I'm writing a CSS book

JavaScript archives Compatibility Thidrekssaga QuirksBlog Donations CSS book Politics Mobile About COH

contact

Why AdMob’s reported iPhone and Android market shares are inflated

AdMob, the mobile advertiser that was bought by Google some months ago, has released its latest market share figures for the mobile browsers.

Their main findings have already been discussed extensively:

1. Smartphones are on the rise; 48% versus 35% last month.
2. Feature phones are falling quickly; 58% to 35%.
3. Still, the absolute number of feature phones rose by 31%, which means that the market as a whole is growing rapidly.

The AdMob report, however, is not about browser market share but about ad impressions. And that may matter a lot. Unfortunately we don’t know how much it matters.

Market share — of what?

Let’s go through AdMob’s careful methodology notes (last page in PDF; emphasis mine):

For every ad request AdMob analyzes information available in the user’s mobile browser [i.e. the user agent string]. From this, AdMob determines device capabilities and more using open source tools and a variety of proprietary techniques. The result is a snapshot of the devices viewing the more than 10 billion monthly ad requests and impressions that flow through AdMob’s network. [...] AdMob also serves mobile ads into iPhone and Android applications. The traffic from these applications is included in the Metrics report.

In other words, the figures AdMob gives are not pure browser market share figures. They include requests from apps that happen to have AdMob ads in them.

That’s vital (and it’s something that’s mostly ignored by other website that report on the AdMob stats). Let’s take a closer look.

1. AdMob allows developers to deploy ads in their iPhone, Android, webOS, and Flash Lite applications. Thus, these browsers tend to be overrepresented in AdMob’s stats.
2. From Adobe’s data we may draw the conclusion that Flash Lite is mainly supported by feature phones, many of them Japanese. The only partial exception is that it’s supported on some (but not all) Nokia Symbian smartphones. Thus Flash Lite ad impressions will not have a meaningful impact on the smartphone market share report everybody concentrates on, except a little bit for Nokia.
3. That leaves iPhone, Android, some Symbian, and webOS. The latter’s market share is marginal; 3% in the US, “other” elsewhere (page 7 of PDF).
4. Thus we’re left with all of iPhone, all of Android, and some Symbian.

I am forced to conclude that AdMob’s iPhone and Android figures are inflated. The true browser market share of Safari iPhone and Android WebKit is less than the AdMob report suggests.

All this is not AdMob’s fault, by the way. It’s being very clear about what it measures. Most analysts, interested only in the horserace between Apple and Google, tend to ignore that, though.

Exactly what is the inflation rate? How many iPhone and Android impressions are from apps, and how many from the browser visiting websites? We have no way of knowing. I’m not even sure if AdMob has any way of knowing.

If it does, I would like to request splitting out these figures. The AdMob monthly overview is one of the most important mobile browser market share reports we have, and any attempt to make it more precise will be gratefully received.

Figures

With all that said, what figures does AdMob report? It does not actually report browsers, just operating systems. Worldwide states, smartphones only:

1. iPhone OS 50%; inflated; includes iPod Touch
2. Android 24%; inflated
3. Symbian 18%; slightly inflated
4. RIM OS 4%; BlackBerry
5. Windows Mobile 2%
6. Other 2%

I find RIM curiously low. Of course the default BlackBerry browser is not very good, and that will certainly negatively affect its market share, but 4% is still very low.

An explanation might be that BlackBerry is now extremely popular among the youth, who mainly use it for chatting with their mates (as well as calls and SMS, obviously), but not so much for browsing.

Adjusted figures

Let’s try to adjust these figures. I’m going to make up numbers; feel free to make up your own.

I will assume that one third of the iPhone and Android numbers and about one ninth of the Symbian numbers come in fact from in-app ad impressions and do not count for browser market share.

1. iPhone OS 45%; includes iPod Touch
2. Android 22%
3. Symbian 22%
4. RIM OS 5%
5. Windows Mobile 3%
6. Other 3%

Not that much of a difference, although iPhone drops below the magic 50% and Symbian is now equal to Android.

US figures

Of course, pundits are mainly obsessed with the US figures. There is no world.

1. iPhone OS 44%; inflated; includes iPod Touch
2. Android 42%; inflated
3. RIM OS 7%; BlackBerry
4. webOS 3%; inflated
5. Windows Mobile 2%
6. Other 3%

Note that the iPhone share is smaller than in the overall world. Why? Does Europe have such a significantly higher iPhone penetration than the US? Or higher app use? I doubt it. But I have no idea what else could be going on.

Anyway, the story of the day is how Android is nipping on the iPhone’s heels, and that’s of course always good for a nice story. Do the figures bear this out?

It could be. After all, I assume that we can subtract roughly the same part from both OSs due to in-app ads. But, and that’s an interesting thought, might the rise of Android in the last six months or so be due to a significant increase in in-app ads, instead of web use and/or increasing sales market share? Have Android apps increased so sharply in the past few months? No way to tell.

Still, the overall take that Android is nipping on the iPhone’s heels in the US seems to be generally accurate. In the world as a whole, the effect is less visible both in absolute and in relative numbers.

To close off, let’s do the same random calculation as with the worldwide figures, and let’s assume that webOS impressions, too, are one third in-app.

1. iPhone OS 41%; includes iPod Touch
2. Android 39%
3. RIM OS 10%; BlackBerry
4. webOS 3%
5. Windows Mobile 3%
6. Other 4%

The main result is that BlackBerry’s share increases, and that’s something that I think is close to the truth. (Again, I’m guessing here. I could be totally wrong.)

Conclusion

Concluding, it is not really possible to draw meaningful conclusions about mobile browser market share from the AdMob numbers until AdMob splits them up into website and in-app ads. Be very careful in using these figures.

Comments

Comments are closed.