Mobile broadband traffic volumes: Watch out for dodgy offload & optimisation claims and stats

The next two weeks are going to be uncomfortable for a lot of "mobile data traffic management" vendors. Whether they have WiFi offload solutions, video optimisation offerings, small cells, policy enforcement boxes, charging and "personalisation" platforms or various other bits of network-control machinery, they will all have to think carefully about their pitch for MWC.

The problem is that the (in)famous scissor diagram of diverging mobile data traffic and revenues, that we've all seen 1000 times..... seems to be wrong. Increasingly, we see evidence that it looks more like a pair of pliers, with the lines curving back together rather than expanding exponentially away from each other, especially in more mature markets.

Now I'll be the first to say that the "big" number of TB, PB or EB delivered doesn't tell you much about actual busy cell / busy hour congestion, but it's still an interesting data-point, especially as it is so central to  many spectrum-lobbying and solution-marketing arguments.

We're in the middle of financial reporting season for telcos, and so far we haven't seen much hand-wringing about unsustainable traffic growth this time around. Vodafone's 20% data growth matches its revenue growth. AT&T's is down to about 40%, far less than it predicted in its FCC filings for its for the ill-fated T-Mobile acquisition.

Telefonica and other operators' results are due soon too. Then we've got the new Cisco VNI study and forecast out this week (I'm expecting a downgrade of its mobile data forecasts), and at some point we should get a full-year 2011 report from Ericsson & Akamai as well. The last two updates have shown different patterns - 8% Q-on-Q growth for Q2 rebounding to 18% in Q3 with the note that there are "large differences in traffic levels across markets and regions".

To be fair, it's still too early to tell if flattening data volume growth in more mature markets is a major trend, or a temporary one-off blip/correction as many operators shift to tiered plans. Yet other factors are also in play - especially WiFi use, as well as some in-network policy and optimisation.

So if the "tsunami" has indeed receded, then a few things will occur:

• Every single vendor will claim "See? It was us that saved the world!" and give some form of data as supporting evidence, rather than admitting the problem was overstated in the first place
• If you add together all the "we saved 20%", "our solution saved 40%" claims, we should be back at zero by now, or even negative traffic.
• The marketing will stress "If you don't carry on buying our solution, congestion will reappear again tomorrow, even worse than before"
• There will be 10x more emphasis on network-resident kit having solved the problem, rather than simple pricing plans (or wider free WiFi availability, or better-designed apps) changing user behaviour
• There will be 20x more emphasis on traffic "tonnage" than on signalling load, even though for many networks that its *still* much more of a real problem
• Lots of vendors will overlook the fact that most traffic on many networks still comes from laptops rather than smartphones.
• There will be subtle digs at the other options, as vendors realise that the operators don't really need ALL the solutions they are being pitched. Expect "WiFi offload is more cost-effective than optimisation", or "Adding extra radio resource with small cells is easier than messing around with complex policies", or "Dynamic pricing & charging by user profile is better than trying to guess applications with DPI"

Most of all, though, we will see a broad range of highly questionable stats ("dodgy", in English English), especially in the PR bombardment at Barcelona, when a big slab of hyperbole is often necessary to get attention from the media.

Offload is probably going to be the biggest culprit. I'm expecting to see a lot of hype from vendors and WiFi aggregation providers claiming that X% of data traffic has been offloaded on Operator Y's network, saving $Z in incremental capex/opex.

The key question to ask (especially if you are at an operator) is "What % of total WiFi traffic do you think is actually genuine offload?". As the chart below illustrates, smartphone WiFi is much more than just offload. There is also

• Baseline private WiFi use driven by the user, which was never intended for 3G/4G use
• "Elastic" WiFi use, where offload actually increases total volumes compared to what would have been done over cellular. I've heard a suggestion from an operator that this can lead to a doubling or trebling of total usage.
• WiFi Onload, where a third-party operator diverts traffic to its WiFi and away from a rival's 3G/4G, in the hope that it can use the channel for extra services like advertising or location-specific deals (eg O2 WiFi)
• App-driven WiFi (not on the chart), where a specific application drives WiFi connectivity and choice of access (for example, Skype's tie-up with Boingo, or Sky TV and its WiFi network The Cloud in the UK)

Also, it's important to think about laptop MBB differently to smartphones. Many laptop users "onloaded" to 3G only because there wasn't enough WiFi about to begin with. Increasingly, there is available WiFi, so they're not really "offloading", but merely going back to what they hoped for in the first place. A subtle difference, but important as it may mean they abandon dongles completely at some point.


A couple of discussions I've seen on Twitter and elsewhere suggests that "elastic" and some portion of private WiFi might get counted as "offload" to pump up the stats. Caveat lector....

We may also see some numbers from the video optimisation vendors, about the amount of traffic "saved". Those numbers are perhaps more realistic (ie volume in vs. volume out), but it's also important to understand "volume the user didn't bother with in the first place" with behavioural change from pricing shifts.

And of course, that's the problem. None of these things is truly independent. There are lots of feedback effects, accidents of user demographics on specific networks, impacts of shifting mix of devices - and perhaps, the impact of application and content providers doing their own "self-optimisation", for example with adaptive bitrate streaming.

Ironically, however, the main real underlying reason for flattening data growth is Cisco's VNI mobile data forecasts numbers themselves. They have been so scary, used by some many people, and seen so many times, that they seem to have prodded everyone else in the industry into action, to make sure they didn't come true.

In other words, the most important technology involved in mobile broadband traffic management is probably Microsoft Excel.

(If you found this post interesting, please sign up for the email list at the top of the page. I'm also @disruptivedean and @DApremium on Twitter)