Almost every operator who sets out to document their operation starts in the same place: at the beginning, intending to write it all down. Almost all of them stop somewhere around the third document, because writing procedures is slow and the shift is now.

The projects that succeed do the opposite. They document very little, pick that little carefully, and get it in use before writing anything else.

Document where the cost of variation is highest

The criterion is not "what does someone need to know." By that standard everything qualifies and you will never finish. The criterion is: where does it cost you money when two people do this differently?

That narrows a hundred candidate procedures to about eight.

Start with these

  • Opening and closing checklists — the highest-frequency, highest- consequence procedures in the building. A missed closing step costs you product overnight; a missed opening step costs you the first hour of service.
  • Portioning and plating for your top ten items — not every item. The ten that drive most of your volume are where portion drift turns into real food cost.
  • Receiving — checking deliveries against the invoice. Rarely documented, and one of the few procedures that pays for itself the first time it catches a short delivery.
  • Prep lists and par levels — the difference between prepping for what you expect to sell and prepping for what you prepped last time.
  • Cash and close-out — because the cost of ambiguity here is not only financial.

Skip these, at least initially

  • Anything performed once a year. By the time it comes round the document is out of date and nobody remembers it exists.
  • Anything only one trusted person does, that is not a bottleneck. Document it when you are training a second person, not before.
  • Company values and culture statements. Worth having. Not an SOP, and not what fixes a Friday.
  • Procedures for equipment you are about to replace.

Format decides whether it gets used

The most common reason documentation fails is not that it is wrong. It is that it is unusable at the moment of use.

A procedure gets consulted mid-shift, standing up, with wet hands, under time pressure. That rules out most of what people produce:

  • One page. If it does not fit on one page it is two procedures, or it contains explanation that belongs in training rather than reference.
  • Checkboxes, not paragraphs. Prose describes; a checklist tracks where you are. Those are different jobs.
  • Photographs for anything visual. A plating photo settles a disagreement that four sentences will not.
  • Located where the work happens. A binder in the office is a binder nobody opens.

Write it with the people who do it

Procedures written by an owner alone describe how the owner believes the work is done. Procedures written with the team describe how it is actually done — including the workaround somebody invented two years ago that everybody now relies on.

There is also an adoption effect. People follow procedures they helped write at a noticeably higher rate than procedures that appear from above, and it costs nothing to get that for free.

The practical version: draft from observation rather than from memory, then walk it through with whoever runs that station and correct what you got wrong. You will always have got something wrong.

Role scorecards, not job descriptions

A job description lists duties. A scorecard states what good looks like for that role — the handful of things the person is accountable for, expressed so that both of you can tell whether they are happening.

This is what makes training repeatable rather than personal. Without it, "is this person working out?" is a judgement call that varies by who is asked. With it, it is a conversation about four or five specific things.

Keeping it from going stale

A library that is not maintained is worse than none, because people stop trusting any of it once they find one page that is wrong.

Two habits are enough. Put a review date on every document, and make updating the procedure part of changing the process — if a prep method changes on Tuesday, the page changes Tuesday, not at some future review. One page, owned by whoever runs that area.

What this is actually for

The goal is not a complete manual. It is that the operation runs the same way whether or not you are standing in it — which is the thing that makes time off possible, makes a second location conceivable, and makes the business worth something to someone else.

Eight well-used pages get you most of that. A hundred unread ones get you none of it. If you want this built rather than attempted, SOPs and training systems is the engagement for it.