Evaluate smart cart idea with hypotheses and experiment

Instacart partners with a local grocery store to introduce a “smart cart” in the physical store.

The smart cart UI lets shoppers:

1. search/browse items that exist in the current store and see in-store prices, and
2. also see items and prices from other nearby stores available on the Instacart app.

You are a Senior Data Scientist asked to evaluate whether this is a good product idea.

1. What are the key hypotheses (primary and secondary) for how this smart cart could impact the business? Consider potential positive and negative effects.
2. What metrics would you track? Specify at least:
   • a primary success metric,
   • 2–4 supporting diagnostic metrics,
   • 1–3 guardrail metrics. Make sure you address tradeoffs like cross-store switching/cannibalization and user experience.
3. How would you design an experiment to measure causal impact? Include:
   • unit of randomization (user, trip, cart, store, day, etc.),
   • how you would handle interference/spillovers (e.g., shoppers seeing others use the cart),
   • required logging/instrumentation,
   • rough power/MDE considerations and how you’d estimate them.
4. If you cannot run a clean randomized experiment, propose a credible quasi-experimental alternative (e.g., diff-in-diff) and key assumptions to validate.