway not change with an order of magnitude or more.

So we have 5 minutes of Mobile Data Network Time Consumption daily versus O2’s Smartphone usage time of 106 minutes (wo Calls & SMS) … A factor 22 in difference!

For every minutes of mobile data network consumption the customer spends 20+ minutes actively with his device (i.e., reading, writing, playing, etc..).

So …. Can we trust the O2 Smartphone data?

Trend wise the data certainly appear reasonable! Whether the data represents a majority of the O2 smartphone users or not … I doubt somewhat. However, without having a more detailed explanation of data collection, sampling, and analysis it’s difficult to conclude how representable the O2 Smartphone data really is for their Smartphone customers.

Alas this is the problem with most of the mobile data user and usage statistics being presented to the public as an average (i.e., have had my share of this challenge as well).

Clearly we spend a lot more time with our device than the device spends actively at the mobile network. This trend has been known for a long time from the fixed internet. O2 points out that the Smartphone, with its mobile applications, has become the digital equivalent to a “Swiss Army Knife” and as a consequence (as Tomi also points out in his Blog) already in the process of replacing a host of legacy consumer devices, such as the watch, alarm clock, camera (both still pictures and video), books, music radios, and of course last but not least substituting The POMTS.

I have made argued and shown examples that Average Numbers we are presented with are notorious by character. What other choices do we have?  Would it be better to report the Median? rather than the Average (or  Mean)? The Median divides a given consumptive distribution in half (i.e., 50% of customers have a consumption below the Median and 50% above). Alternative we could report the Mode which would give us the most frequent consumption across our consumer distribution.

Of course if consumer usage was distributed normally (i.e., symmetric bell shaped) Mean, Median and Mode would be one and the same (and we would all be happy and bored). Not so much luck!

Most consumptive behaviors tends to be much more skewed and asymmetric (i.e., “the few takes the most”) than the normal distribution (that most of us instinctively uses when we are presented with figures). Most people are not likely to spend much thought on how a given number is calculated. However, it might be constructive to provide a %tage of the customers for which their usage is below the reported average. The reader should however note that in case the percentage figure is different from 50%, the consumptive distribution is skewed and

onset of Reality Distortion has occurred.