hard20 pts · first answer countschurn-preventionchurn-predictionexperimentation

The model looks worse after we act

After CSMs start calling every account the churn model flags, the flagged accounts churn much less than predicted. A leader says the model is wrong. What is the better explanation?

A.The model must have been trained on the wrong customersB.Flagged accounts were never at risk in the first placeC.Churn models stop working once CSMs see the scoresD.The calls changed the outcome, so keep a holdout to measure

Discussion 0

Join free to post in the discussion.

The model looks worse after we act

Discussion 0

Related questions