Predictions Update - More Magic Metrics


We’re always striving to improve Localytics with updated features and functionality, but it’s especially exciting when we can deliver on a customer request in just a few weeks. That’s why we’re so excited to announce significant enhancements to the Related Behaviors section of Predictions that will deliver a new level of insight into exactly why your users churn or convert.

If you recall, we introduced Predictions back in October and by December, we were already receiving great feedback from our launch customers. As we looked through the requests, a central theme emerged: knowing which users were likely to churn or convert was a game changer for our customers, but now they wanted a better understanding of why. Our team immediately got to work and delivered this update to Prediction’s Related Behaviors in under a month.


What are Related Behaviors?

Related Behaviors are the ‘magic metrics’ that serve as indicators of future behavior like churn or conversion. Here's a great talk from Facebook’s former Head of Growth discussing how important it was for them to discover their magic metric: Getting any given individual user to add 7+ friends in their first 10 days. Facebook built their entire acquisition strategy around this one metric and credit it with their ability to get over 1 billion users.


What’s new in Related Behaviors?

Today’s update introduces two new features:


1.  Attribute-Level Information


Predictions has always shown you which events are associated with a user abandoning your app (churning) or performing an action that is valuable to you (converting), but now you can go deeper and see the impact of individual event attributes. As a reminder, events are the actions that a user has taken inside the app (read an article, clicked on a specific button, etc.). Attributes are the details for a specific event (the title of the article that was read, the type of button that was clicked, etc.).

With the inclusion of attribute information, you can obtain a more detailed understanding of a user’s behavior inside your app. For example, you can discover if articles with videos reduce churn more than articles without videos or find out if a specific product plays a key role in driving conversion.


2.  Enhanced Filters


With so much new information being included in Related Behaviors, we have also added enhanced search, sort and filter capabilities so you can always find the data you want. Of course, Predictions still defaults to highlighting the most impactful takeaways, but these new organizational tools make it easy for you to dive in and take advantage of every possible insight.


How can I use this update to improve my marketing efforts?

Since launching Predictions, Related Behaviors have shown what impact an event has on a user’s likelihood of churning or converting. That means insights like:


This example features a news app and it shows that users who received 4 or more push messages during their initial 5 days of usage were 15% less likely to churn. Now, it’s probably not too surprising that users who received more push messages are less likely to churn (although of course there’s going to be a limit to how many push messages you can send before a user starts getting annoyed). What’s important to notice is that this insight has given you a target for your onboarding efforts. You now know you should build an onboarding campaign that is designed to send 4 or more push messages within the first 5 days of a user downloading the app. Without this magic metric, you might send a series of push messages to new users, but would you know that there’s a significant difference between sending 3 and 4 messages?

As valuable as this insight is, it’s only scratching the surface. With today’s update, we can go much deeper and explore a whole new dimension of magic metrics. Here’s an example of what insights can look like in the updated version of Predictions:


By adding event attributes to Predictions, we’ve uncovered a whole new level of understanding. We are looking at the same event (receiving push messages), but we’ve added the event attribute “Campaign Name” into the calculation. The results are amazing. We can see that users who receive 3 or more push messages from the “High Churn - Retention” campaign are an astounding 61% less likely to churn. By including attributes in the analysis, we’ve identified a far more significant opportunity to reduce the likelihood of churn.


It’s always gratifying to work on projects that are the result of a close collaboration between Localytics and its customers. There’s more to come so please keep sending us feedback and click here to learn more about Predictions.


I want to learn more about Predictions