Multiple connections per user: The impact on penetration and ARPU - Multiple connections per user distorts real penetration and ARPU

GSMA Intelligence recently commissioned primary research in key markets in Western Europe and North America to understand the impact of consumers’ usage of multiple cellular connections. The study aimed to address to what extent this phenomenon affected – and distorted – established key performance indicators in the cellular industry such as market penetration and average revenue per user (ARPU).

The study measured the number of multiple connections or multiple SIMs per user – either active SIM cards or unique mobile phones numbers where SIM cards are not used – in order to calculate values of real penetration per user and average revenue per user, unlike the per-connection values reported by operators today.

In both cases, these two metrics take into account multiple user connections rather than using the traditional method of simply dividing revenue (ARPU) or population (penetration) by the number of connections.

In North America (USA/Canada), we found that consumers on average held 1.3 cellular connections each in Q3 2009. This means that real market penetration in the region stands at 71% compared to the reported figure of 92%. We can also conclude that whilst reported revenue per connection appears to be falling year-on-year, revenue per user is in fact increasing, which means that North American consumers are still spending more on mobile but spreading this spending across multiple connections. Our study reveals that real revenue per user in North America has increased from USD $60 to almost USD $64 since 2006.

Real market penetration per user in Western Europe has reached 87% compared to 130% reported penetration per connection. However, in contrast with North America, our research shows that consumer mobile spending in Western Europe is decreasing, even though revenue per user is showing a slower rate of decline than the reported revenue per connection. Our study found that reported revenue per connection in Western Europe will fall to EUR €23 this year compared to EUR €24.7 in 2008, while revenue per user will fall to EUR €33 this year from EUR €34.2 in 2008. Western European consumers on average held 1.5 cellular connections each in Q3 2009.

In summary, our study reveals that North America has greater potential for per-user connections growth and individual users in the region are spending more on mobile services, countering fears that the economic situation has led to a slowdown in spending in the sector. However, the situation in Western Europe is less encouraging and reflects a real need for cellular operators in the region to ramp up investment in value-added services and network quality in order to take advantage of new opportunities through data services.

The phenomenon of multiple connections per user

Cellular operators have traditionally reported their customer base in terms of ‘subscribers.’ Historically, this made sense when the industry was in its infancy and an individual user would rarely have more than one connection, be it a unique number associated with a separate SIM card (GSM) or a unique device and/or phone number (CDMA).

Today, however, this method of reporting customers can distort analysis of market penetration or ARPU because individual ‘subscribers’ are increasingly likely to have more than one cellular connection. For instance, during the second half of 2009, cellular penetration is reported to have reached 151% in Italy and 193% in Greece, while ARPU growth reportedly fell by 5% on a compound annual basis in Western Europe between 2006 and 2009. However, these statistics demand a greater level of scrutiny by looking at real revenue and real penetration per user – which factor in multiple connections per user – to fully understand current market trends.

The usual methods of reporting market penetration and ARPU are derived from total cellular connections reported by cellular operators, which in many cases additionally include redundant connections (e.g, inactive SIM cards) as well as a multitude of non- voice devices available today, e.g. modems and e-book readers. This implies high levels of multiple connections and, therefore, significantly lower levels of market penetration.

Since 2006 we have studied multiple connections and this report builds on our previous analysis, aiming to provide an update on the phenomenon as well as push forward further research in this area.

To illustrate the potential impact of multiple connections, on traditional cellular industry metrics, we offer the following scenario, which highlights the importance of tracking real revenue per user:

• A fictitious cellular operator has one user;
• The user has a primary connection (SIM or device) and spends USD $60 per month – this gives the operator an average revenue per user of USD $60;
• The user acquires a secondary connection and spends an extra USD $20 on this connection;
• According to standard reporting, the operator’s revenue per connection has now reduced 33%, dropping from USD $60 to USD $40 (i.e. USD $80 per month across two connections);
• Calculating true revenue per user however, the figure has actually increased 33% from USD $60 to USD $80.

Market penetration

The most fundamental of metrics affected by this phenomenon is market penetration. In markets with very low penetration or those where cellular deployment is in its infancy, it is statistically valid to consider that each user has a single connection. As market penetration rises, there is an increase in multiple connections per user, driven by second devices and data-only connections (USB modems, dongles, data cards, etc.) at which point it becomes more difficult to ascertain ‘per user’ metrics.

Given that penetration is a population-based metric (i.e. based on individual users), it is clear that reporting market penetration using the number of connections rather than users distorts the true value for any market where the number of connections per user exceeds 1. This is most evident when looking at markets where penetration by connections has exceeded 100% (a full table for the markets covered in this research follows this analysis). With the reported data we have for the number of connections per user, we are able to extrapolate the connections per user factor for other markets which share similar demographics and penetration levels. To do this, we fit an asymptotic power curve to the plotted data for the (reported) number of connections per user against the market penetration by connections. For any market with a given penetration by connections, we can deduce the weighting factor for the number of connections per user. The curve is asymptotic along the line for 1 connection per user since for any number of users there must always be a minimum of 1 connection per user.

From the connections penetration and the number of connections per user we can simply divide one by the other to ascertain the penetration by user. Another power curve can be fitted to this ‘per user’ data to model the real relationship between penetration by user and the number of connections per user.

Looking at the reported data, and firstly in North America (US and Canada), we find that the formulae that represents the power curve for the penetration by connections relationship is:

Y = 1 + nX^z (n = 0.4, z = 2.0)

Adjusting these data points and their fitted curve for the number of connections per user as per our methodology, we reach the graph below, showing real penetration by user in the North American region.

Thus, in Q3 2009, while reported penetration by connections reached 92% in North America, real penetration by user stood at a more modest 71%. We also see a ceiling to this real measure of penetration, which rises asymptotically towards 80% with an increasing number of connections per user. The curve shows the true extent of the room available for growth in even mature markets. This is especially evident if we apply the model to our forecast, which in North America in 2013 shows 77% penetration by user for a surplus of over 100 million connections.

Secondly, we can apply the same methodology to our research in Western Europe. Here, for 5 markets, the formulae that represents the power curve for the penetration by connections relationship is:

Y = 1 + nX^z (no = 0.26, z = 2.2)

Again, adjusting for mutiple connections per user, we reach the graph below.

These results are somewhat different – a scenario that is rapidly approaching user saturation. In Q3 2009, reported penetration by connections is just below 130% yet real penetration by user stands at 87% given almost 1.5 connections per user. As per North America, the curve for real penetration by user implies a ceiling of 90-92% which highlights the somewhat higher level of saturation. Our adjusted forecast for 2013 shows a real penetration by user of 88% and a surplus of over 200 million connections. Overall, the trend in the number of connections per user is closely aligned with the clear slowdown in net additions across Western Europe in recent quarters. Most cellular operators have been reporting lower – and some negative – quarterly net additions due to the high levels of market maturity and fierce competition.

At a market level, Germany and the UK are showing particularly strong signs of user saturation given real penetration by user of over 90%. We have modelled Finland, Greece and Portugal which follow the same trend as Sweden and Denmark, all about to cross the 90% line. In these markets, multiple device ownership is almost solely driving the growth, propelling it well above any meaningful direct comparison with the logical maximum number of customers.

In contrast, France has a real penetration by user of just 68%, showing the true room for growth in the country. As a market, France has reached the point in its lifecycle where multiple device ownership is common, but the customer base has not yet been saturated. The entry of a fourth player in the country in the coming quarters is likely to boost competition and increase the level of market maturity.

For analysis and comparison, we provide the full data set below for both regions, showing the reported data alongside our modelled estimates and the real penetration by user these connections per user weightings imply.

Average revenue per user

In addition, we can assess the impact of multiple connections per user on average revenue per user. In the same way that operators use the term ‘subscribers,’ when actually the figure reported is connections, ARPU as reported is actually an average revenue per connection, not per user. In a similar manner to market penetration per user, we can take the figures reported for average revenue per connection and adjust these by our weighting for the number of connections per user to arrive at a more statistically valid metric for average revenue per user.

In markets with high penetration, revenue per user is seen to be falling when in fact, as shown by our hypothetical operator in the introduction, users (or at least some of them) are actually spending more. This also drives the fact that connections don’t have price elasticity; users do.

In North America, the results are particularly revealing: while revenue per connection declines year on year, revenue per user continues to rise. This means that North American consumers are actually still spending more on cellular despite the often quoted trend in falling reported ARPU. On aggregate, growth in revenue per connection fell on average by 0.4% between 2006 and 2009 yet in reality, individuals are choosing to spend more except they do so over multiple connections. In the region, average revenue per user has grown by a compound yearly average of 1.8% during the same period. In Q3 2009, during which our base metric of connections per user has been recorded, average revenue per user was USD $63.5 compared to a reported average revenue per connection of just USD $51.0.

It is worth noting that this trend comes from a market where 82% of connections are contract based – a segment that drives a majority of the revenue generation. In addition, a sizable number of these multiple connections are most likely allocated to business customers on high-end tarrifs using voice and data services.

In Western Europe however, the situation is somewhat different. In this region, growth in revenue per connection has been falling by a compound yearly average of 5.4% over the last 3 years. In addition, real revenue per user has also been falling but by a lower value of 2.0%. We estimate that the average revenue per user in Western Europe in Q3 2009 is EUR €33 compared against a reported average revenue per connection of EUR €23. A year ago these figures were a marginally higher EUR €34.2 and EUR €24.7, respectively.

Historically, there is an implied transition point in 2004-5 where net additions and revenue growth both slowed due to the prevalence of multiple device ownership. Over the last 3 quarters, the region saw its sharpest ever slowdown with quarterly connections growth falling below the 1% mark, making it the most mature region in the World.

The overall findings for Western Europe seem reasonable if you consider that:

• 56% of the installed base in Western Europe is prepaid which generate less value than the contract customers.
• In terms of device usage we believe that approximately 50% of mobile phones sold in Western Europe are located in the sub-EUR €150 price segment (average selling price) and are used mainly for voice usage and messaging.
• According to our research, less than one third of laptop owners currently use a datacard/USB dongle/HSPA-enabled laptop to connect to the Internet through a mobile network.
• Due to the economic environment, consumers are rationalising their spending and are likely to choose cheaper and more affordable offers.
• Finally, the region is reaching individual user saturation, forcing mobile operators to focus all their efforts on value share and customer retention.

Methodology and limitations

Our primary research has been conducted by Consumer Analysis Ltd in the following markets: Germany, France, Spain, Italy, United Kingdom, United States of America and Canada. For each we have selected a survey sample statistically representative of the population. Consumers were questioned through a hybrid survey comprising phone interviews and email Q&A.

This analysis has a number of potential limitations:

• It is based on a snapshot in time. The modelling of historical data for each country is not based on reported time-series data but on informed assumptions.
• It is a cross-sectional analysis – countries may follow different patterns and consumers may behave differently. The aggregated relationships for modelled countries may not hold in some cases.
• Power curve fitting may not always be optimal and coefficients may change over time.
• Operators often use different definitions for cellular metrics and these may change through time depending on reporting policies.