Currently, the rising costs of acquiring customers for various products have highlighted the importance of increasing user stickiness. To evaluate the effectiveness of product updates or channel promotions, we often need to evaluate the subsequent behavior of users who entered the product during the same period or used a specific product feature during the same period, which is called cohort analysis.
At this time, we use the retention analysis of Sensors Data to conduct evaluations.

1. Step 1: Set the initial and subsequent behaviors for retention.

Assuming that we consider users who have completed the "registration" behavior as new customers and users who have performed the "payment order" behavior as retained users. In the retention analysis of Sensors Data, we set the initial behavior as "registration" and the subsequent behavior as "details of the paid order."

2. Step 2: Obtain the retention data of the cohort users.

Then, group the users based on the time of customer registration, either daily or monthly, to obtain the cohort group, and observe the retention/loss situation of the user within that group who made purchases. By comparing the retention performance of different cohort groups, we can infer whether the product/operation/channel actions during the initial behavior period have an impact on user retention performance.

It is worth mentioning that it is essential to set valuable user behaviors as subsequent actions for retention analysis in order to provide substantial guidance and recommendations for product optimization and improvement.