Skimming cash is the only kind of theft that leaves your books in perfect order. Nothing is missing, nothing is unbalanced, and no report flags anything. The money is taken before it is ever recorded as a sale, so your records are not wrong. They are just describing a smaller business than the one you actually ran.
That is what makes it so frustrating to chase. An owner who suspects skimming usually has nothing to point at except a feeling that the cash numbers are softer than they should be. There is no shortage to investigate because no total was ever created for the money to be short against.
So you do not find skimming by looking for missing money. You find it by looking for sales that should exist and do not. This guide covers the checks that actually produce an answer: the register patterns, the drawer and deposit comparisons, the inventory-to-sales test, and the surprise count that confirms it. For the wider subject this sits inside, see what employee theft is and how to stop it.
Key Takeaways
- Skimming is theft of cash receipts before recording. Cash larceny is theft after recording. Only the second one creates a discrepancy, which is why skimming is harder to catch and more common.
- The strongest signal is cash as a percentage of total sales drifting down with no change in your customer mix or pricing.
- The second strongest is a product-to-sales mismatch: units leaving inventory faster than the revenue they should have generated.
- The median occupational fraud case ran 12 months before discovery, and losses roughly scale with how long it runs (ACFE, retrieved 2026-11-16). Speed of detection matters more than certainty.
- Blind drawer counts plus deposit-to-sales reconciliation stop most skimming without anyone being accused of anything.
Lower your delivery costs by 23%
How we reduce costs:
- No delivery vehicle expenses
- Optimized local routes
- Pay-per-delivery model
- Average 23% delivery cost reduction
What skimming cash actually means
Skimming is the theft of incoming funds before they are entered into your accounting system. In the ACFE’s classification of occupational fraud, it sits on the cash branch of asset misappropriation as “theft of cash receipts,” alongside cash larceny (ACFE Fraud Tree, retrieved 2026-11-16).
The distinction between the two is the whole reason this article exists:
| Skimming | Cash larceny | |
|---|---|---|
| When the money is taken | Before it is recorded | After it is recorded |
| What your books show | Everything balances | A shortage or an adjustment |
| How you detect it | Indirect analysis of patterns | Direct reconciliation |
| How common in small business | More common | Less common |
Skimming takes a few recognisable forms:
- Unrecorded sales. The sale never happens as far as the system is concerned. Cash in hand, no ticket, no receipt.
- Understated sales. The sale is rung for less than the customer paid, and the difference is pocketed. A $40 order entered as $25.
- Receivables skimming. A customer pays an outstanding invoice in cash or by check, the payment is taken, and the balance is later written off as bad debt or covered by a later payment.
- After-hours sales. Transactions during a period that is recorded as closed, so no till session exists for them.
Each one works for the same reason: there is no entry. Which means your approach cannot be an audit. It has to be a comparison against something the thief does not control.
Why skimming is the hardest theft to catch
Every other scheme has to leave a mark somewhere in your records. A false refund creates a refund record. A ghost employee appears on a payroll register. A fake invoice sits in the vendor file. Those can be read, sampled and reconciled, which is why the schemes described in our guide to the 10 types of embezzlement are mostly discoverable by someone who reads the right report.
Skimming has no record to read. You are trying to prove the absence of something.
There is a second reason it survives: it is usually small and continuous rather than large and one-off. Ten or twenty dollars a shift does not trip any threshold you have set, does not create an obvious dip in any month, and feels to the person doing it like rounding rather than stealing. Over a year at $20 a shift, six shifts a week, that is more than $6,000 from one person, and the national detection picture says a year is about how long you should expect a scheme to last before someone notices.
The practical consequence: do not wait for proof before you act on controls. You can put blind counts and deposit reconciliation in place tomorrow without accusing anyone, and both make skimming substantially harder regardless of whether it is currently happening.
The signs an employee is skimming cash
These are reasons to pull a report, not conclusions. Taken alone each has an innocent explanation; taken together they get specific fast.
In the sales mix:
- Cash as a share of total sales declines without a shift in your customer base, your pricing, or your payment options.
- Gross margin softens while your product mix and supplier costs stay flat.
- Average transaction value on one employee’s shifts runs consistently below everyone else’s, on the same products at the same times.
At the register:
- No-sale and drawer-open events concentrated on one login.
- Receipts not offered or not printed, particularly on cash transactions.
- Transactions voided immediately after being rung, repeatedly.
- Till sessions left open across shift changes, so no individual is attributable.
In reconciliation:
- Deposits that are regularly a little under the daily close, explained as miscounts.
- Deposits made a day late, or by the same person every time.
- Cash sales that look flat on days the business was visibly busy.
In inventory:
- Product leaving stock faster than the revenue it should generate, which is the single hardest signal to explain away.
In behavior:
- Insisting on working the register alone, or on handling the deposit personally.
- Resistance to anyone else being trained on the close.
- Never taking a day off, especially around the end of a period.
That last cluster is not a character judgement, it is a structural tell: most continuous cash schemes need the same person in the same seat. The ACFE found 84% of fraudsters displayed at least one behavioral red flag, most commonly living beyond their means or visible financial difficulties (Whispli, retrieved 2026-11-16).
The cash-to-sales ratio test
This is the first check to run and it takes about twenty minutes.
Pull cash sales as a percentage of total sales, by month, for the past 18 to 24 months. Then do the same split by employee, by day of week, and by shift. You are not looking for a particular number, because every business has its own cash mix. You are looking for a trend with no business explanation behind it.
A real decline in cash usage is gradual, applies across all shifts and all staff, and tracks something you can name: a new card terminal, a new customer type, a price point change, a shift to online ordering. Skimming looks different. It appears on specific shifts or with specific people, and the rest of the business holds steady.
Then run the same comparison on average ticket. If one person’s cash transactions average $18 while everyone else selling the same things at the same hours averages $26, that gap is the shape of understated sales. One or two quiet shifts prove nothing. A gap that holds across sixty shifts is not a coincidence.
The inventory-to-sales test
This is the strongest evidence you can produce without catching someone in the act, because it uses a source the person skimming does not control: the product itself.
Pick a few items that are easy to count and hard to confuse. Bottled drinks, a particular pastry, single-stem roses, a specific case size. For a defined period, work out how many units left your stock. Then work out how many units your sales records say you sold. Then compare.
If 400 units left inventory and the register recorded 340 sales, you have 60 units of product that walked out and generated no recorded revenue. That is either theft of the product or skimming of the cash from selling it, and you need to know which. Unlike a cash ratio, this is a concrete, documentable gap.
Do it on a short, recent window first, a week or two, so the count is reliable and the staffing is known. Food and floral businesses need to subtract genuine waste and spoilage honestly, which is exactly why you pick items with predictable shrink rather than your most perishable lines. The scale here can be significant: in food service, employee theft has been estimated to account for 75% of inventory shortages (Restroworks, citing the National Restaurant Association, retrieved 2026-11-16).
Reconciling deposits against sales
Skimming of receivables and of after-hours sales shows up between the close report and the bank, so compare the three documents directly rather than trusting a summary.
Line up the close report, the deposit slip, and the bank statement entry for the same day. Do it for thirty consecutive days rather than spot-checking, because the pattern is what tells you something, not any single day.
What to look for: differences that run one direction only. Genuine counting errors scatter: some days over, some days under, roughly balancing out. Theft does not scatter. If twenty-two of thirty days are short and none are over, that is not arithmetic.
Also check timing. Deposits consistently made the next day, or only by one person, give a window in which cash can be removed before it reaches the bank and after the close report was filed. Narrowing that window is a control in itself.
For receivables, call or email a sample of customers whose balances were written off and ask, politely and neutrally, whether they recall settling that invoice. A customer who says they paid in cash months ago, against a ledger that says the debt was written off, is a complete case on its own.
Reading void, refund and no-sale reports
Your point-of-sale system is already recording the behavior around skimming even when it cannot record the skim itself. Opening the drawer requires an event. Cancelling a transaction requires an event. Both are logged with a user, a time and an amount.
Run these four reports by employee login:
- No-sale / drawer-open events. There is a legitimate need for these — making change, correcting a jam — but the volume should be broadly similar across staff doing the same job.
- Voids, with time-to-void. A sale rung and voided within seconds, repeatedly, is a pattern. One person accounting for most of your voids is a pattern.
- Refunds without a matching original transaction. These point at register disbursement fraud rather than skimming, but they travel together.
- Discounts and comps by employee. An unauthorized discount that lets the customer pay less, with the difference pocketed, looks like generosity in the log.
This only works if every person has their own login and manager override codes are not shared. If three people use the same PIN, none of these reports mean anything, and fixing that is a free afternoon’s work that makes every other check on this page usable.
Running a surprise cash audit
When the analysis points somewhere, confirm it with an unannounced count. The unpredictability is the point, and surprise audits were measured at roughly a 50% reduction in median fraud loss (PBMares, retrieved 2026-11-16).
How to run one cleanly:
- Arrive mid-shift, without notice, and take control of the drawer before the employee can reconcile it.
- Print the till reading for the session before anyone counts anything.
- Count the cash with a second person present and both of you recording the figure independently.
- Compare the physical count to the till reading, including the opening float.
- Write it down, dated and signed by both counters, and keep it.
Then read the result correctly. A drawer that is over is as informative as one that is short: cash that has been collected but not yet rung is money being parked for removal later, and it is one of the few direct physical signs of skimming in progress. A drawer that is exactly right tells you nothing conclusive, which is why you repeat it on different days and different shifts rather than concluding from one count.
Do these on everyone, including your most trusted people and including yourself if you work the till. The moment they are applied selectively they become an accusation.
Cash handling controls that make skimming hard
Most of this is process rather than purchase, and you can put all of it in place without suspecting anybody.
- Blind drawer counts. The person counting does not get to see the expected total first. This removes the ability to count to the number instead of counting the cash.
- Dual counting at close. Two people, independent figures, both signed. Most continuous cash schemes need solitude.
- Receipt discipline. Every transaction prints a receipt and the customer is offered it. A receipt is a record the employee cannot suppress, and customers unknowingly enforce the control for you.
- Unique logins, no shared overrides. Without attribution, no exception report can be acted on.
- Separate who takes money from who records it. Where headcount allows, the person running the till is not the person reconciling the day.
- Deposit the same day, logged sequentially. Shorten the window between close and bank, and keep a numbered log with the close report attached.
- Rotate shifts and enforce vacation. Job rotation and mandatory time off were measured at around a 49% reduction in median loss (PBMares, retrieved 2026-11-16). A scheme that needs the same person every day cannot survive a week off.
- Reduce how much cash exists. Every card, tap or app payment is a transaction that cannot be skimmed. Pushing cash down from 30% of sales to 10% cuts the exposure by two thirds without a single conversation.
- Make it possible to tell you. Tips were the leading detection method at 43% of cases, with more than half coming from employees (ACFE, retrieved 2026-11-16). The person on the next register usually knows.
Owners often underrate the last two because they feel indirect. They are the two with the best measured return, and neither requires you to prove anything first. For the full set of figures behind these controls, see employee theft statistics for 2026.
When the numbers point at someone
Resist the confrontation. The analysis above gives you a strong indication, not a case, and acting on an indication is how owners end up unable to terminate, unable to claim, and facing a complaint of their own.
Work in this order. Preserve what you have: export the reports, save the counts, keep the video, and do it before the person knows you are looking. Get the amount verified by someone outside the situation, normally your accountant, because the number you reach alone will be challenged. Take advice before any conversation, since termination, a police report and an insurance claim each have separate requirements and the wrong first move can compromise the other two. And keep the inquiry narrow: do not discuss suspicions with other staff, do not search personal belongings without knowing your state’s rules, and do not suspend anyone publicly.
Then change the control regardless of the outcome. Across the cases studied, 80% of affected organizations modified their controls after a fraud was found (Whispli, retrieved 2026-11-16). The change is the only part that prevents the next one, and there is no reason to wait for a confirmed case to make it.
Frequently Asked Questions
What is skimming cash?
Taking cash from a business before the transaction is recorded. Because no record is ever created, the books stay balanced and nothing appears missing. The recorded sale is simply smaller than the real one, or does not exist at all.
What is the difference between skimming and cash larceny?
Timing. Skimming happens before the money is recorded; cash larceny happens after. Larceny must create a discrepancy somewhere, which makes it easier to find. Skimming creates none, which is why it is more common and harder to prove.
How do you prove an employee is skimming cash?
Rarely with a single document. You build it from converging evidence: a cash-to-sales ratio that drops only on specific shifts, an inventory-to-sales gap showing product that left without revenue, deposits short in one direction only, drawer-open and void patterns concentrated on one login, and surprise counts that confirm it. Have your accountant verify the figures and take legal advice before confronting anyone.
Can you catch skimming with cameras?
Cameras help but rarely decide it alone, because a cash transaction with no receipt looks identical to a legitimate one on video. They are most useful once your analysis has told you which register, which shift and roughly what to look for. Pointing a camera at the drawer without doing the number work first mostly produces footage nobody watches.
How much cash skimming is normal?
None is normal, but small persistent shortages are common and usually a training and process problem rather than theft. The distinction is direction and consistency: genuine errors scatter either side of zero, while theft is one-directional and tied to particular shifts or people.
Start with the two checks
If you take one thing from this page, run the cash-to-sales ratio by employee and the inventory-to-sales comparison on a handful of countable items. Between them they will tell you within an afternoon whether you have a real problem or an uneasy feeling, and neither requires any new software or any conversation with staff.
Then put blind counts, unique logins and same-day deposits in place whatever the answer. Those controls cost nothing and make the scheme difficult to run at all. For the schemes that do leave a paper trail, see 10 types of embezzlement, and for the whole subject including what the law lets you do next, start from what employee theft is and how to stop it.