After a long day of hard work, you just put the kids to bed and settled comfortably on your couch looking forward to watching the 3rd episode of Breaking Bad. Not a trivial milestone, because you will binge-watch the whole season after watching the 3rd episode : the tipping point according to Netflix’ statistics. You pour yourself a drink while waiting for Netflix to load, only to notice a static “Loading 25%” appearing on screen. Network error, slow internet connection. “Not now!”, you mumble in silence, while hitting the reload button.
According to Cisco, the amount of Video-on-Demand will be the equivalent of 10 billion DVDs per month by 2022. An estimated 80% of all internet traffic will be from video. To enable all this explosion of data traffic, the average speed of broadband connections will roughly double between 2017 and 2022. And we will consume 10 times more data through our mobile devices. This demands huge investments in telecom networks. The telecom sector is forecast to be the third largest sector when it comes to infrastructure investments, only after electricity and roads, and on par with water supply1. At the same time, telecom providers are facing flat revenues in most mature countries due to saturated markets and heavy competition. So how can telecom operators upgrade their networks by rolling out 5G and Fiber-to-the-Home, while also earning a decent return on their investments? Upgrades cannot all be done at once, so where should they invest first? When will you benefit from Fiber roll out, and get a seamless Netflix experience? Traditional approaches to telecom network investments leave money on the table, because they tend to focus primarily on the cost side. The most cost-efficient areas go first. However, costs are not the only driver of investment returns. Equally important is the customer lifetime value impact: how do network upgrades influence customer loyalty (and hence market share) and contract value? These factors have shown to differ substantially, and depend on factors such as demographics and current network speeds. For instance, areas with current DSL bandwidths 7-12 mb/s have shown a 4 times higher market share growth during the first 18 months after fiber roll out compared to areas with 12-50 mb/s as current network speed2. As it turns out, providers find it hard to predict these customer effects and thus make very rough assumptions when making investment decisions. Telecom Operators have an opportunity to use AI to predict customer behavior related to network upgrades. This allows to account for all customer-level effects on very detailed customer-by-customer level, and use that to increase investment returns. An example case from a European operator shows that prioritizing network investments based on this AI-powered solution increases returns by over 120% compared to rudimentary methods.
1 According to MGI (“Bridging Global Infrastructure Gaps”, 2016) 0.6% of GDP is spent on Telecom on average globally.
2 Real example from a European country.
Exponential growth in data and connectivity continues need for infrastructure investments
Telecommunications is one of the most capital-driven industries. European operators invested €27bn in capex in 2017 – the equivalent of €83 per capita. These investment levels have steadily increased by on average 3% per year between 2012 and 2017, largely driven by infrastructure investments for both fixed and mobile (both roughly in equal measures). Current growth rates in connectivity and data traffic are by no means putting a break on investment levels. Overall global data traffic is forecast to grow by 26% per year on average between 2017 and 20221. Not surprisingly, most of this growth is caused by mobile traffic with a 46% compound annual growth rate, versus ‘only’ 15% per year for fixed/wired traffic. Internet video is the prime driver of growth, with a share in overall traffic going up from 55% in 2017 to an expected 71% in 2022. To enable all this traffic growth, connection speeds for both broadband and mobile will have to improve dramatically. Global average fixed broadband speeds will roughly double from 39 Mbps to 75 Mps between 2017 and 2022. Similarly, average mobile network connection speeds will triple from 9 to 29 Mbps over the same period.
Infrastructure upgrades are not only growing in size, but also in complexity. Mobile networks are being expanded with 5G, while also building more WiFi hotspots for mobile offloading. Traffic from mobile devices on Fixed/WiFi connections is forecast to grow by a whopping 53% per year. Fixed networks are being upgraded with Fiber-to-the-Home, while also investing in new copper/DSL technologies. Finding the optimal balance of fixed, mobile and Wi-Fi assets among European operators to meet the growing expectations of good coverage in every type of indoor and outdoor physical location is challenging.
1 IP traffic – Cisco Visual Networking Index: Forecast and Trends, 2017–2022
Source: Cisco
Operators face increasing need for capex optimization due to maturing markets
These growth trends create unprecedented investment challenges for telecom operators. While market demand for communication services grows, telecom revenues and average price levels have declined. Mass market (B2C) telecom revenue in Europe (ETNO perimeter1) is down from €182bn in 2012, to €165bn in 2018 – an average drop of 2% per year. Heavy competition in certain areas causes price pressure. Overall European revenue per user has declined for mobile, and is stable for fixed broadband.
To fund the upgrades in infrastructure, European operators spend around 17% of sales per year on capex. The combination of growing capex and sluggish revenues have led to an increase of this ratio by about 6% per year over the past 5 years2. This shows that network capex does not easily translate into revenue growth, at least not in the short term, and in particular for larger operators3. Evidently, optimizing network investments to enhance returns over the long term is a high priority for operators.
1 European Telecommunications Network Operators
2 ETNO Annual Economic Report 2017
3 Horvath Partners 2017, PwC
The return on a network investment for a particular area depends on many different factors, and requires a data-driven multivariable prediction. Traditional
approaches fail at this.
Optimizing network investments is a complex challenge and traditionally suboptimal
While the need for infrastructure investments is obvious, it is less clear why operators struggle to enhance returns from their investments. To answer this question, let us first examine why this is a challenge.
In recent years, around 50% of European network capex has been spent on fixed network upgrades – both for fiber roll out and copper network improvements[1]. These investments are made at local area levels. For DSL lines this implies upgrading hardware in street cabinets. For fiber roll-out, this implies connecting households to a fiber network and thus civil works. This requires scarce resources and diligent planning, which makes fixed network upgrades not only capital intensive, but also extremely time consuming. Reaching meaningful coverage can take years to accomplish in a typical European country.
Given these capex and capacity constraints, the challenge for telecom operators is to optimize the sequence of investment areas. Ideally, this is done based on expected returns. The crux of the matter is that returns can vary considerably from one area to the next, and between technologies. Hence, the ability to prioritize investment options based on expected returns is a key driver of the overall capital efficiency of operators2. The return on a network investment for a particular area depends on many different factors, and requires a data-driven multivariable prediction. Traditional approaches fail at this.
Investment costs typically depend on physical circumstances and population density, which are relatively straightforward to estimate. Future revenue development on the other hand is a different story, and depends on many different factors. First, market share evolution depends on the particular demographics in the area and the quality of the current network compared to the competition. Second, average customer revenue will differ for instance due to household composition and income levels, or, for B2B-customers, industry sector. Third, market needs will evolve over time as the need for bandwidth increases. As such, developing a business case for each individual area requires a multifactorial prediction over a 10+ year period.
To illustrate how ROCE pans out in practice, consider 2 different investment areas. Both areas are in the same city, and both have around 3000 households. Market shares are very comparable: 29% versus 28%. Area 1 however consists of terraced houses, while area 2 has more condo’s. Social class also differs, with higher income levels in area 1 than in area 2. Also, average household size is 2.5 in area 1 versus 1.9 in area 2. If we had to choose, which area to prioritize for FttH investment?
Before Smart ROCE, the marketing department would argue that Area 1 should be chosen because the wealthier demographics would lead to higher market share and higher ARPA. Operations would say that the more densely populated Area 2 should be prioritized, because average investment costs per household are much lower. The result: endless discussions, without one overarching criterion for the trade-off.
Now enter Smart ROCE. The algorithms indeed predict slightly higher commercial benefits over the coming 10 years for Area 1 if we upgrade the network. However, the total investment costs for area 2 are 15% lower. After all the calculations are run, Area 1 shows a negative ROCE of -0.14. Area 2 on the other hand has a positive business case with a ROCE of 0.19 – that is a 19% return on investment. In this case, the cost side outweighs the revenue side.
1 ETNO Annual Economic Report 2019
2 For instance measured in terms of returns on (newly) invested capital for the operator as a whole
Traditionally, operators make investment decisions for specific regions based on simple heuristics, such as looking plainly at average capex per household, or population density. Customer-level considerations, such as the expected lifetime value at household level resulting from the upgrade, is rarely accounted for. Moreover, the lack of one holistic metric that incorporates all underlying drivers – for both cost and revenue – creates inefficient back-and-forth discussions between departments. The network department typically emphasizes costs, while marketing looks at the potential benefits from happy customers. The result is often an arduous decision process, with suboptimal outcomes.
Smart ROCE: leverage AI to make optimal investment trade offs based on long term value creation
To help telco operators optimize their network investment decisions, we developed an AI powered solution – called Smart ROCE – that calculates one single metric for each individual investment option. This metric is called ROCE (Return on Capital Employed) and is determined with help of multiple machine learning algorithms that analyze the entire customer base, enriched with external data. The first two algorithms predict for each location in an investment area the churn and acquisition probabilities over a 10 year period, taking evolving market needs for bandwidth into account. Another algorithm predicts the average revenue per location in the area by determining the likelihood of cross and upselling based on demographics such as household composition, income levels, age, etc. The combination of these algorithms provides a prediction of market share and revenue evolution for a particular area, depending on the infrastructure. The outcomes of these predictions are used to calculate the change in Customer Lifetime Value (CLV) for each individual household after a network upgrade. The ROCE can subsequently be defined as this value creation compared to the investment. A ROCE larger than 0 indicates a positive business case.
Devising an optimal investment strategy
The Smart ROCE engine calculates ROCE for all possible investment options per geographic area, for different potential carriers (Fiber versus VDSL upgrade). On the one hand, this allows for choosing between different carriers for a specific area. Thus, using the output of the Smart ROCE engine, one can identify multiple investment scenarios.
For instance, figure 2 shows a prediction for three different investment scenarios for a particular area. The first scenario is to invest in FttH, showing a steady increase in expected market share. In scenario 3, FttH investment is postponed by 5 years.
ROCE as a new holistic metric leads to counter intuitive insights
To illustrate how ROCE pans out in practice, consider 2 different investment areas. Both areas are in the same city, and both have around 3000 households. Market shares are very comparable: 29% versus 28%. Area 1 however consists of terraced houses, while area 2 has more condo’s. Social class also differs, with higher income levels in area 1 than in area 2. Also, average household size is 2.5 in area 1 versus 1.9 in area 2. If we had to choose, which area to prioritize for FttH investment? Before Smart ROCE, the marketing department would argue that Area 1 should be chosen because the wealthier demographics would lead to higher market share and higher ARPA. Operations would say that the more densely populated Area 2 should be prioritized, because average investment costs per household are much lower. The result: endless discussions, without one overarching criterion for the trade-off. Now enter Smart ROCE. The algorithms indeed predict slightly higher commercial benefits over the coming 10 years for Area 1 if we upgrade the network. However, the total investment costs for area 2 are 15% lower. After all the calculations are run, Area 1 shows a negative ROCE of -0.14. Area 2 on the other hand has a positive business case with a ROCE of 0.19 – that is a 19% return on investment. In this case, the cost side outweighs the revenue side.
This leads to a drop in market share in the first 5 years, which has to be recouped after the investment. In scenario 2, this drop in market share is prevented by applying VDSL upgrade as bridging technology.
This optimization allows operators to maintain market share in areas where FttH returns do not lead to immediate prioritization for upgrade, but VDSL investments keep customers on board with a relatively cost efficient upgrade (with positive returns already over a 5-6 year period).
Figure 2
This example illustrates how an optimal investment strategy can be devised. If the objective is to maximize value creation, i.e. return on investment, a simple strategy is visualized in figure 3.
In this example, areas with the highest ROCE for Fiber are prioritized. Furthermore, if copper (VDSL) investments can be recouped over a 6 year period (yielding ROCE >0 in that period), then investing in VDSL is a no-regret move to safeguard market share while averting scarce fiber roll-out capacity to other areas. This hybrid approach yields 19% higher NPV, than working with fiber alone. So not only do we prioritize areas with the highest expected returns, we also upgrade more areas by optimally trading off different carrier options.
Next to trading off different technology options, the output of the Smart ROCE engine allows for optimal sequencing investment areas over time. By prioritizing areas with the highest ROCE to go first, overall value creation is maximized. See figure 4 for a real world example of how expected returns can differ for quarterly capex batches.
Figure 4Prioritizing investments with Smart ROCE can double NPV
Applying the smart ROCE approach consistently over the entire network upgrade horizon allows to maximize returns over time. Running a simulation of different investment strategies demonstrates the value of Smart ROCE. Figure 5 shows a 126% increase in 10-year NPV if fiber investments are consistently prioritized based on ROCE versus based on market share or churn rates. Also, the overall benefit of a hybrid approach is shown.
Implementing Smart ROCE: AI application for prioritization and monetization
To reap the full benefits, it is imperative that the outcomes of the smart ROCE algorithms are applied in the investment decision process in a sustainable and consistent manner. To achieve this, an application has been developed which guides the entire network investment process end-to-end.
This process encompasses two phases. The first phase consists of investment prioritization, in which for each investment period the relevant areas are prioritized for upgrades. This is done using an interactive interface in which, per area, the optimization results of the Smart ROCE engine are presented. The user can then analyze and possibly overrule optimization results (for instance, for operational arguments), but also plot them in time. Moreover, the application can be used to perform top-down analyses on the optimization results; what would these decisions mean for the types of customers that benefit from network investments, how would the distribution of network speed throughout a macro-area evolve over the coming years, etc.
The second phase starts once these upgrades are implemented. During this phase, the Smart ROCE application is used to perform Benefits Management: track market share development, and compare actual uplifts with predictions. For instance, when market share lags predictions, additional marketing efforts might be needed to reach full potential.
In conclusion: Setting the standard in network investments optimization with Smart ROCE
Making network investment decisions is traditionally a tough challenge for telecom operators. Prioritization decisions are often based on (a mix of) heuristic measures, which do not optimally account for the potential customer lifetime value creation related to a particular network upgrade. Smart ROCE offers an AI solution which incorporates all relevant value creation effects, and determines one single metric that quantifies the expected return of an investment. This allows for very detailed investment optimization, making optimal trade offs between technologies and areas. Based on this approach, operators can set a new standard for internal decision processes for network investments, for both fixed and mobile. The financial benefits of this approach can run into hundreds of millions of euros, depending on budget size and current practices. At a minimum, Smart ROCE creates a smooth decision process, by eliminating all the guess from return calculations.