· PM Editorial · Product Sense · 6 min read
Design a Grocery Delivery App: Metrics and North Star
How to choose a defensible North Star metric and metric tree for a grocery delivery app, covering order accuracy, delivery on-time rate, basket size, repeat rate, and NPS.
Why This Question Follows the Design Prompt
After you design a grocery delivery app in a product sense interview, the natural follow-up is “How would you measure whether this is working?” This tests a distinct skill from design: translating product decisions into a measurable framework that a real team could run against. This article builds a North Star metric and supporting metric tree for a grocery delivery app, then walks through common interviewer pushback.
Step 1: Reject the Obvious Wrong Answers First
Two tempting but flawed North Star candidates: Gross Merchandise Value (GMV) and Number of Orders. Both are lagging, top-line metrics that can be moved by discounting or paid marketing without reflecting whether the actual delivery experience is good. State this explicitly to the interviewer: “I want a North Star that reflects experience quality, not just transaction volume, because in a trust-sensitive category like groceries, volume can grow for a while even as quality quietly degrades, right up until churn hits.”
Step 2: Propose a North Star — Repeat Order Rate Within 30 Days
A strong North Star for grocery delivery is Repeat Order Rate within 30 Days (the percentage of customers who placed an order who place another order within the following 30 days). Groceries are a recurring-need category — unlike a one-off purchase, a satisfied grocery customer should reorder within weeks. This metric captures the compounding effect of every trust factor (accuracy, freshness, delivery reliability) in a single, hard-to-game number: a customer only reorders if the prior experience was good enough to trust again with their food budget.
Step 3: Build the Supporting Metric Tree
Break Repeat Order Rate into its primary drivers:
- Order Accuracy Rate — percentage of orders delivered with zero item discrepancies (missing items, wrong items, unwanted substitutions). This is the single strongest predictor of whether a customer trusts the service enough to reorder.
- Delivery On-Time Rate — percentage of orders delivered within the promised window. Late perishable deliveries directly damage trust and repeat likelihood.
- Basket Size — average order value per transaction. While not a trust metric per se, basket size indicates how much of a customer’s grocery budget they’re willing to route through the app, which correlates with confidence in the product.
- NPS (Net Promoter Score) — a direct qualitative pulse on whether customers would recommend the service, useful for catching sentiment issues that lagging behavioral metrics haven’t caught up to yet.
Each of these maps to a specific team’s roadmap: Order Accuracy Rate is owned by the inventory/substitution product area, Delivery On-Time Rate by logistics and routing, Basket Size by merchandising and cart-experience, and NPS by the cross-functional experience team as a holistic check.
Step 4: Add Guardrail Metrics
Guardrails prevent gaming the North Star. For grocery delivery: cost per delivery (should not spike as you improve accuracy, e.g., by over-staffing pickers), picker/store partner satisfaction (should not degrade if you push harder substitution rules that create friction for pickers), and refund/guarantee payout rate (should stay within a sustainable range even as the freshness guarantee reduces friction for customers). Track these weekly alongside the North Star, and treat any experiment that improves repeat rate while blowing through a guardrail threshold as a rollback candidate pending redesign.
Step 5: Handle Interviewer Pushback
“Repeat Order Rate is a lagging indicator — how do you make fast decisions week to week?” — Use Order Accuracy Rate and Delivery On-Time Rate as fast-moving proxy metrics for weekly experiments, since they move within days and are strong leading indicators of the 30-day repeat metric, reserving the full North Star for monthly cohort-based strategic review.
“What if basket size grows but repeat rate falls — which do you trust?” — Repeat rate wins, because a customer who orders a large basket once and never returns has extracted no compounding value and likely reflects an unsustainable promotion (e.g., a heavy first-order discount) rather than genuine product-market fit for the recurring use case.
“How would you segment this metric?” — Segment by customer cohort (first-time vs. established), by store partner (since accuracy varies by store’s own inventory systems), and by delivery region (since traffic and picker density vary geographically), because an aggregate repeat rate can hide serious problems in a specific region or partner store that would otherwise get averaged away.
Comparison Table: Candidate North Star Metrics
| Metric | What It Rewards | Gameability | Verdict |
|---|---|---|---|
| Gross Merchandise Value | Any transaction volume, including discounted one-offs | High — inflatable via promotions | Reject as primary North Star |
| Number of Orders | Raw order count regardless of satisfaction | High — same issue as GMV | Reject as primary North Star |
| Repeat Order Rate within 30 Days | Genuine trust and recurring-use satisfaction | Low — requires real experience quality | Recommended North Star |
| Order Accuracy Rate | Correct, complete deliveries | Low | Use as fast-moving proxy/leading driver |
| Delivery On-Time Rate | Reliable logistics | Low-Medium | Use as fast-moving proxy/leading driver |
| Basket Size | Customer confidence to spend more per order | Medium — can be inflated by upselling low-value items | Use as supporting metric, not North Star |
| NPS | Overall sentiment | Medium — survey response bias | Use as qualitative cross-check |
Step 6: Tie It Back to the Business Model
Close by connecting the North Star to unit economics: a higher repeat order rate directly reduces customer acquisition cost amortized per lifetime order, since retained customers require no new marketing spend to reorder, and it increases the predictability of demand for both the platform and its store partners, who can then commit to better wholesale terms. This kind of metric-to-unit-economics reasoning is exactly what’s expected in senior product sense interviews and is covered at length in the 100x Product Manager Interview Playbook (Amazon: https://www.amazon.com/dp/B0DBC1FQWH?tag=sirjohnnymai-20), which includes a dedicated framework for recurring-purchase and marketplace metrics.
Common Mistakes Candidates Make
The most common mistake is proposing GMV or order count as the North Star without acknowledging their gameability through discounting. The second is proposing a North Star with no supporting metric tree, leaving the interviewer unable to see what levers actually move the number. The third is ignoring guardrails, which signals a lack of awareness of unintended consequences, such as improving accuracy by simply over-staffing pickers at unsustainable cost.
Final Answer Summary
For a grocery delivery app, reject GMV and raw order count as the North Star and instead propose Repeat Order Rate within 30 Days, supported by a metric tree of Order Accuracy Rate, Delivery On-Time Rate, Basket Size, and NPS, protected by guardrails on cost per delivery, partner satisfaction, and refund payout rate. Be ready to defend why repeat rate beats top-line volume metrics, how you’d segment it by cohort, store partner, and region, and how faster-moving proxy metrics support weekly experimentation while the full North Star anchors quarterly strategy.