Crypta specialists developed a model built on the AppMetrica infrastructure that predicts churn in mobile apps. Using the model, a service provider marketplace was able to more accurately calculate the number and nominal value of promo codes they should issue to bring back customers.
Here, we’ll use this example to demonstrate how the Crypta model helps apps increase LTV.
If you want to grow your mobile app, analyze user behavior and acquire top LTV traffic, it’s crucial to rely on accurate data for all your product and marketing decisions.
AppMetrica is an all-in-one mobile app analytics solution that allows you to take your app to the next level at an affordable price.
The predictive churn model is an ML model which predicts the likelihood that a user will stop using a service. Churn prediction can help apps that send push notifications to users or offer them promo codes.
This approach may lead to more orders, but it reduces profits. That’s why some businesses build predictive models to avoid giving out promo codes to users who are already likely to stay on the app.
In September 2022, Crypta specialists developed a model on the AppMetrica infrastructure that predicts user churn in mobile apps. They asked the service provider marketplace to compare models and decide which was more effective.
It was important for the team to understand whether the model would become a separate, boxed solution or an add-on to an already existing algorithm.
Crypta experts assumed that the model would be more effective if it was based on more extensive data. To test this hypothesis, they launched a two-week A/B test where they compared the following:
Sample sizes were all the same: 30% of the entire test audience. During the two weeks of the experiment, the only users considered were active app users who had ordered cleaning or water delivery at least once and who did not have an active current order.
There conclusions were reached regarding how more accurate forecasting affected business profits:
The predictive churn model is currently in closed beta testing. But you can already implement it as a test by filling out the AppMetrica feedback form. Get access.