Plan DS approach for biker delivery project

You are a Data Scientist supporting a “biker” (delivery rider) product/project for a food-delivery platform.

An interviewer gives only a short description of the project and asks you to explain how you would approach it as a DS.

Prompt

1. Goal framing: What clarifying questions would you ask, and how would you translate the business goal into measurable objectives?
2. Business process & data: Describe the end-to-end biker workflow (dispatch/assignment → accept/decline → travel to restaurant → pickup → travel to customer → dropoff → post-delivery). For each step, list what data/events you would expect to capture and how they map to metrics.
3. Metrics: Propose:

• Primary success metric(s)
• Diagnostic metrics (to explain why the primary metric moved)
• Guardrail metrics (to prevent harm)

4. Causal inference / evaluation: If you cannot run a perfect randomized experiment immediately, how would you estimate impact and reduce bias (confounding, seasonality, selection effects)?
5. Misclassification + precision/recall tradeoffs: In this context (e.g., detecting problematic deliveries/riders/orders or triggering interventions), explain where false positives vs false negatives matter, and how you’d set thresholds.
6. A/B testing design: If city-level clustering is one option, what other randomization units could you use (order, rider, customer, restaurant, zone, time-based switchback, etc.)? For each, discuss pros/cons, interference/spillover risks, and when you would choose it.