USE CASES
Market research
Research needs comparability. Only one variable may change between samples, and if the exit region drifts on its own, the comparison is void.
Make region the controlled variable
One collection script across ten countries should vary only in exit country. Everything else gets pinned: the same headers, the same time window, the same parsing rules, the same currency baseline.
Allocate samples by variance
Do not split evenly. Large markets have wider price distributions and need more samples; small ones converge in a few dozen. Run a small round, measure variance, then spend accordingly.
- Start with 50 records per country and compute price variance
- Add samples where variance is high, trim to a floor where it is low
- Keep raw responses so a method change can be recomputed instead of re-collected