The drawer is eighty dirhams short and everyone looks at the same person. That is almost always wrong: theft is the fifth cause of cash variance, not the first. Ahead of it come change given badly, the wrong payment method, an untracked void, and a float that was never counted at open. Here are the five, in the order they actually happen, and what makes each one go away.
In short
- A variance is a measurement, not an accusation. Treated as an accusation, it stops being reported.
- The first cause is giving change during a rush, and it produces variance in both directions.
- The second is the payment method: a card taken as cash creates a perfectly clean shortfall.
- Count the float at open, or you spend the evening hunting a variance created in the morning.
- One account per person turns an anonymous variance into a precise question, which is what makes it shrink.
The five causes, in order of frequency
The order matters more than the list. Searching out of order means spending an hour on the rarest cause and missing the four that explain most variance.
| The cause | What it produces | What makes it go away |
|---|---|---|
| Giving change | Variance both ways, usually small, every day. | Let the till calculate the change and display it, customer screen included. |
| Wrong payment method | A drawer short by exactly the amount of a card sale. | Take payment on the right method, and a dedicated method per platform or channel. |
| An untracked void | Variance with no explanation and no record of who voided. | Require a reason and a signed-in user on every void. |
| The float | A constant variance, identical every day, created at open. | Count and confirm the float at open, not only at close. |
| Unrecorded petty cash | The taxi, the bread, topping up a supplier, paid from the drawer. | A recorded cash-out, with a reason, every time. |
| Theft | Regular variance, one direction, often the same shifts. | Named accounts and an activity log: it shows up fast once everything else is clean. |
Note the last row: it is not that it does not exist, it is that it can only be diagnosed once the other five are eliminated. Regular variance, always the same direction, always the same shifts, in a till where everything else is clean, is serious information. The same variance in a till with no named accounts proves nothing at all.
Why the order changes everything
Three practical consequences, visible in the mood of the team as much as in the numbers.
- Variance assumed deliberate stops being reported: If the first reaction is suspicion, the cashier who is forty dirhams out puts forty dirhams in from their own pocket and says nothing. You lose the information, and the real cause carries on.
- The small causes are the most profitable to fix: Change and payment method produce variance every day. Fixing them removes most of the noise, and it was that noise making everything else unreadable.
- Variance is only useful if found early: Same-day variance can be explained: you remember the sale, the customer, the moment. The same variance found at month end cannot be explained, only recorded.
The close that surfaces the causes
Six steps, ten minutes, and in this order above all: count before reading the total, never the reverse.
- Count the float at open and confirm it in the till. Without that starting point, no evening variance can be dated.
- Every payment on its own method: cash, card, and one method per platform or channel if you deliver.
- Every cash-out is recorded, even twenty dirhams of bread, with a reason.
- Count the drawer before looking at the till’s total. Seeing the expected figure before counting influences the count, always.
- Compare, then record the variance even when it is zero. A log with only bad days is useless.
- Search in the table’s order when variance exceeds your threshold: change, payment method, voids, float, cash-outs, and only then the rest.
Step 4 is the one everybody inverts, and it is the most important. A count made after reading the expected total finds that total nearly every time: that is not dishonesty, it is how attention works.
What variance is acceptable?
We publish no threshold, and nobody should: fifty dirhams does not mean the same thing in a grocery taking a thousand a day and a restaurant taking twenty thousand, nor in a 90% cash till and a 90% card till. Build your own: record the variance every day for a month, look at the real spread, and set the point beyond which you investigate. It is your figure, not an article’s.
What weighs in Morocco
The share of cash, first. Many shops take most of their sales in cash, which means more change given, and therefore mechanically more chances of error than in a card-heavy business. That is not a lack of rigour, it is arithmetic: the number of variances follows the number of times change is handed over.
Then the float, and small coins in particular. A till short of coins forces rounding, borrowing from the neighbouring shop or leaving an IOU, and every one of those creates variance that has nothing to do with anyone’s honesty. Sizing the float in denominations, not only in total, removes a good share of morning variance. The hardware side is covered in https://blog.bellocommerce.com/en/cash-register-price-morocco/; here it is a question of organisation.
What genuinely reduces variance
In order of effect, strongest first.
- One account per person. Anonymous variance has no cause; named variance has a question.
- Change calculated and displayed by the till, with a customer screen so the customer sees it too.
- One payment method per channel, without exception.
- A mandatory reason on voids, and a log that keeps them.
- Counting before reading the total, every day, without exception.
Mistakes to avoid
- Starting with suspicion. It makes the information disappear, not the variance.
- Looking at the expected total before counting. The count aligns to it without anyone deciding to.
- Only recording bad days. Without the good ones you have no spread, so no threshold.
- Taking a card as cash to go faster: the variance is guaranteed and completely unexplainable that evening.
- Sharing an account. Everything becomes anonymous, and anonymous variance never gets fixed.
Frequently asked questions
What causes cash drawer variance?
Five things, in order of frequency: giving change, the wrong payment method, an untracked void, a float not counted at open, and unrecorded petty cash. Theft comes after, and can only be diagnosed seriously once the other five are eliminated.
What level of cash variance is acceptable?
There is no universal threshold: it depends on your turnover and your cash share. Record the variance every day for a month, observe the real spread, and set your own investigation threshold. A threshold borrowed from another shop means nothing.
How do I know who is responsible for a variance?
By giving each person an account. Without named accounts a variance is anonymous and no conclusion is possible. With them, sales, discounts and voids carry a name and a time, and the question becomes precise instead of general.
Should I count the drawer before or after reading the total?
Before, always. Knowing the expected amount before counting steers the count without anyone deciding to, and you find the figure you were looking for. Count first, compare second.
Should the float be counted every day?
Yes, at open and not only at close. A float miscounted in the morning produces an evening variance that gets hunted in the day’s sales, where it is not. Counting it at open takes two minutes and removes that whole family of variance.
What to take away
Cash variance is a measurement, not a verdict. Work it in order: change, payment method, voids, float, cash-outs, and only then the rest. Give each person an account, count before showing the expected total, and record the variance every day including the zeros. Within a month you will have your own threshold and, in most shops, far less variance left to explain.
An account per person, a close that balances
BelloPOS gives every employee their own account, records voids and discounts with the time and the user, and compares the counted drawer against the till’s total at close. It all runs offline, and the Lite licence is free for life.
Read next
Other practical guides on the same subject: