By Charlie Billingham

In their standard form on-app analytics data has always been about telling us what behaviours our users have carried out in the past. However, with the all new Firebase Predictions, Google have taken a big step forward in being able to provide us with insight into what the future might hold for our users.

Back to blog home

Predicting the future

Predictions is a new report and functionality within Google Analytics for Firebase which aims to shed insight into the probability or likelihood that a user is going to carry out a particular action or behaviour within the next 7 days. This user level dimension comes in the form of a value between 0 - 100 with 0 being the least likely to convert and 100 being the most likely.

Although Google have decided to keep the finer details of their methodology under wraps, our analytics team has been carrying out similar advanced analysis techniques for many of our clients for quite some time:

  • Google use regression modelling which is a machine learning advanced statistical technique used to find to what degree certain variables correlate and affect one another.

  • Regression modelling requires a significant amount of data to ensure accurate predictions which is why Google recommend at least 10,000 monthly active users.

See Will's blog for a more detailed explanation of regression analysis and statistical modelling techniques.

The report

Below is an example of the report you’ll see in Google Analytics for Firebase (Grow -> Predictions). If it’s your first time using the report you will see predictions for just churn/not-churn and spend/not-spend - in which case you can set up custom predictions based on custom events you’ve implemented.

Firebase Predictions interface

The report provides us with a visualisation of our configured predictions, how many users sit within the risk tolerance buckets (more on this another time) by dragging the slider, and then the option to target these users.

You’re also able to click into each prediction to get a view of the accuracy of the predictions made over the last 14 days so that you can be confident you’re targeting the correct users.

How can you use this insight?

It’s all well and good having the insight in the reporting interface but it needs to be translated into actions that drive improvement. Here are a few ways you can make use of this data in an effective way:

  • By predicting the likelihood of a user spending money in-app, users who are unlikely to spend can be pushed into Remote Configuration to monetize them using alternative methods such as showing them more in-app ads.

  • Optimise in-app promotions at a user level based on their likelihood of spending. For instance you could promote more expensive items to users likely to spend and less expensive items to users less likely to spend.

  • Use predictions to identify users that are likely to churn over the next 7 days and employ a more aggressive user retention strategy for these users. For instance grant these users a gift as incentive to continue using the app.

Put it into action

If used in the correct way Firebase Predictions could prove to be an extremely valuable tool to aid retention and re-engagement of users. Get in touch today to find out how we can help you leverage Firebase Predictions effectively!

Share this article