About Me

My photo
Web person at the Imperial War Museum, just completed PhD about digital sustainability in museums (the original motivation for this blog was as my research diary). Posting occasionally, and usually museum tech stuff but prone to stray. I welcome comments if you want to take anything further. These are my opinions and should not be attributed to my employer or anyone else (unless they thought of them too). Twitter: @jottevanger

Friday, November 16, 2012

The Mystery Trend

This is a little case study of digging into our web stats a bit to understand a recent trend. Doing this sort of digging is pretty addictive and can, of course, throw up some interesting insights.
Recently we noticed a bit of an odd trend in our user stats on iwm.org.uk. Although direct comparisons are imperfect we're happy overall with how they've looked in the year since we swopped to the new site, with decent improvements in all the main crude metrics that might mean, well, something or other that's meant to be good. Last month we exceeded 500,000 visitor sessions for the first time. Here's how things have looked week-by-week from early May to mid-November

What with a general above-target performance, an odd summer (a big dip - thank you, Olympics!) and peaks in months that historically haven't been so prominent, the trends have been a bit of a departure from the norm so I thought I'd dig into it. It turned out that direct visits had increased noticeably from mid-September onwards - like, doubled.

Further, the growth was in the mobile segment and, more specifically, in iOS devices - equally iPhones and iPads. But why? This is weird. I mean, we're seeing the rapid growth in mobile traffic like everyone else, but whilst the numbers of iOS users overall increased by a couple of percent between May and November, direct visitors increased 4-fold just from September to November (whilst direct visits from Android are pretty much flat). Yep, that much.

To me directs from mobile means visits referred from apps, and in this case I certainly doubt that all those Apple fans - and only them - have suddenly bookmarked our site or emailed it to each other. So it's probably about referrals from apps, but which ones? There are few self-explanatory clues, either in where they arrive or in the timing. Plenty land on the home page:

our branch pages and visiting info:
(but not Duxford)

and a healthy number also land on collections record pages:

If the trend is evident for all of these it can't be from a single suddenly-popular app, right?
Predictably, our site gets healthy spikes of traffic from places like Reddit, the BBC or newspapers, which we shows up in our referrer traffic; perhaps the app equivalent of these websites is responsible for this other trend. But at first look the spikes don't seem to correlate, as for instance this graph of iOS visits from Reddit shows. I'd expect referrals from a Reddit app to be similar and is not much like the "directs" trend:

It just seems odd that there's been such growth in whatever apps are responsible in that period. We have a couple of apps of our own, of course, but again looking in detail at the pages they may link to doesn't provide an explanation. At first I suspected social media apps. Facebook & Twitter have not suddenly quadrupled in users over 8 weeks, but it seemed a plausible explanation for a rise in direct traffic to such a range of pages. Then I thought about search. Perhaps there's a newly popular search app on Apple devices? Aha, perhaps we're onto something:


So, referrals from Google drop off amongst iOS users at the same point that direct visits increase. Not being an iOS user I still have no idea of what's caused this, though I assume that there's a new appified version of Google that's being taken up rapidly. I think I'll need to do some more detailed analysis to make sense of this, but it looks like (a) the rise in directs doesn't mean an absolute rise in iOS users, and (b) our overall increase in visitors has nothing to do with this trend. This actually makes things a little worse for us because we now have less information about searches that bring traffic to us. Hmm.
I'd be interested in whether any of you lot seen a similar trend in direct traffic on iOS devices over recent months? Comments please!

PS if you found this blog post hoping to download stuff by the Mystery Trend, sorry to disappoint you. Interesting band though. More here.

4 comments:

bigiain said...

Does the Sept 19 iOS 6 launch line up with the start of your anomaly? Perhaps you're losing Google maps referrals?

Frankie Roberto said...

I think this is down to the iOS 6 update, which switched the in-built Google search in the standard Safari app from non-SSL to SSL.

This means that Google Analytics no longer has access to the referer [sic] header, so visits could as direct traffic rather than organic search traffic. See: http://searchengineland.com/ios-6-removes-all-google-search-referer-data-134560

I've seen the same pattern on all the sites I have google analytics access to.

I'm not sure why this has an overall positive effect for you - in theory it should be neutral (simply with a displacement in how iOS-search-originating traffic is classified).

Seb Chan said...

Hey Jeremy

Take a look at the OS versions and you should be able to see if it is iOS6 - in which case, given the nature of the landing pages, might suggest the impact of Apple Maps.

Maybe.

Jeremy said...

Good call, all of you, thank you for the pointer. That's it. It coincides perfectly with iOS 6, and indeed the version number supports that - thanks Seb, I'd not looked at that.
I'd got myself confused about when 6 happened, because in July there's another instantaneous switch in Apple data: traffic that had been reported as iPhone or iPad all became iOS (presumably a GA change, on reflection, since it was overnight).
Frankie gets the prize though - it looks very much like search rather than maps, given the huge variety of landing pages, and the SSL explanation is perfect. What a bummer - even more irritating than the "(not provided)" search term, which at least lets us know it's a search referral.
Thanks again for your input, no more mystery!