Best practices
In practice, you can combine various models to analyze from different angles.
Next-day churn analysis
Next-day retention is a metric that every project cares about. Only by finding the real reasons for user churn can you make targeted adjustments and improve the next-day retention rate.
First, use the Retention Analysis model with registration as the first event and login as the return event to observe next-day retention. After you find the dates when the data drops, you can save the users who registered on that day, and the users who registered on that day but didn't log in the next day, as two separate result cohorts.
Next, with Funnel Analysis, you can see which levels these users stopped at on the day they registered. Group by whether users are churned to compare the data of churned users and retained users.
You can also use Flows Analysis to view the behavior paths of churned users starting from registration, find the last core gameplay of churned users, and analyze whether that gameplay frustrated users so that they stopped being active, such as losing an arena battle or being disappointed with Gacha results.
Payment behavior analysis
Payment behavior is a focus of data analysis, and you can analyze it from multiple angles.
The most basic is Events Analysis. With the total count, number of users, and total amount of payments as metrics, you can see how their data changes over a period of time.
Going further, in Distribution Analysis you can select a time range and divide users into ranges by their total payment amount during that period, to view the number and percentage of users in each range.
Interval Analysis helps you analyze how long it takes users to go from registration to their first payment. Its box plot shows the median and the upper and lower quartiles at a glance, giving you a basis for adjusting the first payment point.
User profile analysis
In Composition Analysis, you can cross-combine dimensions, see the number of users for each combination, and explore whether two dimensions are correlated. For example, by grouping by Source Channel and Cum. Revenue, you can find out whether the share of big spenders in some channels is clearly higher than in others.
If you find abnormal user counts for some dimension combinations, you can click through to the User List, or go further into the Behavior Sequence to observe all the characteristics of individual users. You can also divide the users in the project into different groups by user property and compare their data.
If the current user properties don't meet your needs, you can also segment users with user tags. Like user properties, user tags can be used for filtering and as group-by items.
When your project has rich characteristics, you can place multiple Composition Analysis reports on one dashboard to build a complete user profile for the project.