All worked cases
Validate grocery segments in Hyderabad
Hyderabad, Telangana · Fictional practice case
The decision
A grocer clusters households using recency, visits and basket value. The largest cluster also lives closest to stores. An initial regression shows a strong association between visits and basket spend.
How can the team turn those findings into a decision?
Worked solution
Reveal the visual and solution
- Standardize features and hold out a later period to check whether clusters are stable. Avoid variables observed after the offer.
- Check whether distance and prior purchase explain the regression association. Treat the coefficient as descriptive until the identification design supports a causal effect.
- Randomly assign eligible households within each segment to offer or control. Compare incremental contribution and use segments only if they change offer selection.