Player Churn Prediction and Retention
Publication date: January 8, 2021 (Diagram history)
This architecture uses Amazon SageMaker AI to build, train, and deploy a churn prediction model. You can use the model for batch or real-time inference to predict which players are likely to leave and trigger targeted retention actions.
Player Churn Prediction and Retention diagram
The following steps describe the architecture:
-
The mobile client sends player events through a game analytics pipeline to an Amazon Simple Storage Service bucket.
-
AWS Glue transforms player data and marks players as ACTIVE, INACTIVE, or CHURN for data scientists.
-
Data scientists prepare player data, test algorithms, and build a model with a subset of the dataset in an SageMaker AI notebook.
-
Data scientists launch model training on 80% of the dataset and store the trained model in Amazon S3. The remaining 20% is used to test model accuracy.
-
The mobile client asks which in-game player retention action should be triggered, such as a promotional offer or bonus items.
The architecture supports two inference paths:
-
Real-time inference: An AWS Lambda function checks the player profile and evaluates churn probability from the SageMaker AI endpoint to choose the most appropriate retention action.
-
Batch inference: A periodic AWS Glue job triggers a batch transformation job on the entire dataset and copies the results to a player profile table in Amazon DynamoDB. The Lambda function checks the player profile churn prediction and chooses the retention action.
Further reading
For additional information, see the following resources:
Diagram history
To be notified about updates to this reference architecture diagram, subscribe to the RSS feed.
| Change | Description | Date |
|---|---|---|
Initial publication | Reference architecture diagram first published. | January 8, 2021 |
Note
To subscribe to RSS updates, you must have an RSS plugin enabled for the browser you are using.
