Bookkeeping for a Restaurant: A Practical Guide

What does restaurant bookkeeping actually look like?

Three rhythms, not one. A restaurant that only does bookkeeping monthly is doing accounting; a restaurant that does all three is running a business it can steer.

The distinction matters because the three answer different questions. The daily work tells you whether today was honest — whether the money in the till matches the money the POS says you took. The weekly work tells you whether the operation is drifting. The monthly work tells you what you earned, and it is the only one most operators do.

The daily half is a routine worth getting right on its own; the six-step version is in our daily reconciliation guide. The rest of this guide is about the numbers the weekly and monthly work produce.

Labour cost, and why it only means anything as a ratio

Labour is usually a restaurant's largest controllable cost and the one most often looked at as an absolute number — which is exactly what makes it useless.

A wage bill that rises is not a problem if sales rose faster. A wage bill that is flat is a problem if sales fell. The absolute figure cannot tell you which of those happened, and it is the one on the bank statement, which is why it is the one people react to.

Split front of house from back of house once the ratio is stable. They move for different reasons: kitchen labour tracks covers, and service labour tracks how long those covers sit. A single blended number rising tells you something changed; the split tells you which.

Covers and average check: the two numbers behind every revenue change

Revenue is a product, not a fact. It is covers multiplied by what each one spent — and a change in the product means nothing until you know which half moved.

Take a month down 8% on last year. That single figure supports three completely different stories, each with a different fix:

• Fewer covers, same average check — You have a demand problem. The people who come still spend what they used to; there are just fewer of them. That is marketing, location, or a competitor — not your menu.
• Same covers, smaller average check — You have a menu or service problem. The room is as full as ever and each table is worth less: dropped starters, no upsell, a price rise that pushed people to cheaper dishes.
• Both down — Something structural changed — a road closure, a review, a shift in who is nearby. Worth investigating before touching either lever, because moving the wrong one makes it worse.

Track both weekly and the story tells itself. This is also the only honest way to judge a price increase: if the average check went up and covers held, it worked. If covers fell by more than the check rose, it did not, and a revenue total alone would have looked flat either way.

Counting stock, and what cost of sales really means

Cost of sales is not what you bought. It is what you consumed — and in a restaurant those two numbers are nowhere near each other in any given month.

Without a count at each end, your cost of sales is just your purchases, which means it swings with your ordering schedule rather than your trading. A month where a big delivery landed on the 29th looks like a terrible margin month. A month where you ran down the walk-in looks brilliant. Neither is true, and both will be reversed next month.

This is where a restaurant differs from a café in kind, not degree. A café counts a handful of lines and can get away with counting rarely. A restaurant's cost of sales is dominated by things that spoil, get over-portioned, or walk — so the count is not administrative tidiness, it is the measurement itself.

Once you have real cost of sales, comparing it against what your recipes say it should have been is where the money is. That gap — over-portioning, waste, comps, theft, recipe drift — is covered in our food cost percentage guide.

Margin by channel, because a blended line hides the problem

Dine-in, delivery and takeaway are three businesses sharing a kitchen. They have different revenue per dish and very different costs, and a single Sales line lets a loss-making one hide inside a profitable one.

The numbers above are illustrative, but the shape is the point: a marked-up delivery price can still leave you worse off per dish once commission is taken, and better off in total if it fills a kitchen that would otherwise be idle. Both can be true at once — which is precisely why you need the channels apart to decide anything.

The bookkeeping that makes this visible is unglamorous: recognise the gross sale on the day it happened, book the commission as its own expense, and clear the receivable when the platform actually settles. Recording only the net deposit as revenue understates your sales and makes the commission invisible — and a commission you cannot see is one you will never negotiate. More on that in delivery receivables.

Closing the month

A month you can still edit is a month you cannot report on. Closing is the act that turns a set of records into a statement someone can rely on.

1. Count stock — On the last day, or the first morning before anything moves. This is what turns purchases into cost of sales.
2. Accrue what you owe and are owed — Invoices received but unpaid, wages earned but unpaid, platform payouts not yet landed. A month that only counts cash movements is not a month's trading.
3. Reconcile every account — Bank, cash, each delivery platform, each supplier balance. An unexplained difference here is an error somewhere else that has not surfaced yet.
4. Read the statements against last month — Not for the profit figure — for the lines that moved more than trading did. That is where mis-postings show up while you can still remember the transaction.
5. Lock the period — After this, corrections are made as adjustments rather than by editing history. It is what makes a report reproducible.

The reason for the last step is not bureaucratic. If a closed month can be edited, then the statement you sent your accountant, your bank or your investor last week may no longer be the statement the system produces today — and nobody can tell which one was right.