I talked about mobile data consumption previously here in reference to some of the Admob produced stats. Since then, Admob has been acquired by Google and their stats have been quoted widely. The October metrics report caused a recent spate of articles like this one from the register. The data was also included in Mary Meeker from Morgan Stanley’s well received presentation at the recent Web 2.0 summit.
As I have said before, I like these metrics from Admob but I do think they are taken at face value for being a direct proxy to data consumption without sufficient scrutiny. I don’t intend to rehash all my criticism of the data again and I believe that Admob report the statistics they get correctly and without bias. What I want to do is point out some inconsistency in their numbers, and also to show some interesting stats from elsewhere that enable us to conclude that measuring data usage in mobile is still a relatively immature field.
The major problem with the Admob stats is geographical skew toward the US. The US is, no doubt, a massively growing market for mobile data, however the Admob stats would have us believe that the North American mobile data usage market is over 50% of the world’s total. Below is a chart based on information from a Cisco prediction of mobile data traffic by region and the Admob data. (compares the full 2009 figures from cisco to the oct ‘09 from admob)
You can see there is quite a difference, with North America only taking a 20% of worldwide total. In the Cisco numbers Asia and Western Europe combined make up over 70%, almost double the Admob combined figure.
OK, so where else can we look at stats for mobile data? One of the sources from Mary Meeker’s presentation was the Net Applications data. Traditional internet metric companies are now publishing data on Mobile usage. The Net Application data for mobile OS’s looks like the following over the last year.
Again, however, some things don’t quite add up; Symbian went from 4% to 30% in the first 3 months of the year, there’s seems no market driven explanation for this. Also, Android and RIM have very small market share, with Android showing little impact over the last 6 months. The only other source of similar data I could find comes from StatCounter, their data is very different.
These numbers show iPhone and Symbian on similar terms, with RIM showing a steady increase and a small upturn in recent months for Android.
So what does this all mean? Unfortunately, I think the answer is “not a lot”, other than you should be careful when presented with stats and predictions around mobile data.
One last point; whilst the comparison of mobile platforms with respect to internet usage is interesting I would be extremely interested to see what impact data plans have on the result. I am sure there is a massive uplift in data usage with those with flat rate data plans (of course) and that the different OS platforms will be sold on these plans to a greater or lesser extent. My hypothesis is that this would even out data usage, per platform, per device quite a lot. I would go so far as to say without including this aspect, comparison of data usage on different platforms is fairly meaningless. Anyone know of such data out there?
Social comments and analytics for this post…
This post was mentioned on Twitter by Puckrin: My blog post on mobile data use across smartphone platforms http://bit.ly/7BN8oz #Symbian #Admob…
I’ve always felt the same way about any stats about web browsers (market share, typical screen resolutions, whether they have JavaScript on or off etc.) – there’s just too many variables to get accurate data. Web-servers typically log requests along with the user-agent string and that’s ultimately where a lot of this data comes from. Trouble is, not every site will make their logs available (or embed something like Google Analytics so that a third party can measure page hits) and user-agents can be faked and/or hidden by proxies, web-browser or ISP caches might be preventing re-visits to pages being counted etc. etc.
As for data usage, that presumably would include all internet traffic which goes way beyond what is consumed by web-browsing. A few minutes listening to Last.fm on Mobbler will probably eat as much data as half an hour of web browsing. Every email sent, every photo uploaded, every online multiplayer game – it’s all data not going via the web. Other than network operators (who must be measuring this data usage for billing purposes) I don’t see how anyone else could accurately measure all the data being used by a particular flavour of mobile.
Even if you do have accurate data what does it actually tell you? If one platform uses more data than another does it mean that its users do more stuff online or does it just mean that platform and its apps are using inefficient protocols for the data transfer?
It’s good to see it laid out this way.
Nobody seems to have pointed out that AdMob have a platform for injecting their ads into iPhone apps *only*. Surely this would put a huge bias towards iPhone? You might ask, why did they choose to create this platform for the iPhone and not any other OS? It probably comes down to the fact that people distributing ads know almost for sure that an iPhone user has disposable income, so they want to target the iPhone. So, if I want to spend a million dollars on mobile advertising and I know that there are 20 million iPhones and 20 million Symbian phones (not real figures, obviously) then I would have to go for iPhone, because I don’t know how many of those Symbian users are rich and how many aren’t.
The other thing to consider is the sites they serve ads to. Apparently it’s 15,000 sites. I’d really like to see a list, but all I could find was 10 case studies on their homepage. These seem heavily biased towards US brands. Obviously, we live in a global economy nowadays, but don’t tell me that MTV for example has as significant a role in Western Europe as the US.
Marketshare figures don’t give you the full picture about the smartphone market, but neither do the stats of an ad agency. The sooner we all learned that, the better.