The success of a mobile app depends on the number of users who make transactions and other target actions. However, most of these users can leave if you don’t have a retention strategy. Churn can be driven by various factors – from users switching to a peer or alternative product or no longer needing the product’s features to being generally unhappy with their user experience or simply bored with the app.
Handling churn and user retention is a priority for a product manager or product owner. Here’s why:
You can’t bring churn to zero as some users quit apps anyway, but you can reduce the attrition rate by predicting the probability of churn in advance and taking preventive action.
In this article, you will learn why you need to predict the churn rate using MyTracker's churn prediction model. You will also learn how to use churn predictions to identify the most effective advertising channels and set up retargeting campaigns to retain users with a high churn probability.
Churn prediction helps you analyze audience engagement proactively and prevent app users from leaving.
If you can predict the attrition of users important for your product segment, then you can take action to retain them.
For example, you can send them promotional push notifications, add more relevant content, or launch personalized offers with MyTracker Personalize.
Churn prediction helps you identify advertising channels that draw a more loyal audience. It enables you to replace creatives or reallocate resources to more effective channels in time to save the budget.
For example, you can use it in the early stages of an ad campaign to see that certain ads or partners attract audiences with a high probability of churning. With this data, you can make timely changes to your ad channels or creatives to secure a more loyal audience.
With MyTracker, you can use the Churn Prediction metric to forecast your churn. It leverages machine learning models to identify users with a high likelihood of churning in 7, 14, 21 and 28 days after the app installation.
This metric can be combined with other MyTracker tools as follows:
Now let’s take a closer look at how to predict churn in MyTracker and use Churn Prediction in combination with various tools.
Churn Prediction can be used in combination with other MyTracker tools for a variety of purposes. Let’s look at each of these scenarios separately:
Suppose we have launched an ad campaign to attract new users and want to find out which advertising partner draws more engaged users.
To do this, we will use the Report Builder tool to create a new report for the required period, with a breakdown by Partner and New Devices.
Now let’s add Churn Prediction to find out the probability of new users churning from these partners in 7, 14, 21, and 28 days after the app installation.
MyTracker builds the prediction using machine learning models. They analyze the collected app data (installations, launches, payments, sessions) and can predict the probability of a user churning as early as the next day after installing the app.
To add a metric, follow Report Builder → Select from list → Devices → Churn Rate → Churn Prediction LT.
If we add Churn Prediction LT 7d, we will see how many users might stop showing any activity in 7 days after installing the app.
Let’s recalculate the report and look at the results:
The above report shows that users who came from Partner 2 campaigns are more loyal than those from Partner 1.
Based on Churn Prediction LT 7d, 87.84% of users referred from Partner 1 will not have any activity between February 14, 2022 and March 22, 2022 (i.e. during the 30-day inactivity window).
The predicted churn does not necessarily mean that the app will be uninstalled, but it shows that some users will be inactive during the chosen inactivity window after the selected date.
The inactivity window is a period of time following the last app launch or website visit after which MyTracker can reattribute the user by identifying the ad that has brought them back to your app.
Don’t forget that you should never draw conclusions about your audience based on just one metric. For example, users referred from Partner 1 may turn out to be more valuable and ultimately have a higher LTV than those coming from Partner 2.
Audience research needs a holistic approach. We regularly publish articles providing audience insights on our blog. Along with this article, we recommend reading about financial metrics and LTV prediction. Altogether, this knowledge will help you better understand your audience and make product development decisions based on a comprehensive data set.
Now let’s look at how to build segments with a high churn probability and to export them.
Using the Segments tool, you can identify paying users that are very likely to leave the app, and launch targeted ads for them or approach them with personalized offers to reduce the churn.
Let’s go to the Segments tool and create a new segment.
On the Add Segment page, we will choose an app for building a segment and select Devices in the Audience Type field.
Now we need to configure additional parameters for our segment, including the period to provide data for the segment. By default, all segments have a sliding window of the last 30 days.
To identify the segment of paying devices, choose Add Parameter and select Revenue as an additional metric. As its value, let’s set >USD 1 to filter out any devices that have brought in less than USD 1 in revenue over the specified period.
We will add the Churn Prediction parameter in the same way.
As its value, we can indicate the churn probability. For example, let’s indicate an interval from 80% to 90% to arrive at a segment of devices that are 80–90% likely to leave the app 7 days after the installation.
If needed, we can add additional parameters for segmentation, such as Partner or Campaign, to build a narrower segment.
Once all the parameters we need have been added, we click Add and wait for MyTracker to perform the calculations.
From the results, we see that our app currently has 21,490 devices that have brought in more than USD 1 in revenue but have a probability of 80–90% that they may leave the app in 7 days.
To do this, we will click Export on the page of our newly created segment, and configure the export settings.
It is only account owners and users authorized by account owners that can export segments.
There is also the Audiences tool, which can help set up an automated distribution of the segment to advertising platforms for retargeting. Let’s look at how it works.
Using the Audiences tool, we can send the segment we have just created to an advertising platform.
To do this, let’s go to Audiences and add a new audience.
In the Segment field, we need to select the segment we have built to send it to an advertising platform (in our case, it’s users that are 80–90% likely to leave the app in 7 days).
After the audience is set up, we will add integration with a partner to start automatically sending the audience. Steps to set up integration may vary depending on the partner, but generally look as follows: click Add → in the pop-up form for adding a partner, again click Add and choose an advertising partner from the list. That’s it: sending is set up, and the audience is updated once a day.
For detailed information on integration, visit our documentation.
MyTracker automatically updates the list of devices in a connected advertiser account by using backend infrastructure to interact with partners.
All you have to do is enable audience data transfer and configure partner integration. After that, MyTracker will regularly update this data in the partner’s advertiser account and send the list of device IDs with a 80–90% churn probability.
Depending on the configurations you choose for your advertising platform, you can approach this segment with ads that will make them interested and help engage with the product.
For now, MyTracker’s Audiences has integration with myTarget, VK Ads, Yandex.Direct and ironSource. We will be adding new partners to the list.
An audience that is likely to churn can be retained through personalization, i.e. exposure to personalized offers that can once again make users interested in your product.
MyTracker supports the exporting of segments with a high churn probability and sending them to the Personalize recommendation engine, which will provide users with customized offers, ranked lists of products and prices in real time. For example, with Personalize, you can rank in-game items from the store according to user preferences or offer discounts for options they are likely to be most interested in.
Here are the steps for managing segments with a high churn probability using Personalize:
Check out how Hustle Castle managed to increase ARPU in one player segment by 23% with MyTracker Personalize.
For an app, it always makes more sense to retain and bring back users that already know the product than to attract new audiences. This makes it important for marketers and app owners to fight attrition by monitoring the audience, improving functionality, and running campaigns to retain users.
With MyTracker, you can get churn analytics data and take steps even before they decide to leave. To do this, we have the Churn Prediction metric, which makes it possible to identify users with a high likelihood of churning in 7, 14, 21 and 28 days after the app installation.
This metric can be used to identify the most effective advertising channels and create segments with signs of churn to later send them to advertising platforms and run retargeting campaigns.