The State of Mobile Subscription Analytics

19th November 2017

By Tasman

Mobile subscriptions are eating the world. This is not just a feeling only we have, it can be quantified: according to Gartner, 2015 saw the mobile subscription market being valued at 30 billion dollar. By 2017, this has grown to almost 80 billion dollars, with the exponential trend only being continued further.

Of course, this makes perfect sense: if you think of mobile apps being monetised, mobile subscription is the only viable path forward. It guarantees recurring revenue (unlike upfront payments) and it simplifies the app experience to two flows: a free one and a paid-for one. The canonical gaming model with a lot of calls-to-action and in-app purchases (target the 0.1% of your users who are ’whales’ spending hundreds of dollars) is not as attractive as it once was: it is confusing for the users, and it profits from a negative upsell (pay so you can continue playing) rather than a positive message (pay to get more access).

These are just a few of the problems:

  • Data flows of App Stores are not easy to set-up and are hard to maintain; typical receipt environment is designed for production and not for analytics. But that makes it very hard to do churn modelling.
  • Existing, paid-for tools have great out of the box event models for mobile subscription, but they make it hard to extract and really own data. Snowplow does offer that ownership, but has a less attractive out-of-the-box environment.
  • A lot of the data modelling and reporting that’s being done in mobile subscription companies is equivalent and straightforward: it means that companies are basically duplicating work as there are no proper open source or third party services available to do this for them.
  • Data Scientists starting out in a mobile subscription company are right now spending a lot of time and energy modelling these simple mobile subscription questions; things like churn, conversion, retention, etc.

With this in mind, we propose the following:

  • A set of canonical events ‘ai_tasman_screen_view, ai_tasman_purchase, etc’ to use in the event platform of your choice that cover most of the mobile subscription analytical needs, and standardise the data structures we operate on. This is so crucial for the rest of the funnel that it can certainly be released as an open source contribution rather than as a proprietary service.
  • Basic set of data models.
  • Sophisticated set of data models setting the stage of churn and conversion modelling; and integration with attribution partners.
  • Billing server flows
  • Churn Reporting
  • Churn models both as structures (untrained) and as REST API’s (trained).
  • Campaign Attribution analysis.

Buckle up for the ride!