What you'll learn: the one equation that decides whether a drawer is short, a worked reconciliation line by line in rupees, the eight reasons cash goes missing ranked by how often they actually happen, why counting after you have seen the target is worthless, how much variance is normal, what to do when money lands after the count, and a 20-minute nightly routine you can hand to a cashier.
It is 11:20 pm. The shutter is half down, the last table has paid, and the drawer is ₹250 short. Somebody will say the machine is wrong. Somebody else will look at the cashier. And tomorrow the same thing will happen, because nobody wrote down what actually went out of that drawer today.
Here is the part most owners get backwards: a short drawer is usually not theft. In the ordinary Indian restaurant the money went out of the till for vegetables at 6 pm, or a cash sale got rung as UPI, or the float was ₹1,800 instead of ₹2,000. Those four causes explain most nights. Theft is real and it happens — we wrote a whole guide on the twelve counter scams — but it is the last item on the list, not the first, and reaching for it first is how you lose a good cashier over a rounding error.
The short answer: why your cash drawer is short
Work these in order. In most restaurants you will find it before you reach number four.
1. Money left the drawer and nobody logged it
Vegetables, gas cylinder, ice, an auto for a delivery, the electrician. Cash walks out of the till all evening in an Indian kitchen. Every rupee of it must reduce what the drawer is expected to hold, or your honest cashier counts short.
2. The opening float was not what you think
Last night's closing float and this morning's opening float have to be the same number, counted, not remembered. A ₹200 assumption at 9 am is a ₹200 mystery at 11 pm.
3. A cash sale was rung as UPI or card
The commonest honest mistake at a busy counter. Cash is short by exactly the amount UPI is long. Compare each tender against its own total, not just the grand total, and the pair jumps out.
4. Split, credit and partner money counted as cash
A bill paid half cash half card, an udhaar bill that will be settled next week, a Talabat or Swiggy order whose money never touches your drawer at all. Each one inflates expected cash if your close treats the bill total as cash.
5. Rounding, tips and change
A rupee here, a five there, a tip added on the card machine but not the bill. This is noise. Set a tolerance and stop investigating below it.
6. Only now: somebody took it
And when you get here, you will not need to guess, because the five checks above have already given you a pattern: which shift, which login, which direction, how often.
The one equation behind every cash close
Every cash reconciliation in every restaurant in the world is this, however fancy the software:
Expected cash = opening float + cash sales + cash paid in − cash paid out − cash refunds − money spent from the till
Variance = counted cash − expected cash
A positive variance means the drawer is over. Negative means short. Zero means balanced — and, as you will see below, zero is not automatically good news.
Six of those seven terms are obvious. The one that ruins reconciliations is the last: money spent from the till. It is not the same thing as "cash paid out", which is a formal drawer movement like sending cash to the safe. This is the ₹1,200 the cashier handed the vegetable man at 6 pm because the vegetable man does not take UPI.
If your close does not subtract that, the arithmetic says the drawer should hold ₹1,200 more than it does, and the cashier is short by ₹1,200 for doing exactly what you told her to do. That single missing term is, in our experience, the most common reason a restaurant believes it has a theft problem when it has a bookkeeping problem.
A second rule follows from it. A payout must be recorded when the money leaves, not when it is remembered. A slip in the drawer is not a record, because a slip can be written after the count is done — and a payout that can be added after the count is a hole big enough to hide any shortfall in.
A worked drawer, line by line
A single-counter restaurant in Madurai. Dinner shift, one cashier, one drawer. Here is the whole close in rupees.
| Line | Amount | Where it comes from |
|---|---|---|
| Opening float | + 2,000 | Counted at shift start |
| Cash sales | + 18,450 | Cash tender only — not total sales |
| Cash paid in | + 0 | Top-up from the safe, if any |
| Cash paid out | − 0 | Cash moved to the safe mid-shift |
| Cash refunds | − 0 | Money handed back to a guest |
| Vegetables (paid from till, 6:10 pm) | − 1,200 | Expense, cash, this shift |
| Gas cylinder (paid from till, 8:40 pm) | − 1,150 | Purchase, cash, this shift |
| Expected cash | 18,100 | What the drawer should hold |
| Counted cash (blind) | 17,850 | Physically counted, before seeing the above |
| Variance | − 250 (short) | Counted − expected |
Now trace the ₹250. Pull the tender breakdown for the shift: cash ₹18,450, card ₹9,300, UPI ₹12,400. Then compare UPI against what actually credited to the merchant account that night: ₹12,650.
The bank received ₹250 more by UPI than the POS recorded. So one ₹250 bill was paid by UPI and rung as cash. Expected cash counted that ₹250 as a note in the drawer; no note ever existed; the drawer is short by exactly that amount. Cash short ₹250, UPI under-recorded ₹250 — the same mistake seen from two sides.
That is not theft. That is one wrong button at a busy counter, and now you have a specific bill to look at instead of a suspicion about a person. Open the shift's bill list, filter to ₹250, and the tender on that bill will say Cash while the customer's phone says otherwise. Fix the bill, note it, go home.
Note the direction, because it is the tell. If cash had been ₹250 over and the bank ₹250 short of the POS, the mistake would be the opposite one: a cash payment rung as UPI. Same size, different fix.
The habit that makes this five minutes instead of an hour
Reconcile every tender separately — cash against the drawer, UPI against the bank, card against the settlement. A single grand total can balance while two tenders are wrong in opposite directions, and you will never see it.
The eight reasons a drawer goes short, in the order they actually happen
Ranked by how often we see each one, not by how dramatic it is. Work down, not up.
1. Cash paid out of the till and never recorded
The single biggest cause. An Indian kitchen buys from the till all evening: vegetables, milk, gas, ice, a bike repair, an auto fare for a delivery that went wrong. None of these people take UPI at 6 pm.
The fix: a payout is entered at the moment the money leaves, with an amount, a reason and a name, and it reduces expected cash for that shift — not for the day, not for the month. And once the drawer has been counted and locked, that payout must stop being editable, because a payout you can edit after the count is a way to make any shortfall disappear.
2. The opening float is a memory, not a count
Whoever opens the shift should count the float and enter what they counted. Rolling last night's number forward without counting means every error compounds until somebody finally counts and the whole accumulated gap lands on one innocent shift.
3. A sale rung on the wrong tender
Cash tapped as UPI, UPI tapped as cash, card tapped as cash. The grand total is right, so nothing looks wrong until you compare tenders separately. Rule: cash reconciles against the drawer, UPI against the bank, card against the settlement. Three checks, not one.
4. Split payments treated as one tender
A ₹900 bill paid ₹500 cash and ₹400 card is not a cash bill and it is not a card bill. If your close buckets it by a single "payment method" field, ₹400 of card money is sitting in your expected cash and the drawer will be short by ₹400. Every split bill must contribute its cash part to cash and its card part to card.
5. Credit and udhaar bills counted as paid
The regular who signs and pays on Saturday. That bill is revenue, but it is not cash tonight. A pay-later or credit bill must sit outside the cash bucket entirely until it is settled — and when it is settled, the money belongs to the shift that received it, not to the shift that made the sale.
6. Aggregator and partner money in the cash line
A Swiggy, Zomato, Talabat or Deliveroo order is money the platform collects and settles to you later, minus commission. It must never touch the drawer's expected cash. If it does, your drawer will look short by the exact value of every online order you took.
7. Rounding, tips and change
Bills rounded to the rupee, a tip added on the card machine, change given from a pocket at a rush. This is genuine noise. It belongs under a tolerance (see below), not under investigation.
8. Someone took it
It happens. But notice that you have arrived here with evidence rather than a hunch: seven mechanical causes eliminated, a variance that repeats, a direction that never reverses, and a shift pattern. That is a completely different conversation from "the drawer is short, was it you?" — and it is the only version of the conversation that is fair to a cashier who did nothing.
If you are at this step, the counter-fraud guide covers what to check next: void and cancelled-bill patterns by login, discount value by cashier, personal-QR substitution, and what not to do when you catch someone.
Count blind, or don't bother counting
This is the cheapest control in cash handling and most restaurants get it wrong.
A blind count means the cashier enters what she physically counted before the system shows her what it should be. If the expected figure is on screen first, the count stops being evidence. A drawer that is ₹300 short gets quietly topped up from a pocket; a drawer that is ₹300 over gets quietly relieved of ₹300. Either way the variance reads zero, everyone goes home, and you have learned nothing at all.
That is why a variance of exactly zero, every single night, is not a sign of a well-run counter. Real cash handling produces small errors in both directions. A perfect zero on 30 consecutive nights means either nobody is really counting, or somebody is adjusting to the target.
Two rules make it work:
- The count locks. Once she has entered her number, it cannot be edited — not by her, not from another device. Otherwise the "blind" count is just a first draft.
- The owner's view is deliberately not blind. You open the shift with every total visible, because you are the one carrying the loss, and because a genuinely wrong locked count needs some legitimate way to be corrected.
The person handling the money produces the evidence. The person carrying the loss reads it. Collapse those two roles into one and you have no control at all, whichever way round you collapse them.
When money lands after the count
A drawer is counted at 11:15 pm. At 11:22 pm a delivery rider comes back with ₹480 in cash. At 11:30 pm somebody edits an expense from earlier in the evening.
Both of those change what the drawer should hold, after the count that was supposed to prove it. If the system keeps showing "balanced" it is now lying, and the lie is in exactly the direction a shortfall would hide in.
The correct behaviour is uncomfortable and correct: the count is invalidated and the drawer must be recounted. Not adjusted, not overridden — recounted, blind, again. It takes four minutes and it is the difference between a control and a formality.
The same rule catches the other direction. If a payout can be added or edited after the count, then a ₹900 shortfall can be dressed up as a ₹900 vegetable purchase at 11:40 pm. Freeze payouts at the count, and that door is shut.
How much cash variance is normal?
For a restaurant taking roughly ₹15,000 to ₹25,000 in cash a day:
| Variance | What it means | What to do |
|---|---|---|
| Within ±₹50 | Change-making noise | Record it. Do not investigate. |
| ₹50 – ₹200 | Probably a payout or a tender | Two-minute check of payouts and tender totals. |
| Above ₹200 | Something specific happened | Work the eight reasons above before you go home. |
| Any amount, repeating | A pattern, not an error | Plot 30 days by cashier before you act. |
Direction matters more than size. Honest mistakes scatter either side of zero. A drawer that is only ever short, never over, is not making mistakes — something systematic is happening, and it may still not be theft: an unlogged daily payout produces exactly that signature.
And record the variance every night even when it is ₹7. A tolerance is a threshold for investigating, never a threshold for recording. Thirty days of small recorded numbers is what turns "I think Tuesdays are bad" into a fact.
Why your day close and your sales report disagree
This confuses more owners than any other part of closing, and it is not a bug.
A day close is a snapshot, frozen at the moment you signed it. A sales report is a live recount, recalculated from the bills every time you open it. So if a bill from Tuesday is cancelled on Thursday, Thursday's report drops it and Tuesday's frozen close still contains it. The two numbers genuinely differ, and both are right about different questions.
Which is why the gap between them is worth reading rather than eliminating. A close that no longer matches the live report is telling you: a bill from that closed shift was voided afterwards. Who did it, when, and why is a question worth asking, and a void that leaves no trail is a hole in the counter.
Never make the close recompute itself
A close that quietly re-totals to match today's data destroys the only record of what you actually counted and signed. Keep it frozen; surface the drift instead.
Two or three counters: whose drawer is short?
The moment a restaurant runs a second billing counter, one combined cash figure stops being useful. You know ₹600 is missing; you do not know from where; and three cashiers can each say, honestly, that it was not them.
Separate everything that can be separated:
- Its own float, counted at that counter, by that cashier.
- Its own bill number series, so a bill can be traced to a till without asking anyone.
- Its own count and its own variance, locked separately.
- One day total, added up from the counters at the end — not instead of them.
Reconcile per counter, then roll up. A ₹600 gap on counter 2 is a question with an answer; a ₹600 gap across the restaurant is an argument.
The 20-minute nightly routine
Hand this to whoever closes. It is deliberately short enough to survive a busy Saturday.
Stop billing. Close the shift on the POS. (1 min)
Nothing else counts until the shift is sealed. A bill rung during the count is a variance you created yourself.
Count the drawer blind and lock the number. (6 min)
Notes by denomination, then coins. Enter the total before looking at anything else. Once entered, it does not change.
Read the variance. (1 min)
Inside ±₹50, write it down and go to step 7. Outside it, continue.
Check the payouts. (3 min)
Every rupee that left the till tonight, with a reason. Ask the kitchen out loud: "did anyone buy anything with counter cash?" This finds it most nights.
Reconcile each tender separately. (4 min)
Cash against the drawer. UPI against the bank app. Card against the machine's batch total. Two tenders wrong in opposite directions hide inside a correct grand total.
Scan cancelled bills, discounts and credit. (3 min)
How many bills were cancelled, by whom, and why. What was discounted and by how much. Which bills went on credit and are therefore not cash tonight.
Leave tomorrow's float. Bank the rest. Sign off. (2 min)
Cash to bank = counted cash − the float you are leaving. Record who closed and at what time. Tomorrow's opening float is tonight's closing float, counted again in the morning.
Seven steps, twenty minutes, every night. The value is not in any single night — it is that after 30 nights you have a distribution instead of an anecdote.
What a day-end close must record
If you are doing this on paper or in a notebook, these are the fields. If your POS records fewer than these, a gap has nowhere to appear.
| Field | Why it has to be there |
|---|---|
| Opening float | The baseline. Counted, not remembered. |
| Sales by tender | Cash, card, UPI and anything else, each on its own line. |
| Split bills, per part | Otherwise card money sits inside expected cash. |
| Credit / pay-later bills | Revenue tonight, cash some other night. |
| Online / partner orders | Money that never enters the drawer. |
| Cash paid in / paid out | Safe top-ups and safe drops. |
| Money spent from the till | The term everyone forgets. Amount, reason, name, time. |
| Cash refunds | Money handed back across the counter. |
| Cancelled bills | Count, value, who cancelled, and why. |
| Counted cash | Entered blind. Locked once entered. |
| Variance | Recorded every night, however small. |
| Closing float & cash to bank | What stays, what leaves. |
| Who closed, and when | A close with no name on it is not a control. |
Every one of those fields, filled in for you
Dinelax closes the shift blind, locks the count, subtracts what was spent from the till, buckets split, credit and partner money away from cash on its own, makes the cashier recount if money lands late, and prints a Z-report — per counter, if you run more than one.
rocket_launch Start free — no card14 days free. Then ₹1 for your first month in India.
Frequently asked questions about a short cash drawer
In most restaurants the cause is not theft. Work the list in this order: money paid out of the till for vegetables, gas or petrol that nobody wrote down; a wrong opening float; a cash sale rung as UPI or card; the cash half of a split bill; a pay-later bill counted as paid; aggregator money treated as cash; rounding and tips. Only after all seven come up clean is theft the likely answer, and by then you have a per-cashier pattern to test it with.
Expected cash = opening float + cash sales + cash paid in − cash paid out − cash refunds − money spent from the till. Variance = counted cash − expected cash. A positive variance means the drawer is over, negative means short. The last term is the one most owners forget: an expense or a supplier payment made out of the drawer reduces what should be in it, and leaving it out turns an honest cashier into a thief on paper.
A blind count means the cashier enters what she physically counted before the system shows her what it should be. If she can see the target first, the count stops being evidence: a short drawer gets topped up from a pocket and a long one gets skimmed, and either way the number matches and you learn nothing. Blind counting is the single cheapest control in cash handling because it costs nothing and changes what the number means.
For a shop taking ₹15,000 to ₹25,000 in cash a day, a variance inside plus or minus ₹50 is ordinary change-making noise and should be recorded, not investigated. ₹50 to ₹200 deserves a look at the day's payouts and tenders. Above ₹200, or any amount that lands on the same person's shifts repeatedly, is a pattern rather than an error. Direction matters more than size: honest mistakes swing both ways around zero, and a drawer that is only ever short is telling you something.
The cashier should count, blind, and lock her number. The owner or manager reviews it afterwards with every total visible. That split is the point: the person handling the money produces evidence, and the person carrying the loss reads it. If the owner counts alone, nobody is accountable for the drawer; if the cashier counts with the target visible, the count proves nothing.
Opening float, cash sales, card, UPI and any other tender separately, money paid into and out of the drawer with a reason for each, cash refunds, cancelled bills with who cancelled them, counted cash, the variance, cash left in the drawer for tomorrow, cash sent to the bank, and who closed the shift and when. Anything less and a gap has nowhere to show up.
A day close is a snapshot frozen at the moment you closed. A sales report is a live recount every time you open it. If a bill from that day is cancelled afterwards, the report drops it and the frozen close does not, so the two legitimately disagree. That gap is information, not a bug: it tells you a bill was voided after the shift was signed off, and who did it.
Record it against the shift the moment it leaves the drawer, with the amount, the reason and a name. It has to reduce expected cash for that shift, or the cashier counts short through no fault of hers. A slip in the drawer is not enough, because a slip can be written after the count is done. Once the drawer is counted and locked, no payout should be editable, otherwise a shortfall can be papered over after the fact.
Each counter needs its own float, its own bill number series, its own count and its own variance. One combined drawer total across three tills tells you money is missing but not from where, and three cashiers can each honestly say it was not them. Reconcile per counter first, then total the counters for the day.
Usually not. Most shortfalls are unrecorded payouts, wrong tenders and float errors, and accusing a cashier over a ₹200 gap on one night is how good staff leave. Theft looks different: it is repeated, it is one-directional, and it concentrates on one login. Build the pattern over 30 days before you build a case.
Start tonight: three things
You do not need new software to begin. You need three habits, and you can start all three at closing time today.
- Write down every rupee that leaves the till, when it leaves, with a reason. A page in a notebook by the drawer will do until something better exists.
- Count blind. Whoever counts writes their figure down before anyone looks at the expected total. Then compare.
- Record the variance every night, even when it is ₹9. Thirty nights of numbers will tell you more than thirty conversations.
Do that for a month and one of two things becomes true. Either the variance settles into small numbers in both directions — in which case your counter is fine and you can stop worrying — or a pattern appears with a name, a shift and a direction attached to it, and then you will know exactly what you are dealing with instead of suspecting everyone.
Dinelax POS was built around this close: a blind count that locks, expected cash that already subtracts what was spent from the till, split, credit and partner money kept out of the cash bucket on their own, a forced recount when money lands after the count, an owner-only void that leaves a trail, and a Z-report at the end of every shift — per counter if you run two to nine of them. You can see what that looks like, or what it costs, before you install anything.
Stop guessing where the money went. Close the shift blind tonight.
Start free — no card14 days free, no card. Android tablets or a Windows till.