Composition Analysis
Composition Analysis is a model designed specifically for statistical analysis of users and user properties. You can compare the data performance of different user groups, or view the number of users for cross-combinations of dimensions, to quickly profile user groups and understand the characteristics of specific user groups from multiple angles, providing a basis for refined operations.
With Composition Analysis, you can answer questions such as:
- How many users are there in each province in the project, and which province has the most?
- Is there a significant difference in average cumulative revenue per user among different cohorts?
1. Only tags or cohorts of entities whose source is the AE user ID or a user property can be used in Composition Analysis
2. When you select a user tag, the latest version is used for calculation and cannot be changed
3. User properties have no concept of time zone, and you cannot select a display time zone
Quickly configure analysis metrics
As in Events Analysis, you can flexibly configure analysis metrics in Composition Analysis.
By default, the analysis metric is the number of users, that is, the deduplicated count of AE user IDs (#user_id) that meet the conditions. You can also select user properties or user tags for statistics.
You can also filter data to exclude the data of certain users. Composition Analysis supports filtering based on user properties, user tags, or cohorts.
For example, to find users who belong to multiple cohorts at the same time, you can configure this in the global filter, then drill down to the User list to view details or save these users as a result cohort to get the intersection of users across the cohorts.
Group by property
When grouping by property, you can select up to two group-by items
- Select a single group-by item
When you select a single group-by item, you can directly view the number of users or other analysis metrics for different group values.
For example, if you select registration time as the group-by item and the number of users as the analysis metric, you can view the number of users registered each day/week/month in the Bar view, or view the proportions in the Pie view.
- Select two group-by items
When you select two group-by items, the Bar view is shown as a stacked bar chart, and you can select one of the group-by items as the X-axis.
In the table, the two group-by items are used as the row headers and column headers respectively, and each cell shows the analysis metric for the cross-combination of dimensions. This makes it easy to spot correlations between group-by items or compare the data results of different group values.
Compare the data performance of different user groups
Unlike other analysis models, in addition to group-by items, Composition Analysis also lets you compare the data performance of different user groups, such as the number of users, or the per-user average or variance of cumulative revenue. You can compare up to 10 user groups at the same time.
In addition to selecting existing cohorts, you can filter by user properties or user tags to define a custom user group, or select All Users as the control group.
In the chart display area on the right, you can visually compare the analysis metrics of different user groups in the Bar view, or view the proportions of different user groups in the Pie view.