Prioritize a new warehouse proposal with data
Overview
This question evaluates demand forecasting, network optimization, five-year NPV financial modeling, experiment and quasi-experiment design for incrementality estimation, sensitivity analysis, and stakeholder communication within the analytics & experimentation domain for data science roles focused on warehouse/fulfillment center decisions.
Amazon
Oct 13, 2025, 9:49 PM
Data Scientist
Technical Screen
Analytics & Experimentation
0
0
Build vs. Lease vs. Defer: New Fulfillment Center Decision
Context
You are evaluating whether to open a new fulfillment center (FC) to improve delivery speed and reduce costs. You have:
- Daily demand by metro/zip.
- Current FC capacities and variable costs.
- Average delivery time as a function of distance.
- Lease and build (CapEx) options with cost details.
Assume the goal is to maximize long-run free cash flow while meeting customer service-level targets (e.g., 95% next-day coverage in target metros) under operational constraints.
Tasks
- Outline an end-to-end analysis plan: demand forecasting approach, service-level targets, network optimization objective and constraints.
- Create a back-of-the-envelope (BOTE) 5-year NPV model with explicit assumptions, including cannibalization of existing FCs and ramp-to-steady-state utilization.
- Propose an experiment or quasi-experiment (e.g., geo holdouts, synthetic control, or historical replay simulation) to estimate incrementality prior to committing. Include success metrics and stopping rules.
- Provide a sensitivity analysis: identify the three most critical parameters, explain why they matter, and how to bound them with additional data.
- Draft a one-slide stakeholder narrative: recommendation, top two risks, and mitigations.
Solution
Login required
Show