The main idea of this project, is to predict the ARPU (Average Revenue Per User) 11 month after subscription of a cohort with a monthly subscription, using minimum number of delays (a delay is a month of data of a cohort, i.e.:
delay 0 --> first month of data for cohort A,
delay 1 --> second month after subscription ...), now the idea is to predict the ARPU using maximum 3 months of data, in a way everytime a cohort has another month of data we predict the ARPU of 12th month :
- cohort A subscribed in 01/2021 --> Predict ARPU of 12/2021 using data of 01/2021
- in 02/2021 --> Predict ARPU of 12/2021 using data of 01/2021 and 02/2021
- in 03/2021 --> Predict ARPU of 12/2021 using data of 01/2021 and 02/2021 and 03/2021
(more data = better accuracy, but for business requirements we cant wait more than 3 month of data for a cohort to get good predictions) I'm wondering how to make this possible, considering that data is stored in Snowflake.