9 Comments

Summary:

To hear Cisco tell it, the world is quickly running out of wireless spectrum. Tim Farrar, of TMF associates, says a look at the numbers shows that Cisco first overstated them, then revised them, and is now overstating them again.

cardshuffle
photo: Alex Garaev/Shutterstock.com

Cisco’s mobile VNI forecast (the shorthand for Visual Networking Index: Global Mobile Data Forecast), issued last week, is widely regarded as the leading source of information on how the mobile data market will evolve over the next five years.  Policymakers including the FCC use it in their decisions about how to allocate wireless spectrum.

However, like every forecast, the VNI has its flaws — namely that it may overestimate the future demand for mobile data on cellular networks, while understating the need for additional unlicensed spectrum allocations.

Earlier predictions didn’t pan out

Last October, I wrote an article for GigaOm pointing out the dramatic slowdown in mobile data traffic seen in the CTIA’s semi-annual wireless industry survey and asking whether the supposed “spectrum crisis” was a myth. Of course, that didn’t go down well with some people, and CTIA executives lined up to proclaim that It is No Trick – There is a Spectrum Crisis, and asserting that “as Cisco’s data shows… there must be more spectrum to meet demands from consumers and businesses across the country.”

However, Cisco has now revealed its latest VNI mobile data forecast that instead of the originally projected 118 percent growth in North American mobile data traffic between December 2011 and December 2012, traffic grew by only 64 percent over that period – which is to say much slower than in 2011 and far below prior expectations.

So is that the end of the spectrum crisis? Not if you take Cisco’s projections of future growth at face value: They expect 10-fold growth in North American mobile data traffic between 2012 and 2017. Indeed, Cisco actually projects that growth in North American mobile data traffic will be even faster in 2013 (70 percent between December 2012 and December 2013) than the 64 percent it estimated for the last 12 months.

Conflicting data sources

There are reasons to be cautious about the weight that should be given to these forecasts. Cisco has retroactively revised its mobile data traffic estimates, reducing the total estimated global traffic in December 2011 by 13 percent (from 597PB/month in last year’s forecast to 520PB/month in the current model). This is largely due to 30 percent and 23 percent reductions in the European and Asia Pacific traffic estimates respectively, partially offset by a 14 percent increase in estimated North American mobile data traffic. The scale of these revisions indicates that there is considerable uncertainty in Cisco’s numbers, and highlights the difficulty of obtaining real traffic data from mobile network operators.

Nevertheless, at least in the U.S. we can attempt to validate Cisco’s numbers, given that CTIA’s mobile data traffic statistics are based on direct reporting by carriers accounting for 97 percent of wireless connections in the U.S.. In its latest forecast, Cisco estimates that mobile data traffic in the U.S. was 128PB/month in December 2011, and increased to 207PB/month by December 2012.

However, CTIA data indicates that 633PB were carried in the first six months of 2012, for an average of 105.5PB each month. Cisco’s estimate for the U.S. is clearly inconsistent with the CTIA statistics: It is hardly likely that monthly traffic declined significantly between December 2011 and June 2012, and equally implausible that total mobile data traffic in the U.S. then doubled in the second half of the year.

Based on the above analysis, it seems advisable to be rather cautious about the use of Cisco’s mobile data traffic statistics to make policy decisions about the U.S. wireless market structure. That has not been the case historically, with the FCC Chairman often citing Cisco’s projections to suggest that the “skeptics” about the so-called “looming spectrum crunch” were simply wrong. We now have a looming battle between advocates of making more unlicensed and shared spectrum available, and those insisting that all available spectrum (such as that freed up in the upcoming broadcast TV incentive auctions) must be auctioned.

But the most critical piece of data that should be used to inform this debate is how much mobile data traffic will be carried on traditional cellular networks, and how much will instead be able to use unlicensed Wi-Fi spectrum in the 2.4GHz, 5GHz and (potentially) the White Space frequency bands.

Offloading a crucial variable

In previous years Cisco’s forecasts substantially understated the impact of Wi-Fi “offloading” on mobile data traffic growth: Just last year, Cisco estimated that the proportion of data offloaded from smartphones and tablets in the U.S. would fall from 49 percent of their data usage in 2011 to 46 percent of their data usage in 2016. Instead, according to Cisco’s latest forecast, offload is already 60 percent of smartphone and tablet traffic.

Cisco remains relatively cautious about future use of Wi-Fi: the proportion of traffic offloaded from smartphones is only expected to grow by 1 percent per year between 2012 and 2017 – despite having expanded from 21 percent at the end of 2010 to 49 percent at the end of 2011 and as much as 59 percent today. If,  instead, as much as 80 percent of traffic were “offloaded” (which is in line with the traffic split for current users of Cisco’s Data Meter application), then the amount of data traffic carried on cellular networks might be nearly halved. That’s a major difference in outlook from what Cisco is predicting.

When policymakers consider an appropriate balance between future allocations for licensed and unlicensed spectrum, let’s hope they take into account the likelihood that Cisco’s estimates of a 10-fold increase in U.S. mobile data traffic over the next five years may not be realized, whether because of an overestimate of recent traffic growth or an underestimate of future Wi-Fi offload. But given the challenges of dispelling the myth of the “spectrum crisis” (and the carrot of those supposed billions of dollars in auction revenues), I’m not holding my breath.

Tim Farrar is president of Telecom, Media and Finance Associates, a consulting and research firm in Menlo Park, Calif., which specializes in technical and financial analysis across the satellite and telecom sectors. Follow him on Twitter @TMFAssociates.

Photo courtesy of Alex Garaev/Shutterstock.com

You’re subscribed! If you like, you can update your settings

  1. Reblogged this on Rational Arrogance and commented:
    Where will the data go? To mobile operators? Or to Wi-fi providers? The answer seems to be bloiwing in the wind – but there’s a storm coming, obviously,,,

  2. To be fair to Cisco, forecasting a quickly changing industry is a very tough job. No forecast is ever perfect.

    Also, the fact that Cisco seems to be underestimating Wi-Fi offload suggests that they are being cautious about making it look biased towards their core business.

    1. I don’t think the people at Cisco who develop the numbers are consciously trying to favor one side over the other. But they have had a tendency to underestimate WiFi across multiple versions of the forecast.

      As a company, Cisco would generally benefit if providers expected excessive growth for either cellular or WiFi data, as it supplies significant amounts of equipment to both camps. However, its worth noting that some parts of Cisco have been accused of waging a “War on TV White Spaces” in the past (http://tales-of-the-sausage-factory.wetmachine.com/my-insanely-long-field-guide-to-ciscos-war-on-the-tv-white-spaces/).

      I think many people have assumed that Cisco has some special insight (through provision of core Internet routing equipment perhaps?) which means its number are based on “measurements” or some such, in contrast to the (often more conservative) “estimates” made by research firms etc.

      The changes (and apparent inaccuracies) in historic numbers highlight that Cisco is relying on significant amounts of guesswork, just like anyone else.

  3. MissEsterMation Sunday, February 10, 2013

    I’d be taking any statements from Cisco on wireless spectrum and cellular networks with a truckload of salt. They bought Navini as an ill-fated attempt to add RAN to their portfolio so it could offer customers a full e2e solution, only to kill the business less than 2 years after they acquired it because no one was buying WiMax, as the big RAN vendors in Ericsson, NSN and ALU backed the winning LTE horse instead.

    65-75% of a mobile Service Provider CAPEX spend is on the RAN, where Cisco does not play, at least not on the macro level. It is not a credible player in the wireless space.

  4. The bottom line is that these companies want the public to believe a resource is scarce when it is not. There is PLENTY of spectrum, it’s just being used in the most inefficient and *profitable* way possible.

    Now we have the added benefit of companies buying spectrum simply so no one else can use it, and then it goes unused. Then the ISPs come back and say there’s no spectrum. Same old song and dance.

  5. Dear Mr. Farrar,

    I appreciate your time to present a well thoughout summary of CSCO’s over-ambitious mobile data forecast. As you have noted; now lowered twice in as many years.

    One metric which caught my attention from CSCO’s forcecast:

    “Cisco’s predictions are scaled back from last year, when they predicted an 18 times growth rate in wireless network data between 2011 and 2016, with users consuming 10.8 exabytes per month by 2016. Now Cisco thinks traffic will hit 7.4 exabytes per month in 2016.”

    In just one year, CSCO’s forecast went from 18x growth rate in wireless to 13x between 2011 and 2016. (10.8 exabytes vs. 7.4 exabtyes; a 31% decrease).

    Definitely a negative trend which does not bode well for CSCO, or any other the carrier ethernet equipment providers.

    I suppose it all comes down to “significant amounts of guesswork, just like anyone else.”

  6. we can always compare Cisco with Ericsson Mobility Report:
    http://www.ericsson.com/ericsson-mobility-report

  7. The Cisco VNI Global Mobile Data Traffic Forecast methodology rests on a combination of in-house estimates and forecasts, direct data collection and third-party analyst projections from organizations such as Informa Telecoms and Media, Strategy Analytics, Infonetics, Ovum, Gartner, IDC, Dell’Oro, Synergy, ACG Research, Nielsen, comScore, Arbitron Mobile, Maravedis and the International Telecommunications Union (ITU).

    The Cisco VNI methodology begins with the number and growth of connections and devices, applies adoption rates for applications, and then multiplies the application’s user base by Cisco’s estimated minutes of use and KB per minute for that application. The methodology has evolved to link assumptions more closely with fundamental factors, to use data sources unique to Cisco, and to provide a high degree of application, segment, geographic, and device specificity.

    Cisco has consistently revised VNI predictions to reflect new data and trends. The recent mobile traffic growth forecast was adjusted for several reasons , including the fact that mobile laptop traffic growth is slowing, particularly in Western Europe. As people replace mobile laptops with smartphones and tablets which use less data, laptops will not be used to access mobile networks to the extent previously projected. If we include offloaded traffic with mobile traffic, the difference between last year’s forecast and this year’s is much narrower. Also, people are using Wi-Fi small cells more often, shifting more data from mobile networks.

    As we explained to Mr. Farrar last week, we have consistently used what we believe to be the most accurate currently available data to support our methodology in updating the forecast each year, rather than artificially imposing consistency with previous estimates. – Jim Brady, Cisco Public Relations

    1. The worst issue regarding these Cisco report is basically that they do not state at all which results have been really measured and which ones are rather approximations. Havin said this, to me it seems that these reports are not suitable at all to give a solid basis even of the state-of-the-art today.

Comments have been disabled for this post