For at least a year now I’ve held to the theory that the huge uptake in Chrome we’re seeing on the mobile web is mostly due to Samsung using its own version (based on Chromium 28) in its high-end smart devices from the Galaxy S4 on.
Yesterday Krijn started tweeting about mobile stats he had, and it turned out he was willing to share. He gave me data on about 100K mobile and tablet hits in Q2 on a project of his he’s worked on for
ages five years. I took the data gratefully and created a table.
Conclusion: Of Chrome users, 25% uses Samsung Chrome — this amounts to about 5% of all mobile visits to the site. On the one hand this proves that Samsung Chrome is a Thing — on the other hand I had expected a much higher percentage. So my theory isn’t right, but Samsung Chrome is still important.
Samsung Chrome is not exactly the same as Google Chrome. It’s still at version 28, and there are some other differences. For instance, it supports scoped styles, something Google Chrome never did without a flag. I feel we should take more care to test specifically on Samsung Chrome and separate it from Google Chrome. So buy that Galaxy S4 or S5 and start working.
The statistics come from the Royal sites, which are targeted at a Dutch audience looking for holidays, primarily on the Mediterranean coast, but also elsewhere. The first surprise was that about 45% of total hits came from mobile devices or tablets.
Snapshot: person on couch wanting to book holiday, too lazy to get up to computer, uses tablet or mobile device within reach instead. Still, 45% remains a lot.
More than 50% of these mobile/tablet hits come from Apple devices. Almost all the rest comes from various Androids, and it’s here that the stats start to become truly interesting.
You can browse the full analysis here and compare it to StatCounter’s analysis of Q2.
This was the general browser make-up:
|Browser||Percentage ROYAL||Percentage STATCOUNTER|
That’s a pretty close match — there’s only a 5% difference that has gone from Android WebKit to Safari. It’s fun to theorise about affluent people (Apple!) booking more holidays than less affluent ones (old Androids!), but I’ll resist that lure.
However, where StatCounter gives you the bare browser numbers at best, I could dig deeper into the Royal stats. I focused on Chrome, since that browser is my biggest worry right now. So what about those 20% Chrome hits?
|Chrome iOS||42%||12.5% of all hits from Apple devices|
|Google Chrome latest (34/35)||31%||Only 7% of this 31% (= about 2% of all Chrome hits) come from Google Nexus devices|
|Samsung Chrome 28||19%||Default browser on Galaxy S4 and up|
|Samsung Chrome 18||6%||Old default browser on Galaxy S4 and up|
|Google Chrome 33||1%||These people didn’t upgrade their Chrome|
Surprisingly, Chrome on iOS is the biggest Chrome. It accounts for 42% of Chrome hits, and 12.5% of Apple device hits. One in eight Apple device users uses Chrome, in other words. That’s WAY more than I expected.
In the middle of Q2 Google Chrome was updated from 34 to 35, so I combine those two browsers as “Chrome latest.” These users use a downloaded Chrome (except the slight number of users actually surfing with a Google Nexus, where Google Chrome is the default browser). Again, this is more than I expected. I thought consumers don’t download browsers — it seems I’m not entirely right here.
Then comes 25% of Chrome users actually using Samsung’s Chrome, which was first at 18 and then got upgraded to 28. This is the hidden group that no web developer except for me pays attention to.
Finally, 2% other Chromes. These are mostly people who didn’t update their browsers, I guess.
The stats were gathered by Google Analytics. Krijn gave me CSV files of the Q2 devices and browsers. Unfortunately these two are separate — despite more than fifteen years of experience with web stats, the first package that actually cross-references browsers with devices or OSs still has to be written. (I mean, it’s not THAT difficult to calculate that 23% of Android users use Chrome, and show that fact clearly in the reports, is it? But this seems to be a curious blind spot of analytics package creators.)
I took the devices file, which also includes a browser version number — but not a name. (Why not? I have no clue. Blind spot etc.) I removed all devices that had less than 10 hits, because I had to manually go through the file and this removed about 75% of the lines without affecting the overall stats too much. I then removed a lot of extra stuff I didn’t need, and was left with lines like this:
Samsung GT-P7310 Galaxy Tab 8.9,4.0,47
The first item is clearly a device. The second is the browser version number, the third the number of hits. My job was to match browser version numbers to browser names, and fortunately that was pretty simple most of the time. The
4.0 above clearly refers to Android WebKit; Chrome has much more complicated version numbers like
35.0.1916.141. I also added a device type: phone or tablet. So I manually expanded the line above to the following (and yes, I had to do this for every single line — more than 300 in total. I’m glad I removed the long tail.)
Samsung GT-P7310 Galaxy Tab 8.9,4.0,47,Android WebKit,tablet
In the table I ignored device types for now; I may delve into those at a later date, but right now it wasn’t my main research question.
If you’re interested, the sanitized data is here.
Granted, 100K hits is not really a lot, and I’d love to repeat the experiment with a much larger set of data. On the other hand, that much larger data set just isn’t there, and any data is better than no data.
Update: Peter Gasston tweeted he’s seeing roughly similar Samsung Chrome numbers: 32% of mobile Chrome users. Chrome iOS at 17% — distinctly lower than in the data I went through.
Update to update: I checked Peter's numbers myself, and it turns out only 10% of his Chrome mobile visitors use Samsung Chrome. Also, it seems Google Analytics can't distinguish between Chrome and Android WebKit.
I’m around at the following conferences: