The restaurant close, in the order that proves things.

A restaurant close is not a generic close with food in it. Money arrives in four forms before the night is over, two of the biggest balances on the sheet belong to somebody else, and the P&L has to be right by the second week or it is decoration. Here is the sequence, what each step actually proves, and the four places it quietly fails.

LedgerInbox · September 2026 · corrections@ledgerinbox.com

Before anything: fix the cut-off

The last Friday and Saturday of the month are the whole problem. Their card sales are revenue in the old month and cash in the new one. If sales are posted on the deposit date, the month ends two days early and every year-over-year comparison is quietly wrong.

This is what the card settlements clearing account is for — see the chart of accounts for where it sits. Sales land in the clearing account on the service date; the processor deposit clears it two days later, in whichever month it falls. The balance at month end should be non-zero — it is the weekend in transit — and knowing what it should be is the first proof of the close.

The sequence

The steps that are actually hard

1 · Count the days

Thirty-one days, thirty-one entries. A POS integration that silently skipped a Tuesday produces a P&L that looks entirely plausible — sales down 3%, nobody asks. Sort the sales journal by date and count. It takes a minute and it catches the error class nothing else will.

4 · Prove cash daily, not monthly

A month of drawer variances netted together is a number that means nothing: $40 over on the 8th and $40 short on the 19th reports as a clean zero, and a steady $30-a-day shortage reports as $900 with no way to see the pattern. The comparison is per service day, per location — declared cash from the POS against what the deposit slip says arrived. The daily math, from a Toast export.

5 · The tip rollforward

Opening balance, plus tips accrued at the register, minus tips paid — through payroll and out of the drawer — equals what the liability should be. Then compare it to what QBO says. These two numbers drift apart slowly and for boring reasons: a payroll item mapped to wage expense, a night of cash tip-outs nobody booked. The entries, and the proof.

7 · Delivery, one platform at a time

Each platform pays net of commission, on its own schedule, sometimes with adjustments for refunds a week later. Three numbers have to agree: gross menu sales in the POS, the commission the platform charged, and the deposit. When they do not, the difference is usually a refund or an adjustment fee that never got booked — small individually, and a four-figure annual leak in aggregate.

10 · The split week

Payroll almost never ends on the last of the month. If the period runs the 26th to the 8th, roughly five days of wages belong to the month that just closed. Accrue them — wages, employer taxes, and the benefits that ride along — and reverse on the first of the next month. Skipping it moves labor between months, which is exactly the line the client watches weekly.

The six balances that have to be proven

Reconciled means agreed to something outside QuickBooks. A balance that has only ever been compared to itself is not reconciled, however long it has looked reasonable.

Where the close actually breaks

It is done monthly when the evidence is daily

Cash and deposits are daily facts. Reconstructing three weeks of them on the 5th, from memory and a stack of exports, is why the close takes two days and why cash gets skipped first when it runs long. Everything in step 4 can be computed the morning after the service day, when a manager can still answer the question.

The automation is trusted past its evidence

A daily sales entry from Shogo or Bookkeep is accurate about what the POS recorded. It is silent about whether the cash arrived, whether the tip liability ties, and which deposit belongs to which service day — because it never sees the bank or payroll. What a sales entry structurally cannot see.

Nobody locks the period

Settings → Advanced → Close the books, with a password. Without it, a reclass in March silently changes February's prime cost after the client has already seen it, and the second version of a signed-off month is worse than a late one.

The variance is calculated but not asked about

Food cost moved 2.1 points. The close is not finished when the number is produced; it is finished when someone knows why. A month-over-month comparison of food cost %, labor % and prime cost, with a sentence against anything that moved more than a point, is the deliverable the client is actually paying for.

Make the close short by moving it earlier

A close that takes a day is usually a close where the daily row was done daily. A close that takes three is one where all of it was left to the fifth.