3 Problems Solved by Building Activation as a Company-Wide Common Layer: Reliability, Cost, and Productivity

The Active User metric only shows the phenomenon, but doesn't explain 'why' it happened. To solve this, a User Activation perspective that analyzes a user's Active State and State Transition is needed.

To improve the fragmented analysis methods across teams and secure data reliability, the Activation layer was designed as a company-wide common data layer. This layer provides information such as user state, state transition, activity continuity, return interval, and churn period based on consistent criteria.

The overall structure follows a layered structure built up in the order Base → Fact → Activation. It was configured so that three models—FirstLast, Activation, and Activation Status—operate as a set based on a single Fact model.

To solve the reliability problem, instead of directly using event logs, the Fact model, which represents a meaningful unit of behavior, was fixed as the reference behavior. In addition, a naming convention of <fact_name>_activation_<time_grain> was introduced so that the model name alone clearly shows which behavior the calculation is based on.