Built Different · 02

The math behind 10-minute delivery

A city about the size of Bengaluru. Orders cluster where people live. Every dark store gets placed where demand is. Drag the sliders and watch which orders arrive in 10 minutes (green), in 10 to 15 (amber), or later (red). The rings show how far a rider can actually get in the time that's left.

under 10 min10 to 15over 15dark store

–%

of orders arrive within 10 minutes

Median delivery–

How far a rider can get in time–

Dark stores 40

Rider speed in traffic 20 km/h

Pick and pack 2.0 min

The last 100 metres (gate, lift, flat) 1.5 min

Where the 10 minutes go, median order

assign riderpick + packhand-offridelast 100 m

The geometry sets the rules

Take away assignment, picking, hand-off and the last 100 metres, and a rider has maybe 5 to 6 minutes on a scooter. At 20 km/h on real roads, that's about 1.5 to 2 km. That's why dark stores sit every couple of kilometres, not in a warehouse on the edge of town.

Seconds matter inside the store

Pickers follow a route-optimised pick list, with fast movers like milk and bread shelved closest to the packing desk. Every minute saved at the shelf is about 250 more metres a rider can reach.

Then it's a real-time location problem

Zepto's engineers describe a geospatial layer that knows where every store, rider and customer is, answering about 1,200 to 1,500 queries a second in roughly a millisecond each. That's what picks the store and the rider the moment you hit pay.