knowledge centre

Cisco slashes its mobile data forecasts, but the numbers still look far too high

Rupert Wood Research Director

As well as revising its forecast cellular growth rates downwards, Cisco has also indicated that it expects Wi-Fi offload to be much higher than it previously thought.

Cisco Systems published its latest mobile data traffic forecasts on 6 February 2013.1 These show a substantial downward revision of traffic volumes for 2012 for Western Europe, and lesser, but significant downward revisions for most other regions. Moreover, projected growth rates to 2017 have been revised downwards. Despite these downward revisions, the volumes for 2012 in Western Europe and North America still look much too high to us, and imply growth rates of around 100% for 2012 over what we already know about 2011. This contradicts all the evidence we have seen for actual rates of growth last year in these regions.

Analysys Mason uses the data published by most national regulatory authorities in Europe, and by the CTIA for the USA as primary source data for its annual wireless traffic forecasts (see Wireless network traffic worldwide: forecasts and analysis 2012–2017). These we take in good faith. Cisco's preferred metric is petabytes (PB) per month for December. Our forecasts show total annual data and year average monthly data. This means we cannot compare our forecasts exactly with those of Cisco, but we can estimate year-end figures using a mid-way point between our year averages.

North America

The CTIA recorded 867PB for the whole of 2011 in the USA, and has since indicated that mobile data traffic increased by 21% during the six-month period between the second half of 2011 and the first half of 2012. On this basis we estimate the December 2011 figure to be 102PB for North America as a whole (rounding up for Canada). This is a little lower than Cisco's previous estimate for December 2011 of 119PB per month. However, Cisco now estimates that the figure for year-end 2012 was 222PB, a growth rate of 87% over its own earlier estimate for year-end 2011, and a huge 118% over our own estimate based on CTIA figures. Cisco also retrospectively revised the 2011 figure (rather counter-intuitively) upwards by 14%. Whichever way one looks at it, 222PB implies an implausible acceleration of growth in the second half of 2012. Our estimate is that mobile data traffic increased by 53% from 102PB to 156PB between end-2011 and end-2012. Given the CTIA's 21% growth figure for the first half of 2012 (which we did not have when we published our own forecasts), this 53% may be on the high side, although it may be true that traffic tends to increase more strongly in the second half of any given year.

Figure 1: Cisco Visual Networking Index 2011–2016 (February 2012), Cisco Visual Networking Index 2012–2017 (February 2013) and Analysys Mason's Wireless Traffic Forecasts (September 2012), year-end cellular mobile data, North America [Source: Cisco Systems and Analysys Mason, 2012 and 2013]


Western Europe

The discrepancies between published data and Cisco estimates are even starker for Western Europe. We can say within a 5% margin of error that total mobile data traffic in Western Europe in 2011 was 980PB. Annualised growth rates for the first half of 2012 are varied. At the top end, Denmark has a 75% annualised rate; at the bottom end, Portugal has –7%. However, the average leads us now to believe that growth in Western Europe in 2012 was about 35–40%. Our published forecasts had 35% growth for the period, so part-year indications confirm our hypothesis. This leads us to calculate year-end traffic for 2011 at 96PB, which we believe increased to 128PB at year-end 2012.

In last year's forecasts Cisco estimated traffic in Western Europe at 366PB at December 2012. It has cut this by more than 50% in this year's forecast to 181PB, but it also indicted in the report that traffic in Europe increased by 44% in 2012, meaning that it has revised its earlier 2011 figure down by 30% from 180PB to 126PB. The growth rate looks about right to us, but the actual volumes – even after the dramatic 50% cut – still look far too high. For 180PB to be true, growth would actually have to be 89% in 2012, a figure higher than any rate of which we are aware in individual Western European countries.

Figure 2: Cisco Visual Networking Index 2011–2016 (February 2012), Cisco Visual Networking Index 2012–2017 (February 2013) and Analysys Mason's Wireless Traffic Forecasts (September 2012), year-end cellular mobile data, Western Europe [Source: Cisco Systems and Analysys Mason, 2012 and 2013]


Cisco has also cut its forecast growth rates

As well as revising its forecast cellular growth rates downwards, Cisco has also indicated that it expects Wi-Fi offload to be much higher than it previously thought. However, 2012–2017 CAGRs of 56% for North America and 50% for Western Europe still look much too high to us. The increase in traffic brought about by subscriber growth tends to be offset by a dilution of average usage by later adopters. This makes growth charts have a broadly linear shape rather than an exponential one. We would expect that annual growth rates, averaged out over whole regions, would decline year-on-year. By our latest estimates, annual growth in 2012 (40–50% in North America and 35–40% in Western Europe) is already lower than these CAGRs. This, combined with a difference of opinion about the actual volumes of data in 2011 and 2012, mean that our forecasts for these two regions and those of Cisco continue to diverge dramatically. Analysys Mason forecasts 3.6-fold growth between 2012 and 2017 in Western Europe (as opposed to Cisco's 7.6-fold) and 4.5-fold growth in North America (as opposed to Cisco's 9.4-fold). The result is that Cisco's forecast of the volume of traffic in 2017 in each of these regions is more than three times that of Analysys Mason.

Forecasting data traffic is riddled with uncertainties, and, when the explicit or implicit warnings in forecasts are taken seriously, involves feedback loops that will contradict the trend forecasted. We certainly get things wrong: for example, we have probably underestimated future traffic generated by LTE FMS-type services, and we underestimated 4G take-up in parts of Asia–Pacific. There is nothing wrong with retrospective revisions: new evidence comes to light all the time. However, trying to get it right by assessing the available evidence is important. The future growth of mobile data traffic should influence the rate of build-out of LTE and LTE-A networks, and it will determine the timing of any spectrum crunch. In October 2012, when challenged about the CTIA's interpretation of its own data, CTIA director Robert Roche said that "dismissing 'a so-called spectrum crunch' ignores not just a consensus in the US, it neglects the global nature of the analysis that calls for more commercial spectrum allocation around the world, to accommodate growing numbers of users and increasingly complex uses." We would not want to dismiss out of hand the idea of a spectrum crunch anywhere. However, we would challenge any consensus view (if this consensus still exists) that predicts the timing of a crunch on questionable analysis of the pretty robust data that is already publicly available.

1  Cisco (San Jose, California, 2013), Cisco Visual Networking Index: Global Mobile Data Traffic Forecast Update, 2012–2017. Available at