LTV prediction for a mobile game
Budget: $750 – $1,500 USD
I have data for an Android mobile app (casual game) at the device level for millions of users. It includes install date, days of activity, device model, and traffic source (channel and campaign). If you need any specific data that could be useful, I will be happy to provide it. I can also provide info on when the app was uninstalled. I'm mainly interested in retention prediction, so no revenue data will be provided.
I expect a script (in Python or R) that can compute and generate retention predictions for each user for day 7, 14, 30, 60, ..., up to 720 days since the install date. All calculations must not require extra computing power and should be possible to do on an average laptop, since the predictions must be generated daily and I cannot afford an expensive server.
I have tried BTYD models for this data; however, they mostly underestimated longer-tail retention, and predictions for older users were mostly pessimistic. I also expect model accuracy estimates based on historical data (using 7, 30, 60, 90, ... days of factual data).
Please include in your proposal which statistical/ML model you would use and if you have experience with retention prediction.
I expect a script (in Python or R) that can compute and generate retention predictions for each user for day 7, 14, 30, 60, ..., up to 720 days since the install date. All calculations must not require extra computing power and should be possible to do on an average laptop, since the predictions must be generated daily and I cannot afford an expensive server.
I have tried BTYD models for this data; however, they mostly underestimated longer-tail retention, and predictions for older users were mostly pessimistic. I also expect model accuracy estimates based on historical data (using 7, 30, 60, 90, ... days of factual data).
Please include in your proposal which statistical/ML model you would use and if you have experience with retention prediction.