About the Client

Our customer was a game development company. The company wanted to leverage the players’ in-game behavior to proactively predict player churn.

Goal

Implement a solution that could periodically evaluate the player base and single out the profiles that were likely to quit playing the game within a specific time frame.

Work Description

1. Define the features that were significant for predicting churn
2. Segment the existing audience according to those features
3. Train the model to calculate the probability of quitting the game within the following 7 days for a given player profile
4. Implement a periodical job that re-evaluated player profiles and saved potential leavers to a database table.

Work Stages

1. Get access to an anonymous dataset of player profiles and their in-game activities
2. Mark profiles of the players who quit playing the game
3. Look for correlations between the profiles and the fact that they quit playing
4. Implement and train the predictive model
5. Validate the model
6. Implement the model as a Hadoop job

Results

Geomotiv’s solution was able to accumulate the required in-game statistics and identify individual player churn risk in 75% of the cases. This enabled the client to proactively address high churn risk players thus reducing the attrition rate by 45%.