Retention Analysis
Retention analysis is a common data analysis method that measures how different batches of new users perform, most commonly through the retention rate. Improving retention metrics such as next-day, 3-day, and 7-day retention is crucial to a project's sustainable growth. With the Retention Analysis model, you can analyze in just a few clicks how users who completed a first event during a period come back, so you can keep the big picture under control and make better decisions.
Retention Analysis can answer questions such as:
-
What percentage of users who register each day are still active the next day?
-
What has the 30-day LTV of new users been over a past period?
-
On which day does the ROI of each ad channel break even?
Quickly configure the retention definition
Before you start the analysis, determine the entity. Different entities represent different retention perspectives. For example, you might want to view the retention of in-game characters, or analyze the next-day retention of accounts.
If you want to analyze whether new accounts are still active on the next day, day 3, and so on, configure the following:
- First event: the event that marks a new user, such as registration
- Return event: the event that marks an active user, such as login
By default, Retention Analysis analyzes by Day. You can also analyze by week or month, that is, whether the entities that completed the first event in a given week or month complete the return event in week or month N afterward, which gives you week N or month N retention. If you want to analyze daily retention metrics but view them summarized by week or month, configure it in the visualization settings.
After you configure the events, the Retention Analysis model automatically generates the number of retained users, retention rate, number of churned users, and churn rate for day, week, or month N. Retention and churn are defined as follows:
- Retention: an entity that completed the first event on a given day and completed the return event on day N after the first event date is a "retained user" for day N.
- Churn: an entity that completed the first event on a given day but didn't complete the return event on any day from day 1 to day N after the first event date is a "churned user" for day N.
You can also turn on Also Show for further analysis and calculate other metrics of the users who completed the return event, such as calculating LTV and ROI.
To get more precise retention data, you can also use Hold Property Constant and select an event property of the same type for the first event, the return event, and Also Show. When it's on, users must not only complete the return event, but the property values must also be identical. If a user triggers the first event multiple times with different property values on the initial date, the user belongs to all of the corresponding constant property groups.
For example, if you want to compare different in-game activities and count a user toward an activity's retained users only if the same user still takes part in the same activity the next day, you can specify the activity type as the property to hold constant. After you turn on Hold Property Constant, the Display constant properties as groups switch appears below it and is on by default. When it's on, the results are grouped by the values of the constant property, with one group per value and each group with its own overview value. When it's off, the results aren't split by the constant property, and only the overall retention of all entities is shown.
You can also filter data or compare with group by. Unlike Events Analysis, if the group-by item you use is an event property or a tag that uses Match by date, only the group value at the time the entity first triggered the first event is used to label the entity, which means an entity belongs to only one group within the time range.
For example, suppose you want to analyze the 7-day active retention of users from January 1~January 7, grouped by the event property VIP level. If a user completes the login event every day within the 7 days and has a different VIP level each time, the user is still placed only in the group of the VIP level at the time the user first triggered the login event on January 1.
Visualize results
Retention Analysis offers multiple visualization styles and shows a table by default.
- Table
Each cell shows the number of retained users or the retention rate on day N for a given initial date. Hover over a cell to see its description. Each cell is filled with a different shade of blue based on the retention or churn rate, so you can quickly spot abnormal dates.
The first row of the table shows the overview data for the time range. To learn how overview values are calculated, see Overview calculation logic. From the second row on, the table shows daily, weekly, or monthly retention or churn. If you've configured group-by items, click + to view the data of each group on that date in a dialog.
To compare the data of different group values more easily, you can also switch the table display format to Group as row. The table then shows the overview data of each group value by default, and you can click + to view the data of each date for that group in a dialog.
Note that cells involving Today are marked with an asterisk (*), which means the data is still changing. For example, when you view retention rates on March 2, the cells for the next day of March 3, day 3 of March 4, day 4 of March 5, and so on all have an asterisk.
You can use Exclude partial dates to keep changing data from skewing the results. This setting applies to the table display and to data export and download.
- Day N Retention
Day N Retention shows, as a line chart, how the day N retention or churn metrics change over time. Compared with the table, it makes it easier to find first event dates with anomalies.
In Day N Retention, the X-axis is the first event date, and each line represents the data of one group.
- Daily Retention
With Daily Retention, you can compare the retention rates of users who completed the first event on different dates on day 1, day 2, and so on afterward.
Unlike Day N Retention, the X-axis of Daily Retention is day N, and each line represents one first event date. If you've configured group-by items, each line represents the overview data of one group.