Here we answer some of the most common questions we hear about the wine club attrition model from our customers.
How do you know the predictions are accurate?
We created and tested the model with real winery data. We used historical data and known outcomes to run the model and test the predictions against what actually happened. Then we tweaked the model and did it all over again. We repeated the process until we achieved our desired outcome, 90% accuracy that what we predicted occurred.
How did you determine the predictions?
A lot of variables play a role in indicating if/when a member will cancel. Emetry used a rainforest ensemble model capable of responding accurately to all the data variables.
Can it predict everyone who will cancel?
No. There are factors in life that data can't account for, like unexpected life events, natural disasters, etc.
I have 4,000 members, so why are only 3,500 in the attrition exports?
Members who don't have enough data for the predictive model to access accurately are excluded, and their predictive attrition fields are left blank.
Is the model custom to my winery?
Yes, the algorithm adapts to each winery and returns predictions custom to each brand.
How often do the member attrition predictions update?
The predictions refresh on the 2nd of each month.