Improving Amazon Fresh: prioritized changes and the metrics that show they worked

You are a product manager at Amazon working on Amazon Fresh, one of the company's grocery businesses. Propose changes that would make the Amazon Fresh experience better, and explain how you would measure whether those changes worked. ### Clarifying Questions - Does "Amazon Fresh" here mean the online grocery delivery service, the physical Amazon Fresh stores, or both? - Should the changes mainly win first orders from Prime members who buy groceries elsewhere, get existing customers to reorder more often, grow basket size, or lower the cost of each delivery? - Are fulfillment and delivery operations, such as picking, substitutions and delivery windows, open to change, or only the shopping interface? ### Part 1 — Decide what to change Set a goal for the improvement, choose which customers you are designing for, find where their grocery journey breaks down, and propose and prioritize changes. ```hint Walk one order end to end Follow a single grocery order for the customer you chose, from planning the week's meals to putting the food away, and mark each point where they might decide a store trip would have been easier. ``` #### What This Part Should Cover - A stated goal and a chosen customer segment, each with a reason - Pain points drawn from that segment's full journey, ranked by how much they hurt - Candidate changes compared on impact, effort and operational cost, with a clear first release ### Part 2 — Measure the changes For the changes you prioritized, define how you would know they worked and how you would test them before rolling them out widely. ```hint Behavior and cost For each change, write down the customer behavior it should shift and the cost it could add, then choose a metric that captures each. ``` #### What This Part Should Cover - One primary metric tied to the goal from Part 1, with a reason it beats raw activity counts - Feature-level metrics showing whether each change is used and working - Guardrails for customer harm and for fulfillment or delivery cost - A test design that fits each change: unit of randomization, duration, and how to separate the effect of one change from another ### What a Strong Answer Covers - A goal and segment that both the proposed changes and the metrics follow from - Reasoning specific to groceries (perishable items, repeat weekly baskets, scheduled delivery) rather than generic e-commerce fixes - Explicit trade-offs between a better customer experience and the cost of picking and delivering orders - A clear line from each proposed change to the metric that would show it worked, plus the main risks ### Follow-up Questions - If your top change made customers reorder more often but lowered average order value, would you keep it? How would you decide? - How would you test a change to picking or delivery operations that cannot be randomized customer by customer? - How would your answer change if the scope were the physical Amazon Fresh stores?