MATO

Because delivering two streets away and delivering to the next city are not the same job, and a platform that treats them as one will price at least one of them wrong. Each city in MATO carries its own zones and its own fees, so the cost is derived from the address rather than guessed at and corrected later.

Keeping that on the order, not in a lookup the checkout consults at the last moment, means the figure the customer agreed to is the figure the operations team and the rider see. A delivery fee recalculated after the fact is a refund conversation, and refund conversations are the expensive kind.

They turn demand into something the shop can staff. Groceries are not dispatched the instant they are ordered: somebody walks the aisles and fills a basket, and there is a limit to how many baskets an afternoon holds. A slot is a promise the business can keep.

That only works if capacity is held when the slot is taken, not counted afterwards. The failure is a plain race condition: two customers reaching for the last place in a window at the same moment, both told yes. A popular window then oversells, and every order in it is late, which costs more trust than showing the window as full ever would.

Because their work is not shopping. A picker needs a guided run through an order — pick, review, pack — that is hard to get wrong while moving. A rider needs the next address, navigation to the exact coordinates, and a way to say they have arrived. Neither needs a catalog or a cart.

Putting them inside the customer app would mean one interface pretending to be two. A separate operations app with role-based picker and rider accounts keeps each person's screen about the task in front of them, while both still write to the same order.

Through one timeline that the people doing the work advance — a state machine, with each step set by the picker or the rider as it actually happens, so the status is a record rather than an estimate. Picking, packed, on the way, delivered, and finally rated.

Once a rider is carrying it, their GPS position and an ETA go on a map, and customer and rider can message in-app — which is what resolves the gate code, the wrong flat number, the dog. After delivery the customer rates the order, which closes the loop the same way it opened.

Because four audiences genuinely want different things: a customer shopping on a phone, a customer shopping on a desktop, a staff member fulfilling, and an administrator running the catalog and the cities. What they must not have is four versions of the truth.

So the split is at the surface, not underneath. One REST API, one MongoDB catalog, one order model, with JWT and OTP authentication across all of them and role-based access control (RBAC) separating a picker from a rider from an administrator. The mobile apps share a React Native and Expo codebase, the web surfaces share React and Vite, and product media sits on CDN-backed object storage so every client loads the same images and video.