Compare two stores’ profits rigorously

Two snack shops are positioned at a school gate and open concurrently. You have a 14-day window to decide which one will earn higher profit over the coming quarter. Create a measurement and analysis plan that produces a defensible recommendation under real-world limitations. Specify:

1. The precise primary metric (for example, profit per passerby per open hour) and why it aligns with the decision; state which costs and revenues are included and how you will normalize for foot traffic and operating hours.
2. The minimum data to collect: hourly foot traffic counts, transaction counts and average order value (AOV), item-level margins, labor hours, rent and utilities allocation, weather, school calendar, and competitor promotions or price changes.
3. Your identification strategy—either an experiment (such as randomized flyer distribution or alternating queueing) or a quasi-experiment (such as hour-level difference-in-differences with fixed effects and weather controls)—that stays valid despite weekday/weekend and exam-week spikes, Store B extending its hours, Store A offering weekend discounts, and one shop changing its price on day 9.
4. The model and uncertainty: write the difference-in-differences regression you would fit (defining outcome, treatment, fixed effects, and controls), how you would compute a 95% confidence interval for the profit difference, and a minimal power check indicating that 14 days is enough (order-of-magnitude inputs are acceptable).
5. Guardrails for detecting stockouts or cannibalization, plus an explicit decision rule (for example, recommend Shop X if the estimated profit difference exceeds $Y per day and the confidence interval lower bound exceeds $Z).

Overview: This question tests a data scientist's abilities in experimental design, causal inference (including difference-in-differences), profitability measurement and metric construction, uncertainty quantification (confidence intervals and power checks), and operational guardrails for real-world data collection.

This question is drawn from a data scientist interview experience.