--- title: "The signs of theft your restaurant's own reports already show | GoSufra" description: "The patterns that separate a sloppy process from a deliberate one — voids after firing, discount concentration, a drawer that is always exactly right — and the one control that closes most of the gap." url: "https://gosufra.com/en/blog/spotting-theft-in-a-restaurant/" language: "en" source: "https://gosufra.com" --- # The signs of theft your restaurant's own reports already show The patterns that separate a sloppy process from a deliberate one — voids after firing, discount concentration, a drawer that is always exactly right — and the one control that closes most of the gap. 2026-09-24 · 10 min read ![Illustration of a report with one line highlighted beside a cash drawer](https://gosufra.com/blog/spotting-theft-in-a-restaurant.svg) ## Key points - A drawer that balances perfectly every single day is more suspicious than one that is off by small amounts. - A void before the kitchen sees the item is a correction. A void after the food was made is the pattern worth investigating. - The strongest control is not surveillance — it is that food cannot leave the kitchen without an order existing first. - Reports show patterns, never proof. Use them to decide where to look, never as the basis for an accusation. Every restaurant owner eventually has the same suspicion, usually at two in the morning with a stocktake in front of them: the numbers do not work, and somebody must be taking something. Sometimes that is true. More often the money is leaving through a process that was never tightened — unweighed deliveries, unlogged waste, generous portions, a discount habit nobody defined. The two look identical in a monthly report, and telling them apart is the actual skill. This article is about reading the difference, and about closing the gaps that make the question necessary in the first place. ## Start from the right assumption Before looking at anyone, accept two things. **Most losses are process, not people.** If your actual food cost exceeds theoretical by four points, the most likely explanations are portioning, receiving and waste — in that order — long before anybody’s pockets. **Opportunity creates the problem.** In practice, losses appear where three things overlap: a gap in the system, someone under financial pressure, and a story that makes it feel acceptable (“they underpay me”, “it is only a few pounds”). You cannot manage the second or third. You can close the first, and closing it protects honest staff too — nobody wants to be the one who was alone at the till the night money went missing. ## The patterns in your reports Here is what to look at, and what each pattern actually suggests. ### Voids and cancellations, by timing This is the most informative single report in a restaurant, and it turns on one distinction: - **A void before the item was fired to the kitchen** is a correction. Wrong table, wrong item, guest changed their mind. Normal, and should be frequent. - **A void after the kitchen made the food** means something was produced and then removed from the bill. The food went somewhere. A modest number of the second kind is unavoidable — a dropped plate, a genuine complaint. A pattern of them, concentrated on one person or one shift, is the clearest signal a restaurant produces. ### Discounts and comps by user Pull discounts by staff member, as a percentage of their own sales, over a month. You are not looking at who gives the most discounts; you are looking at who is an **outlier against their own peers on the same shifts**. A manager with authority to discount who uses it three times more than the other managers is worth a conversation — which may well end in “I did not realise I was doing that.” ### A drawer that is always exactly right This is the counter-intuitive one, and it is worth internalising. A cash drawer handled by humans should be off by small amounts, in both directions, most days. Perfect balance every single day usually means the count is being *made* to balance — someone adjusting the declared figure to the expected one rather than counting and reporting honestly. Small variance in both directions is health. Zero variance forever is a flag. Consistent variance in one direction is a process error, usually in how change or tips are handled. ### Refunds and post-close activity Refunds processed after service, adjustments made after a shift is closed, orders reopened and modified. Every one of these may be legitimate. All of them should be **rare, logged, and attributable to a named user**. A system that allows a closed shift to be edited without a trace cannot be audited at all. ### Sequence gaps Receipt numbers should run without holes. A missing number is either a system fault or a removed record, and both need an answer. This is also exactly what a tax audit looks for first, which makes it worth fixing for two reasons at once. ### Theoretical versus actual, at item level The aggregate food cost variance tells you there is a problem. The item-level variance tells you where. Compare, per ingredient: what your recipes say should have been consumed against what your stocktake says was consumed. Then look at the shape of the gap: | Gap pattern | Likely reading | | --- | --- | | Spread thinly across everything | Portioning and waste — a process issue | | Concentrated in high-value items only | Worth a closer look; expensive items are the ones that travel | | Concentrated in easily-resold items | Bottles, cigarettes, oil, protein — the classic list | | Appears only in certain weeks | Correlate with rosters before concluding anything | That last row matters. Correlation with a roster is a reason to look more carefully. It is not a finding. ### The one reports cannot catch Here is the honest limitation: **a sale that was never entered into the system does not appear in any report.** Cash taken, food served, nothing recorded. The reports are clean because nothing happened, as far as they know. This is why the fix below is structural rather than analytical. ## The control that closes most of the gap If you do one thing after reading this, do this one: **make it impossible for food to leave the kitchen without an order existing in the system first.** When the kitchen works from a screen rather than from shouted instructions and handwritten slips, an order must exist to be cooked. There is no path from ingredient to guest that does not create a record. The unrecorded cash sale stops being difficult to detect and starts being difficult to perform. This is worth understanding properly, because it is the opposite of surveillance. Nobody is being watched. The process simply has one entrance, and everything that comes out of the kitchen came through it. Kitchen display screens are usually sold on speed and ticket accuracy; this is the quieter reason they change a restaurant’s numbers. The equivalent control on the buying side is **separation of duties**: the person who places the order, the person who receives and signs for it, and the person who pays the invoice should not all be the same person. In a small restaurant that is hard, but even splitting it two ways closes most of it. ## The controls that cost nothing - **Weigh deliveries, at random, in front of the driver.** You do not need to weigh everything. You need it to be known that anything might be weighed. - **Rotate who counts stock.** A person who counts the same section every month, unsupervised, is a structure you created. - **Count high-value items weekly**, everything else monthly. - **Set approval thresholds.** A waiter can void before firing; a manager approves anything after. A discount above a set percentage needs a second person. - **Give every person their own login.** A shared manager PIN destroys attribution entirely, and it is the single most common control failure in restaurants. - **Log waste as waste.** Staff meals as staff meals. Everything given away has a category, or it turns into unexplained variance three weeks later. - **Count drawers blind.** The person counting should not see the expected figure until after they have declared their count. None of these require software. All of them are harder to do on paper. ## What the system should give you When you are evaluating whether your current setup can answer these questions at all, test it against these: - Every action attributable to a **named user**, including voids, discounts, refunds, price overrides and stock adjustments. - An **audit trail that cannot be edited**, where a correction is a new entry reversing an old one rather than a rewrite of history. - **Role-based permissions**, so authority to discount, void or adjust stock is granted deliberately rather than shared informally. - **Per-branch isolation**, so a manager sees their own site and not the others. - Reports that compare **theoretical to actual at item level**, not just in aggregate. This is what GoSufra’s roles, permissions and audit trail are for, and it is worth being clear about what that does and does not achieve: it does not catch anyone. It makes the gaps small enough that the question stops coming up at two in the morning. ## How to act on a pattern This part matters more than the analysis, and it is where owners most often do damage. 1. **Verify the data first.** Before anything else, rule out a system explanation: a misconfigured recipe, a double-counted delivery, an item ringing up under the wrong category. A surprising number of “theft” investigations end here. 2. **Look at the process before the person.** Could an honest employee produce this pattern by following current practice? If yes, fix the practice. 3. **Ask, do not accuse.** “I am seeing a lot of voids on your shifts, help me understand what is happening” gets you information. An accusation gets you a defensive answer and, if you are wrong, a good employee who leaves. 4. **Document what you find**, in writing, with dates and figures. 5. **Follow the law and your own contracts.** Disciplinary action, deductions and dismissals have legal requirements in every jurisdiction, and getting the procedure wrong can cost more than the original loss. Take advice before acting. A report is a reason to look. It is never, by itself, a conclusion about a person. ## The month-one plan 1. Give every staff member an individual login. This week. 2. Turn on manager approval for post-fire voids and for discounts above a threshold. 3. Start blind drawer counts. 4. Write recipes for your top twenty items so theoretical cost exists at all. 5. Count your ten highest-value ingredients weekly, and read the item-level variance. 6. Route kitchen orders through a screen, so nothing leaves without a record. After two months of this, you will still have variance — every restaurant does. But it will be a number you can explain, which is the only version of this problem that can actually be managed. - theft - shrinkage - internal control - audit trail ## In the product ### [Point of Sale & Orders](https://gosufra.com/en/features/restaurant-pos/) Explore ### [Inventory & Purchasing](https://gosufra.com/en/features/inventory-management/) Explore ### [Analytics, Branches & Control](https://gosufra.com/en/features/analytics-and-reports/) Explore ## Run all of this from one system POS, kitchen, inventory, recipe costing, staff and accounting — connected, and free to start. [Create your free account](https://app.gosufra.com/register?lang=en) ## Related articles ![Illustration of a cash drawer beside a variance figure](https://gosufra.com/blog/cash-shifts-and-reconciliation.svg) Operations 2026-08-17 · 7 min read ### [Cashier shifts and the day-end count: making the drawer explain itself](https://gosufra.com/en/blog/cash-shifts-and-reconciliation/) How opening float, cash in, cash out and refunds produce an expected drawer, what a variance actually tells you, and why a close that always balances is the one to worry about. Read article ![Illustration of neatly stacked inventory crates](https://gosufra.com/blog/inventory-control.svg) Inventory 2026-05-18 · 9 min read ### [Inventory control: stopping leakage before it reaches the P&L](https://gosufra.com/en/blog/inventory-control/) A practical stock system for restaurants: purchasing and receiving, count cadence, batch and expiry tracking, and reorder points that prevent running out mid-service. Read article ![Illustration of a key beside a tree of permissions](https://gosufra.com/blog/roles-and-permissions.svg) How it works 2026-08-25 · 7 min read ### [Nine roles, one system: who can see and do what](https://gosufra.com/en/blog/roles-and-permissions/) How role-based permissions work in practice, why a branch manager should not see another branch, and what an audit trail is actually for when something goes missing. Read article [← All articles](https://gosufra.com/en/blog/) [Contact us](https://t.me/GoSufraSupport)