Fixing the Post-Checkout Experience in Quick Commerce

Context

Pincode was PhonePe's quick-commerce play (sunset in December 2025). Everything after the customer pays — order tracking, delivery promise, exception handling, support — was the experience bottleneck: breached delivery promises, a support queue where "where is my order?" (WISMO) queries alone were ~40% of all tickets, and fulfillment logic too fragile to expand beyond grocery.

What I owned

The entire post-checkout and fulfillment roadmap. Three builds:

1. Order state machine and exception-handling logic. I rebuilt how an order moves through its lifecycle, with explicit states and automated handling for the failure modes (rider unavailable, partial fulfillment, delayed handoff) that previously fell to manual ops. This is what made expansion from grocery to high-value retail SKUs possible — those orders can't tolerate ad-hoc exception handling.

2. ETA rearchitecture. The delivery promise was breaking because the ETA itself was unreliable — sourced from static 3PL data, with no reliable origin event. And the display logic was worse: if a rider wasn't assigned on time, the app kept showing an ETA that was physically impossible to meet. I rearchitected the calculation on our own backend timestamps plus Google Maps API routing layered over Loadshare (logistics partner) data, moved live rider tracking forward from "out for delivery" to the moment of rider assignment, and built an explicit delayed state: when the promise is breached and no rider is assigned, the app says so — and surfaces a self-serve cancellation option instead of routing the customer to an agent.

3. Rider–customer communication sync. Redesigned how delivery status reached the customer in real time, alongside automated support workflows that resolved WISMO queries before they became tickets. For damage and mismatch claims, an AI agent ran first-layer image verification against the SKU's catalog appearance — above a 90% confidence threshold it triggered an instant refund; below it, a warm handoff to a human agent with the evidence pre-highlighted.

Impact

• Order breaches down 11% — via the exception-handling and fulfillment logic
• WISMO queries down 15% — customers who can see a trustworthy live ETA don't ask where their order is
• NPS up 8 points on the post-checkout experience
• Resolution time for damage/mismatch claims down 65% (48 → under 17 minutes) via the AI verification layer
• Unlocked the high-value retail SKU expansion

Skills: logistics/OMS product design · systems thinking (state machines) · third-party API integration · support automation