Client
Pottsalat GmbH
Unit
prodct /skills · + [Labs] adj.
Duration
8 months · 2022
Shape
Custom build · internal app pair
Headline
A custom order-handling app for the kitchen and a driver-routing app for dispatch — one operational loop, no spreadsheet.
// prodct /skills case · Pottsalat · 2022 · food delivery · internal apps

Operations on spreadsheets.
Shipped as a real app.

The kitchen and dispatch were running on a shared spreadsheet — one lag, one cell collision, one mis-typed address and the whole shift stalled. We built the operational loop as real software: an order-handling app for the kitchen, a routing app for drivers, wired to the same live state.

2 apps
// order handling + driver routing, one live state
// end-to-end
0 spreadsheets
// ops moved off shared Excel entirely
// clean cutover
Live state
// kitchen and dispatch see the same order in the same second
// no latency
// 01 / the situation

A spreadsheet running a kitchen.

Ops was a shared Excel — orders on one tab, drivers on another, addresses copy-pasted by hand. Every shift picked up the same drift: stale rows, cell collisions, one-off typos that moved pins to the wrong postcode.

It held together on good days and broke under any spike. The ask was a real operational tool — owned by Pottsalat, fitting how they actually work.

  • 01

    Shared spreadsheet as the system of record.

    Locking, merge conflicts, silent overwrites during the lunch push.

  • 02

    Kitchen and dispatch on different views.

    Same order, two states, reconciled verbally.

  • 03

    Routing by memory.

    No driver-side tool — addresses relayed, sequences improvised.

// 02 / approach

Two apps, one live order state.

We designed the app pair around the actual operational loop: order arrives → kitchen confirms → route is assigned → driver delivers. One state object flows through it; both apps read and write to it in real time.

We shadowed real shifts before writing a line of code — the kitchen dialect, the dispatch gestures, the micro-frustrations of the current setup. The apps are shaped by that, not by a generic ops template.

  • A

    Order app for the kitchen.

    Incoming, in progress, ready. Big type, no dropdowns, gloves-safe controls.

  • B

    Routing app for drivers.

    Sequenced stops, address-verified pins, handoff confirmation — works on a shift phone.

  • C

    Shared live state.

    Both surfaces react in the same second; no polling, no reconciliation calls.

// 03 / what we built

Two apps, one state, a clean handover.

[App]
Order app · kitchen surface
Incoming queue, confirmation, in-progress board, ready-for-pickup lane. Designed for a rushed kitchen, not a demo.
[App]
Routing app · driver surface
Optimised stop sequence, verified addresses, delivery confirmation. Runs on any mid-range phone, works under cellular.
[Core]
Shared live order state
One source of truth behind both apps. Kitchen and dispatch see the same order transition in the same second.
[Handover]
Runbook + training + cutover plan
Spreadsheet retired with a planned switchover — not a sudden one. Staff trained, fallback documented, ops owns it end-to-end.
// 04 / outcomes

One loop. Fewer dropped orders.

2 apps
// kitchen + driver, shipped and operating
// end-to-end
0 spreadsheets
// ops cut over cleanly from shared Excel
// no fallback
Faster delivery
// sequenced routing, address verification, fewer retries
// customer-visible
Owned internally
// runbook + training + support path all with Pottsalat
// no lock-in
// similar engagement?

Ops running on a spreadsheet? We ship the real tool.