62. A platform recommends higher bids because its model predicts unusually high customer lifetime value. What would you verify?
MediumAnalytics & budget
Case
Show model answer
Check the target definition, training cohorts, observation window, and whether recent acquisition sources resemble the data used to train the model. A prediction can reflect immature outcomes, selection effects, or changes in pricing and retention.
Compare predictions with realized cohort results and uncertainty using the responsible analytics team. Limit commitments until the model is fit for this use, rather than treating a precise forecast as already earned customer profit.
What interviewers listen for
- Inspect the model target and cohort maturity
- Validate against realized outcomes
- Bound spending to evidence of fitness