Daily Cash-Up: Closing a Café Till Without the Nightly Argument
12 August 2026 · MidaOne
Closing time in a lot of cafés looks the same: the drawer is counted twice, the number does not match the till, someone starts asking who served table four, and forty minutes later everyone goes home slightly annoyed. Nobody stole anything. The count was never going to match, because the shift was never recorded in a way that could be checked. A good cash-up takes about ten minutes and ends in a number you either accept or investigate — not an argument.
What the daily cash-up is actually for
It is not about catching a thief. Most cash-up differences are ordinary: wrong change handed over during a rush, a sale rung up on the wrong payment type, a refund given from the drawer and never recorded, a delivery driver paid in cash out of the till. The point of counting every night is that you find these while you can still remember the shift. A difference you spot the same evening is a five-minute conversation. The same difference found at the end of the month is a mystery nobody can solve, and it becomes the cost of doing business — which is exactly how a small leak turns permanent.
Set a float and never break it
Your float is the fixed amount of cash the drawer starts every shift with — small notes and coins, enough to give change through the first hour. Pick a number that suits your ticket sizes, write it down, and keep it constant. The whole of cash reconciliation rests on this: if the float moves around, there is nothing to measure against, and every count becomes an estimate.
- Count the float in, not just out. The opening person confirms the drawer is right before the first sale, so a shortfall belongs to a known shift.
- Never pay a supplier or a driver from the drawer without recording it. If you must, put a signed note in the till for the exact amount — that note is cash until it is entered.
- Keep tips out of the drawer. A separate tin removes the single most common reason a count comes up strange.
- Bank the surplus daily. Cash sitting in a café overnight is a risk you are not being paid to carry, and a fat drawer makes miscounts easier.
The ten-minute close, in order
Sequence matters more than speed. Do these in the same order every night and the count stops being a debate about method:
- Stop selling. Close the till to new sales before counting, or you are counting a moving target.
- Print or open the day's takings by payment type — cash, card, and anything else you accept. You are looking for what the system says the drawer should hold in cash, not the day's total revenue.
- Count the cash physically, notes then coins, without looking at the expected figure. Knowing the target before you count is how counts quietly bend towards it.
- Subtract the float. What is left is your actual cash takings.
- Compare to what the till says. Write both numbers and the difference in the same place every day, even when the difference is zero.
- Investigate anything past your threshold now, tonight, while the shift is still in someone's head.
How big a variance should actually worry you?
There is no official figure here and anyone who gives you one is guessing at your business. What matters is that you set a limit, write it down, and apply it the same way every night. A common approach is a small fixed amount for a quiet day and a small percentage of cash takings for a busy one, whichever is larger — a AED 5 difference on a AED 400 cash day is noise, and the same AED 5 on a AED 4,000 day is even less. The bands below are a starting point to adapt, not a standard:
| Difference | What it usually means | What to do |
|---|---|---|
| Under ~0.5% of cash takings | Ordinary change errors | Log it, move on |
| 0.5%–2% | A specific mistake in one shift | Check refunds, voids and payment types tonight |
| Over 2%, or repeating on the same shift | A process problem, not an accident | Sit down with the shift, review the pattern before assuming anything |
The pattern is what carries the information, not any single night. A drawer that is short on Tuesday and over on Wednesday is almost always someone giving change badly. A drawer that is short every Thursday evening by roughly the same amount is telling you something else entirely, and it is worth watching quietly for a couple of weeks before you say a word to anybody.
Where the money usually went
Work through these in order, because they account for most differences you will ever see:
- Payment type mismatch — a card sale rung in as cash, or the reverse. Your cash count is short or over by exactly one ticket. This is the single most common cause.
- Refunds and voids — money handed back but never entered, or entered twice. Requiring a manager to approve a void makes this visible instead of invisible.
- Cash paid out of the drawer — the milk delivery, the gas bottle, the plumber. Ordinary and legitimate, and it wrecks the count unless it is recorded.
- Staff meals and drinks — if they are not rung in at zero or at cost, your stock and your cash tell two different stories. Our guide to café inventory management covers the stock half of the same problem.
- Change fund raided for the float — someone topping up small coins mid-shift from the takings, without a note.
Where MidaOne fits
A reconcilable night depends on one thing: the expected figure has to already exist when you start counting, rather than being rebuilt from a spool of receipts. MidaOne records every sale as it is rung, applies the 5% VAT per sale, and scopes what each staff member is allowed to do by role and by branch — so a void or a refund is attached to a person and a time rather than being a gap. Its offline mode also matters more than it sounds at closing time: if the internet drops during the evening rush, sales keep going through and sync when the connection returns, so you are not counting a drawer against a record with a hole in the middle of it. If you are still running a basic till, our comparison of a POS versus a cash register explains why the nightly count is usually the first thing to break.
Close the day against a record you did not have to rebuild. Free for 14 days, no card.
Start your free trialFrequently asked questions
How do you reconcile a café till at the end of the day?
Close the till to new sales, pull the day's takings split by payment type, count the cash without looking at the expected figure, subtract the opening float, and compare the two numbers. Record both figures and the difference every day, even when it is zero, because the record is what makes a pattern visible later.
What is a normal cash variance for a café?
There is no official standard, so set your own limit and apply it consistently. Many small cafés treat a difference under roughly half a percent of cash takings as ordinary change error and investigate anything larger the same night, while the size of the number matters far less than whether it repeats on the same shift.
Why does my till never match the cash in the drawer?
Usually a sale rung up under the wrong payment type, a refund handed over without being recorded, or cash paid out of the drawer for a delivery. Work through those three before considering anything else — between them they explain the large majority of differences.
Should staff count their own drawer?
It is generally better for the person counting not to be the only person responsible for the shift's takings. A common arrangement is that the closing staff member counts and a manager or the owner confirms the figure, which protects honest staff as much as it deters anything else.
Do I need to keep daily cash records for tax purposes?
You need to keep proper business records, and your daily takings are part of that. Ask your accountant what form and retention period apply to your business — daily cash-up sheets alongside your system's sales records are the usual answer.
The cafés where closing time is calm are not the ones with the most honest staff. They are the ones where the expected number already exists before anyone opens the drawer, and where the same six steps happen in the same order every night. Get that in place and the nightly argument does not get resolved — it stops happening.