A Guide to Understand App Tracking in the Mobile Space Part Two: Tracking conversions and solving download lag
About The Author:Michael Dewhirst is a veteran senior IT professional, entrepreneur and a pioneer in the mobile industry having worked in the sector since the early WAP days of 2000. Michael began his career in high tech development roles with British Aerospace, Deutschebank and JP Morgan, and worked on the launch of Vodafone Live! in 2002. Prior to joining StrikeAd, Michael was CTO of The Daily Mail Digital Property Group, driving the development of their real estate portals and apps, including an augmented reality property search mobile app. In 2003 Michael founded DevzeroG, a global advertising software company with clients such as News International, The Guardian and NTT and was sold to Adstream Pty in 2007. At StrikeAd Michael is responsible for all products, systems and integration with third parties.
In his last piece, Michael Dewhirst, CTO at StrikeAd, discussed numerous problems which need to be addressed in the mobile advertising world, looking particularly at the shift away from UDIDs, and back towards the old veteran, the cookie, for tracking mobile app downloads. But here he takes it one step further, and examines the benefits of using cookies as opposed to UDIDs, particularly regarding the download lag often encountered with apps, and how cookies can address this issue.
As well as being able to track downloads, advertisers also want to know what happens afterwards.
Just think – if a user goes to a website, do the developers of that website want to know which parts of the site the user uses most and how? Of course they do! That way they can work out which parts of the web site are most useful, make them better, add to them and so on.
It's the same situation with the app developers, except that often this information is not collected and those installed apps out there are in a ‘black box’. But app developers do want to know which parts of the app are most used and make them better.
The solution is to use the same code inside the app that is used for tracking the download, to also record various events such as registrations, game level play starts, completions, etc.
This ability is a reality and a solution now and the StrikeAd app tracking SDK does all this and more. The StrikeAd app reporting solution can even pass data into third party analytics packages such as Omniture so it can be examined and analyzed there, together with other tracking data, allowing all reporting to be viewed in one place.
As well as collecting and reporting this data, the other challenge is collecting huge volumes of it and aggregating it into reports, which can be viewed in many different combinations. For example, you could pull out a report that would tell you “how many iPhone users in London during lunch time on Sunday completed a level” or “how many new issues of the magazine were purchased in the app in New York during a particular week” and so on. All in real time. Imagine the data that needs to be processed for this, bearing in mind that Angry Birds, for example, has been downloaded more than 500M times across the globe.
This sort of data is pretty standard on the web, but in the app world it’s actually quite new.
Dealing with the Download Lag
The other thing to bear in mind is that downloads are just like other conversion events and can happen sometimes days after the click.
You see a banner for some exciting new app.
You click on it and this event is recorded.
You end up in the app store, download the app and then turn your phone off as you get on a plane before you get a chance to launch the app.
You then forget about the app.
You get off the plane, go about your business, and days pass.
Finally – days later – you see the app and launch it.
Since many current solutions on the market will not try and sift through days of UDID data – as there could be billions of records to process – the above situation will not be accounted for, and whichever media buyer ran the ad will not be attributed to having driven the download.
On first look this is good for the agency and advertiser, since it means fewer downloads you have to pay for. However, if you look deeper, all it means is that the statistics will be incomplete. Potentially productive inventory is not being re-targeted, there are fewer downloads than there could be in the long run, and CPD stays relatively high, due to the long lag between the click on the banner and the launch of the app.
Until you’ve launched it – nobody knows you have it
Don’t forget, the app only “wakes up” for the first time and starts to do anything when it is first launched after the download on the device. Until you’ve started it at least once, it is asleep on your phone, like a new Sky Box you just bought that’s still in its cardboard box that hasn't yet been unwrapped or plugged in.
With a cookie approach it doesn't matter how many days the app lays dormant, as the information is kept in the cookie – on the users’ device – and you can attribute downloads which may have happened days ago extremely easily.
There is no need to sift through data on the servers, trying to match up UDIDs, since the cookie and the tracking data is stored on the device.
The 'Post-View' Download
Post-view conversions are an extremely common approach in online advertising and the process has been around for years. It consists of attributing a conversion such as a purchase, form fill – to a banner view, rather than a click. In mobile, however, this is still not very common, even in mobile web campaigns.
In fact, for the mobile app industry it is a quantum leap in the future.
With an app, the post view experience would be something like:
You see an ad for some exciting new app a few times (and a cookie is set to record this).
You never click on the banner but you decide to have a look at the app later on.
You go to the app store or Android Market, find the app, read the overview, download it.
The app starts, pops up a browser, the cookie in step one is read and the attribution occurs.
The trouble is, even though the above solution is technically feasible right now, many advertisers and agencies will not currently consider such a download attribution process.
When downloads are sold and bought on a cost per download basis, it is in the interest of the person fitting the bill to pay for as few as possible, whilst getting the supplier to generate as many as possible, arguing that they did not truly do so.
However, it is actually in the advertisers’ direct interest to take those post view conversions into account, since the media buyer would then be able to optimize them.
The result to the advertiser would be lower CPA/CPD and more downloads – all very worthwhile benefits! In turn it would allow for lower CPA/CPD to be commanded in the market, perhaps $0.20 - $0.30 cents instead of the current $1.00-$2.00.
Adoption from all Parties
As I stressed in my first piece about mobile app tracking, it's clear that there are in fact many solutions out there which can make mobile app advertising much more successful and cost effective, but it will take the triumvirate of the advertiser, agency and media to adopt them. Without all three of these parties understanding the options available and deciding to utilize them, the industry will not move forward – or at least not easily and not quickly.