Web analytics tracking mobile

The recent PR furore about the reception issues on Apple’s iPhone4 has got me thinking again about the impact of smart phones on digital analytics. As the world moves to mobile the digital analytics industry may be taking a significant step backwards in its ability to provide actionable insight. Tracking of mobile devices is still developing, so just when the Web Analytics industry thought it was getting to grips with clean data it now has a real challenge on its hands. Whichever way consumers react to Apple’s current problems, the smart phone is here to stay and advertisers who will spend £61 million on mobile in the UK this year, rightly want to know the impact of this spend across display and applications (apps).

Apple have taken a strong stance on the inclusion of analytics within apps to date. Steve Jobs was rather vocal following the outing of the iPad by Flurry Analytics earlier this year  and there was speculation that this impacted Apple’s view on all analytics tools for the iPhone. Apple restricted the use of embedded analytics tracking code in any App and also stated that that app tracking must be an opt-in system. This obviously has significant user experience issues, not to mention questions over how representative the data is. Apple will provide some data to developers and advertisers: developers can see how many downloads their app has had, along with revenue and some limited geographical usage; advertisers can see the aggregated anonymous geographic location install base via Apple’s new mobile advertising platform, iAd. The app tracking also allows developers to see how people are using their applications, which areas of the app are popular, which buttons are used most frequently, how often the app is launched, and so forth.

So how does this work and why is this limiting to the digital analytics industry? Well, most analytic tools embed a small section of code into the app that make a call to a data collection server hosted on the Internet – which obviously means it must be online. This raises various challenges with mobile devices, as they change location and more importantly will have varying levels of Internet access. This raises several questions, firstly if you have a particular work flow in your app, then you may have embedded tracking on each key stage of the process. If the phone then goes offline, then the tracking calls may not be completed in a timely fashion, or even at all. As such it would appear from the data that the user didn’t complete the process leading to very misleading results. Secondly, now that all of the UK carriers have introduced limits on how much data can be used, it raises an interesting question over how many aspects of the app should be tracked. Admittedly the size of the tracking requests is minimal (typically less than 5Kb per call), but none-the-less, repeated use of features with tracking included will push up the data usage on the device.

So, with the apparent antennae issues with the iPhone4, you have to wonder how reliable and representative any tracking data will be? If certain geographical areas have low signal strength, then it is far more likely that the tracking calls will be lost and as such it would skew the audience figure of your app to geographical areas with stronger signal strength.  

So how reliable and representative is data collected from mobile devices? Well, given the nature of the data collection it’s extremely difficult to validate – you cannot easily test your app in every location that your target audience are going to be using the app. What happens if they use your app extensively whilst on the tube without Internet access? As such a definitive answer on validity is probably unlikely. However, some data is definitely better than none. The issues with reception loss will likely average out over time as it will probably affect all tracking calls equally. Companies such as Flurry Analytics and AppClix are working on these problems and already have some techniques to deal with them. However, as signal strength will vary with location and device there is no universal answer and so we’ll be keen to see how this progresses.

The key, as with all analytics work, is to put the data in context and take account of any data collection issues that may affect our clients mobile activity. Data may be king, but insight is divine!