The Science Behind App Uninstalls Prediction – Part 2/2

App uninstall part 2
So fellow app marketers and product developers, as promised, we build on from where we left in the previous blog. We talked about compelling app uninstalls stats and also drew insights on an app marketer’s woes. If we do not have means to retarget and bring them back, we lose the customers forever once they uninstall the app. That’s where we touched upon the point that you and I need a magic wand to prevent these customers from uninstalling in the first place.

Machine learning predictions are the magic we are all looking for. Sounds like a challenging implementation task, don’t you think? Well, it is and it is not. At Tatvic, we call it the PredictN Model. Our data science team and implementation team have built this product which accurately predicts the cohort of users who are likely to uninstall. Let’s jump into the how’s and what’s of our PredictN Model.PredictN Model Features

PredictN Model Features

  1. Tatvic’s Uninstall Library


    machine learningThis library captures all the critical mobile device specific attributes and then some ranging from battery usage, data connections to advert ids.

    Key features of library –

    1. Light weight of 24.2 KB size and 220 bytes of static config file
    2. Easy to integrate into native Android app
    3. Data can be sent to App Analytics tools/ server
  • App Analytics Tools

    App Analytics tools like Google Analytics. Firebase and Localytics, allow us to extract behavior and transactional properties of users. This minute user-level data (for e.g. average time on app, days since last visit, number of sessions, number of products checked, acquisition source and campaign, etc) are being tracked. These data points will help us understand App usage pattern to be used in the model.

Model Process

predictive analytics for apps,


Our model looks at historical user behavior data, history of uninstalled user data and collected data using uninstall library to predict the probability of likely to churn users.

  • Connect your App Analytics Tool with our PredictN model
  • Integrate our Uninstall library to start collection of advanced attributes
  • Build & Run PredictN Uninstall model
  • Generate list of high probable churning users
  • Target the probable churning users using different marketing channels

Our PredictN model gets trained based on historical app usage data of users and uninstall library parameters. As more data is collected into app analytics/server, the PredictN model prediction accuracy gets improved.

prediction model


Once we build & run the predictN model, it will allow us to generate a list of high probable churning users using advanced segment. Those users’ advertising ids/PII which will help us to retarget using different marketing channels/activities like:

  • Email Campaigns
  • Push Notifications
  • Display Channels

We can also reach out to those users to obtain additional feedback which will help us to understand reasons for uninstalling your app. Prediction of uninstalling users is a very important step to improve user engagement and app retention rate significantly.

To know more how we developed PredictN for a marketplace app and more actionable insights, we invite you to attend our upcoming webinar.

The following two tabs change content below.
Jigar Navadiya

Jigar Navadiya

Jigar is a computer science graduate and leads the technical team at Tatvic. His interest is in solving complex data collection problems using tools and technologies. He is always keen on exploring new analytics tools, cloud platforms and technologies like Python, scripting languages. When not working, he spends time on reading blogs, watching resourceful videos on YouTube.
, , , , , , , ,
Previous Post
The Science Behind App Uninstalls Prediction – Part 1/2
Next Post
Strengthen Your Personalization Strategy with Google Optimize 360

Leave a Reply

Your email address will not be published.

Fill out this field
Fill out this field
Please enter a valid email address.
You need to agree with the terms to proceed