Deep dive
What a quick commerce MVP must get right
The first release succeeds if a customer opens the app, sees what is really in stock, pays in seconds and gets the order quickly and complete. Everything else can wait. That means a reliable stock ledger, a fast picking flow, a rider waiting at the store and an ETA based on real conditions, not marketing.
Launch with one or two dark stores in a dense area and a curated catalogue of the products people buy most often. You will learn how long picking really takes, where stock goes missing and what customers order at different times of day. The MVP development approach keeps scope small enough to change after those first weeks.
How a dark store runs on software
Each dark store is a small warehouse with numbered bins. When an order arrives, the backend reserves stock, creates a pick list sorted by walking path and pushes it to the next free picker. The picker scans each item, which confirms the product and decrements stock, packs the bag and marks it ready. A rider at the store is assigned the moment packing finishes, or slightly before, so no time is lost.
Inward stock follows the same discipline. Goods received are scanned against purchase orders, put away into bins and become sellable only when the system knows where they are. Expiry dates are captured at receipt so the picker app can enforce first-expiry-first-out.
- Reserve stock at checkout and release it on cancellation or timeout.
- Run cycle counts daily on fast movers and reconcile every difference.
- Measure pick, pack, wait and ride times separately for every order.
Forecasting and replenishment
A dark store holds limited space, so the question is always which items, in what quantity, at which store. In the growth tier, simple rules based on sales velocity and supplier lead times drive reorder suggestions. In the scale tier, item-by-store forecasts by hour and day, built by our machine learning team, reduce both stock-outs and wastage on perishables.
Forecasts are only as good as the data behind them. Record every out-of-stock moment, every substitution and every write-off from day one, even if you will not use the data for months. That history becomes the training set later. Until then, a weekly review of the top sellers per store, with par levels adjusted by hand, captures most of the benefit at almost no cost.
Riders, safety and scaling
Fast delivery should never mean unsafe riding. Design the promise around store-side speed, not rider speed: short distances, quick handover and realistic ETAs. Rider apps should avoid timers that pressure riders on the road, and incentive schemes should reward reliability rather than speed alone. Data protection rules such as India's DPDP Act apply to customer addresses and rider location histories, so set retention periods.
Scaling adds stores, cities and complexity in planning: where to open the next store, how to adjust service radii and how to balance riders between busy and quiet stores. Plan roughly 15-20% of the build cost per year for maintenance and support, on top of cloud, maps, messaging and payment fees.