About the client
A multi-site restaurant franchise taking a large and growing share of its orders through third-party delivery platforms alongside walk-in service.
The challenge
The franchise's registers predated the delivery economy. They had no API, no integration path and no realistic upgrade route from the vendor — so when Glovo and Uber Eats orders started arriving, staff bridged the gap by hand.
During a dinner rush, a person stood at a tablet reading orders off one screen and typing them into another. Every re-typed order was an opportunity for a wrong item, a wrong modifier or a wrong address, and the errors landed at the worst possible moment: peak service, in front of the customer.
The re-typing also destroyed throughput. Orders queued behind a human bottleneck that got slower exactly when volume got higher.
The solution
The integration layer came first: our agents automated the connection to each delivery platform so orders land in a single normalised queue regardless of which app they came from.
Replacing the hardware followed. Rather than a proprietary register, we built a native tablet POS running on commodity hardware — cheaper to buy, cheaper to replace, and writing directly into the same order and inventory records as everything else.
- Unified order queue — every channel normalised into one stream, so the kitchen works from a single screen rather than three.
- Native tablet POS designed around service pace, on tablets you can buy anywhere.
- Automatic kitchen routing — orders reach the right station on arrival, with no transcription step anywhere in the path.
- Live inventory link — sales decrement stock in the same system, removing the nightly reconciliation between POS totals and inventory counts.
What changed
The transcription step disappeared entirely, and with it the class of error it produced. Throughput stopped degrading under load, because the bottleneck that slowed down as volume rose was a person, and that person is no longer in the path.
The takeaway
When someone is manually moving data between two systems, that person is the integration — and integrations built out of people fail hardest under load. Automating the connection removed the errors and the bottleneck in a single change.