Predicting Your Uninstall Apps: data and a little bit of science
Mobile App marketers spend a significant amount of their resources to driving installs. However they find it increasingly difficult to improve or even maintain their user retention rates because of uninstall rate. Marketers need to look beyond just installs to grow their App usage.
App uninstall prediction
At Vizury, we have developed a prediction engine that will utilize historical data to determine exactly which user is going to uninstall your app. This is our latest addition into the Engage Commerce platform and has delivered some terrific results.
Initial results for a re-engagement use case
We have tested our model accuracy across advertisers and have seen more than 75% accuracy in uninstall prediction. Which means that, 75% of users whom we predicted have actually ended up uninstalling within the next 7 days.
Vizury’s Engage Commerce platform allows you to reach these users through Push and Display channels with relevant messaging, offers, additional benefits, interested products etc. to ensure they start engaging with the App and thereby increasing retention rate. So, how does all of this work? Let’s deep-dive and find out.
User Data collection:
We integrate our proprietary SDK in the app which collects and sends data at each interaction/event of the user with the app. We then aggregate this event level data at a user level and extract various user properties.
- Device: screen size , memory , processor
- OS and OS version
- ISPs used WiFi/3G
- Time of interactions
- Volume of interactions: Number of interactions per day, per week and per month
- Quality of interactions: This is an app vertical dependent property. For example, in an ecommerce app, a user-visit to the Product page holds much more value than a visit to the Home page.
- Days of activity
- Number of events and many more
Our algorithms look at the history of users who have uninstalled your App in the past and then map these attributes with your existing App users. This helps us predict the probability of uninstall for existing users.
Use Cases for Marketers:
These parameters are continuously monitored at a user level. As more data gets ingested into the platform, the prediction model gets trained to predict with higher accuracy.
- Identify users who are likely to uninstall and target such users through Push and Display channels
- Get reasons and patterns of users who have uninstalled or are likely to uninstall. At Device , Location, OS, frequency and engagement quality levels. Take action based on detailed information to bring down uninstalls
- Reach out to users who have uninstalled the app to obtain feedback
You might have a million installs for your app and it would mean nothing to your brand if your retention rate is low. Predicting uninstalls is the crucial first step towards effectively re-engaging with users and eventually prompt the elusive in-app purchase.