How would you prevent wrong items in deliveries?

### 1) Structure the answer (shows leadership) Use a clear frame: 1. **Define the problem + metric** 2. **Find root causes** (data + qualitative) 3. **Generate mitigations** (process + product + engineering) 4. **Prioritize** by impact/effort and stakeholder friction 5. **Pilot + measure + iterate** --- ### 2) Define metrics and scope Primary metric (choose one and define precisely): - **Wrong-order rate** = wrong-item reports / completed deliveries - Or **cost rate** = refunds + credits due to wrong items / GMV Guardrails: - Pickup time / courier wait time - Merchant acceptance / churn - Customer satisfaction (CSAT), re-order rate - Fraud rate (false claims) Segment metrics by: - Merchant, merchant size, cuisine - Courier experience tier - Order complexity (#items, modifiers) - Packaging type (sealed vs unsealed) --- ### 3) Diagnose root causes Combine: - **Customer support tags** (wrong item, missing item, wrong order, tampered seal) - Courier app events (arrived, picked up, photo, barcode scan) - Merchant POS/integration signals (order acknowledged, printed, packed) Create a simple attribution taxonomy: - **Merchant packing error** (wrong/missing items) - **Courier pickup mix-up** (picked wrong bag) - **Handoff labeling issue** (no label, illegible label) - **Customer misreport / fraud** Then identify top contributors via Pareto: e.g., top 5% merchants cause 40% of incidents. --- ### 4) Intervention ideas (process + product) #### 4.1 Low-cost process controls (fastest wins) - **Standardized packing checklist** in merchant tablet/POS (especially for modifiers). - **Prominent order label**: customer name + order ID + item count. - **Sealed packaging requirement** and seal integrity guidance. #### 4.2 Pickup verification (the “extra check” mechanism, but done thoughtfully) Goal: verify without adding too much friction. - **Scan-based verification**: barcode/QR on the bag label scanned by courier. - App validates it matches the order ID. - Works best when the platform prints labels or provides QR. - If scanning isn’t feasible: **photo confirmation** at pickup (bag label visible) for high-risk merchants/orders. - **Two-person confirmation** at merchant for high-incident locations (operational playbook, not everywhere). Trade-off: Added pickup time. Mitigation: only enforce on **high-risk segments** (merchant/order score). #### 4.3 Reduce merchant packing errors - UI improvements in merchant app: - Highlight modifiers ("no onions") - Group items by station (hot/cold) - "Mark item packed" flow for complex orders - For integrated merchants: send structured item/modifier data; ensure printer formatting is clear. #### 4.4 Courier-side improvements - Stronger pickup UX: show **bag count** expected, store pickup notes. - Training/nudges for new couriers; "repeat offender" coaching. #### 4.5 Fraud/false-claim controls (careful) - Require photo evidence for certain claim types. - Use anomaly detection to flag repeated claim patterns, but avoid punishing legitimate users. --- ### 5) Prioritization and rollout Prioritize by **Impact × Confidence / Effort**. - Start with top incident merchants (targeted rollout). - A/B test verification flows (scan vs photo vs none) with guardrails on pickup time. - Roll out gradually (1 city → 10 cities) with monitoring. --- ### 6) Measurement plan Success looks like: - Wrong-order rate down (overall and in targeted segments) - Refund cost down - No significant regression in pickup time or courier churn - Merchant satisfaction acceptable Also monitor substitution effects: - Wrong-item reports may drop but **missing-item** reports rise (packing checklist partially helps both). --- ### 7) Close with stakeholder alignment Call out that this is cross-functional: - Merchant ops (training, compliance) - Courier ops (workflow) - Product/eng (labeling, scanning, telemetry) - Support (taxonomy + feedback loop) A strong interview answer explicitly balances **customer trust**, **operational friction**, and **scalability** of the solution.