Using our predictions functionality, you can predict what your users are likely to do and serve up marketing and content to match. This is how truly personalized engagement happens.
Conversion and Churn
There are two key marketing events you can predict:
- Conversion
- Churn
Conversion predictions forecast user action. They tell you when users are most likely to perform key conversion events such as view products, make referrals, or finish a checkout.
With Conversions in Predictions, you can predict when users are likely to convert and send them promotional messaging that pushes them to take action.
Churn predictions forecast user inaction. It's typically an indication of users likely to disengage with you. You want to know churn predictions so you can anticipate and prevent user disengagement by sending marketing messaging that reignites their interest and keeps them coming back.
Using the Predictions Dashboard
Navigate to Predictions by selecting it from the sidebar (under Analytics). On this page, you'll see a list of all of your predictions in Localytics.
Click into any of your current predictions to view the status of them. We'll talk about assessing their performance in further in this guide.
Predictions Use Cases
The purpose of Predictions is to anticipate and influence user engagement. For example, you can use our prediction analytics to:
- Identify only those least likely to perform a conversion event so you can send them a discount to complete the sale
- Discover key behaviors that indicate retention and loyalty such as the number of referrals a user makes
- Identify users at high risk of churn so you can send them offers and messaging to rekindle interest
Creating a New Prediction
Localytics makes creating predictions to guide your marketing efforts easy. To create a new Prediction, hit the green plus button at the top of the Predictions dashboard.
You'll be prompted to build your prediction by selecting what you'd like to predict from a series of drop-down menus (for churn) or by selecting the conversion event.
Our predictions engine is divided into two pathways: conversion and churn. First choose which pathway you want to take, depending on whether your goal is to retain users in danger of leaving or convert those ready to take further action.
For conversion predictions, you will also be given a series of drop-down menus. In this case, you can choose from a series of events (actions) you want the user to take.
If you choose the churn path, you will be given a set of drop-down menus that allow you to refine your predictions by events or actions users are not likely to take. You can also set a time frame from 7 to 90 days.
For both churns and conversions, the ability to choose multiple events is what lets you refine predictions in order to tailor your marketing to the expected user behaviors.
When you're finished building your prediction, don't forget to Save and Close.
How to Use Prediction Reports
Use your Localytics predictive analytics report to shape more effective marketing strategies and campaign messaging and offers.
For example, for users highly likely to not view a product within 30 days, send them a time-sensitive offer to entice interest before then.
For users who are multiple purchasers, reward their loyalty with an offer to purchase before the general market.
The predictive report can also help you identify the ideal time to send your marketing. Certain behavior attributes can indicate when certain users are more likely to open push messaging and even on which device types.
To access an individual predictions report, select the report from the Predictions dashboard.
You'll see all of the data Localytics has on your current prediction including:
- Prediction Info
- Type of prediction, criteria definition, next update, and baseline conversion rate
- Likelihood of Conversion or Churn
- Divided into low, medium, and high and broken out by users and % of active users
- Related Behaviors
- Compare the conversion or churn event by different related occurrences
- Related User Attributes
- The user attributes associated with the conversion/churn you're measuring
Let's dive into the last three of these.
Likelihoods
The results of each Prediction will group users into buckets of High, Medium, and Low Likelihood for churn or conversion based on the Events you have specified.
You should use these buckets as Audience criteria for targeted messaging. For example, you could send users at high likelihood of churn a 20% off coupon. At the same time, you could send your most loyal users a discount referral code.
To follow-up with any of these groups, click the ellipsis at the end of the line to access the Actions menu for that group. Target them with either a Push Campaign or an In-App Campaign.
Related Behaviors
Below your the list of Likelihoods, you'll see Related Behaviors.
While it’s important to know which users fall into which buckets, it’s also important to understand why. Related Behaviors are the usage patterns that serve as lead indicators of future behavior.
These behaviors describe the "aha!" moments in your app when users discover core product value, and we observe the greatest shift in retention or conversion as a result of that discovery.
We comb through all of the event and attribute data sent from your app to find these inflection points over all time, as well as within certain key time bins like the first 1, 3, 7, 14, and 28 days of your users' lifetime. By default, Related Behaviors are both ranked by z-score and grouped by event.
Filter Related Behaviors by different Time Bins, Events, and Attributes.
Related User Attributes
Finally, the last piece of data we surface on your Predictions Report is Related User Attributes.
Related User Attributes are the attributes most related to the predictive target behavior. For each attribute, you will see the proportion of active users observed to have each attribute. You will also see the relative difference in churn/conversion between users observed to have that attribute vs. all active users.
Use top Related User Attributes to inform additional layers of segmentation on top of the High, Medium, and Low likelihood segments to create even more targeted and relevant predictive Audiences and messages.