MILKRUN.
The promise on the homepage is an engineering commitment, not a marketing one.
- Client
- MILKRUN
- Sector
- Grocery, quick commerce
- Scope
- App + web + admin
- Stack
- React Native, Node.js, PostgreSQL
Grocery delivery with rider dispatch and live tracking, engineered for sub-ten-minute promises.
What made it hard.
A short delivery window requires store-level inventory accuracy, forecast rider supply and dispatch decisions made in seconds, all holding at once. Any one of them failing cancels the order.
Rider supply
Demand spikes in twenty-minute windows. Supply has to be forecast, not reacted to.
Slot honesty
Showing a delivery window you cannot meet costs more than showing a longer one.
Unit economics
Every extra minute and every failed drop is measurable margin.
What we shipped.
Customer ordering and live tracking, a rider app with assignment and proof of delivery, dark store picking tools, and operations dashboards per store.
How it holds up.
Dispatch was built as its own service with replay tooling, so any assignment decision can be audited after the fact rather than argued about.
The systemStore-level inventory is the source of truth and the product refuses to display a delivery slot the system cannot back. Location data passes through smoothing before it reaches a customer-facing map.
More grocery, quick commerce work.
All 42 case studies →Building something like MILKRUN?
Tell us what you are building. A senior engineer replies within one business day. NDA on request.