A change that arrives on a Friday arrives badly. Kitchens are running systems and you are modifying them while they are switched on, which is why the implementation matters as much as the choice — plenty of good decisions have been undone by a good machine turning up in the middle of a busy week with nobody trained, nothing written down and no way back. This topic is the boring procedure that makes the previous three worth anything.
🧪 One station, one service, two weeks
Change one thing at a time, in the quietest service that still resembles a real one. A Tuesday dinner is ideal: busy enough to be a test, calm enough that a failure is an inconvenience rather than a disaster.
The first shift with anything new is slower, always, and it is worth telling the team so. If nobody warns them, the conclusion after night one is "this doesn't work" and the change is dead before it was ever tried. Two weeks is roughly how long a new step takes to stop requiring thought — and until it stops requiring thought it is not saving anybody anything, so measuring in week one is measuring the learning, not the change.
Then move it to a Friday, deliberately, with the person who ran the Tuesdays on the line. And only after that Friday works, start the next change. A kitchen that is changing three things at once cannot tell you which one helped, which is how restaurants end up keeping the expensive change and quietly dropping the one that actually fixed the problem.
👤 Somebody owns it, by name
Every change needs one person whose job it is, and "the kitchen" is not a person. The owner of a change does four things: learns it properly first, teaches the others, keeps the written version up to date, and reports what is going wrong without it counting against them.
That last part is what makes the difference. If reporting a problem with the new system is heard as complaining, you will get silence and a workaround, and the workaround will become permanent without anybody deciding it. The useful question at the end of each shift for two weeks is short and specific: what did the new way make harder tonight? Ask it, write down the answer, and fix two of them.
How to train people properly — the sequence, the follow-up, who signs off — is the training course's subject and it is worth using here rather than improvising. What belongs in this course is narrower and easy to forget: the machine or the process needs a named owner and a named backup, because the day the owner is off sick is exactly the day the change gets tested.
✍️ Write it into the paper that already exists
Do not create a binder. Put the change into the documents the kitchen already touches every day, which by now are the ones the previous courses built.
- The production sheet gets the new quantities, the new prep step and the new timing. If the fries are now portioned at four o'clock, that is a line on the sheet with a tick box, not a conversation.
- The opening checklist gets the switch-on time and any setup — programmes loaded, temperatures set, baskets in place.
- The closing checklist gets the cleaning, with the minutes it really takes, in the order it is done. This is the single most important line in this topic.
- The recipe sheet from the cost control course gets updated if the yield, the portion or the method changed. A machine that changes portioning changes your plate cost, and nobody ever remembers to go back and adjust it.
- The plate photo gets retaken if the dish now looks even slightly different.
Five documents, all of which already exist. The rule is simple: if the change is not visible in at least one of them, it is not implemented — it is being remembered, and memory leaves with people.
🚨 The night it fails
It will fail, and probably at the worst possible moment, because that is when it is under load. The difference between a bad hour and a closed station is entirely decided in advance.
Have a manual fallback for anything on the critical path and rehearse it once, on a Tuesday, on purpose: turn the thing off and serve the shift without it. Ten minutes of discomfort in a controlled week buys you a Friday that does not collapse. Keep the service number and the model on a sticker on the machine, not in an email somebody has to find; know which spare parts fail first — seals, baskets, belts, probes — and keep the cheap ones on a shelf; and know before you buy how long a repair takes in your city, because that answer is the real risk you are taking on.
The design rule that comes out of all this is worth stating plainly: never let a single machine be the only path to a popular dish at peak. If it is, either keep a manual method the team has actually practised, or buy two smaller units instead of one large one. A kitchen that cannot serve its best-seller because one appliance stopped is a kitchen that made an operational decision by accident, in a showroom, months earlier.