Internal fraud in restaurants is a silent problem. It doesn't show up in the financial statements as its own line item, and most owners don't know it exists until the damage is already significant. According to the National Restaurant Association in the United States, losses from fraud and internal theft represent between 2% and 5% of gross sales in a restaurant without proper controls.
In a restaurant billing $400,000 MXN a month, that's between $8,000 and $20,000 MXN that simply vanishes. Every month.
The 5 most common types of restaurant fraud
1. Fake comps
A server or manager applies a comp to a check the customer actually paid in full. The employee pockets the difference. It's the most common fraud because it's hard to trace: the system shows an "authorized" comp, but there's no way to verify whether the customer actually received it.
Typical pattern: an employee applies 3-4 comps in a shift when their historical average is 0-1. They concentrate the comps on high-spend tables to maximize the amount.
2. Suspicious voids
A ticket is opened, the customer is charged in cash, and the ticket is then voided in the system. As far as the POS is concerned, the sale never existed — but the money is already in someone's pocket.
Typical pattern: repeated voids by the same server, especially at the end of the shift. Voids of tickets that had been open for a while (not data-entry mistakes).
3. Unauthorized discounts
Similar to comps, but instead of wiping out the whole check, a partial discount is applied. The customer pays the full amount, the system records a smaller one, and the difference disappears.
Typical pattern: discounts that are always the same percentage (20%, 30%) applied to cash tickets, not card tickets.
4. Tip theft
In restaurants where tips run through the system, an employee can modify the amounts before reporting them. There's also the case where cash tips simply never get reported to the pool.
5. Unrecorded staff consumption
Employees who consume products without ringing them up. A coffee here, a dish there. Individually it looks insignificant, but it adds up to thousands of pesos a month in inflated food cost.
Red flags: the warning signs
These are the patterns that should trigger an investigation:
- Concentrated comps: one employee applies more comps than the team average over a short period
- Late voids: tickets voided after being open for 30+ minutes
- Cash-only discounts: a disproportion between discounts applied to cash payments vs card payments
- Low sales with full tables: a high-occupancy shift with sales below projection
- Anomalous tips: a server consistently receives fewer tips than peers with similar volume
- Repeat refunds: the same dish shows up voided or returned multiple times a week
- Specific time slots: irregularities cluster in shifts with less supervision
The problem with manual monitoring
Most restaurants try to control fraud with a combination of cameras, manual ticket review and "trust" in the team. The problem is threefold:
- Volume: a restaurant doing 30-40 tickets a day generates 900-1,200 transactions a month. Nobody reviews them all.
- Subtle patterns: smart fraud isn't an isolated event. It's micro-anomalies spread across weeks that can only be detected with statistical analysis.
- Trust bias: owners tend to trust their team, especially long-tenured employees. But tenure isn't immunity.
"The most expensive fraud isn't the new hire stealing from the register. It's the trusted manager who's been applying phantom comps for 6 months without anyone noticing."
How automated detection works
An effective anti-fraud system for restaurants needs three components:
1. A baseline per employee
Before you can detect anomalies, you need to know what "normal" looks like. The system computes historical averages per employee: how many comps they apply per week, what their void ratio is, how much discount they give on average. Each metric has a standard deviation.
2. Deviation detection
When an employee exceeds their historical pattern by more than 2 standard deviations, an alert is generated. It's not an accusation: it's a signal that something changed and deserves attention.
3. Cross-correlation
The most powerful detection comes from crossing variables: comps + time of day + payment method + occupancy. If a server applies 3 comps during a busy shift where 80% of the tables paid in cash, the probability of fraud is significantly higher than if they applied the same comp during a slow shift with card payments.
Real case: at a restaurant in Monterrey, the system detected that an employee had applied $12,400 MXN in "authorized" comps over a 3-week period. The pattern was subtle: never more than $600 per shift, always on tables paying cash, always during high-traffic shifts where supervision was lighter.
Prevention vs detection
Detection matters, but prevention is more efficient. Some measures that reduce fraud before it happens:
- Permission segregation: a server shouldn't be able to apply comps or discounts. Only the manager on duty.
- Dual authorization: any comp or discount requires approval from a second employee.
- Register close per shift: not per day. That way discrepancies surface in hours, not days.
- Table rotation: avoid having the same server always working the same section.
- Data transparency: when the team knows automated monitoring is in place, incidents drop dramatically.
The cost of doing nothing
Take a restaurant with gross sales of $500,000 MXN/month. At a conservative 3% fraud-loss rate:
- Monthly loss: $15,000 MXN
- Annual loss: $180,000 MXN
- Over 3 years: $540,000 MXN
That's more than half a million pesos that could have been net margin. For most restaurants, that figure is the difference between a business that grows and one that barely survives.
How Fullsite scans your operation every week
Fullsite's anti-fraud agent runs an automatic analysis every Friday. It reviews the full week, computes per-employee baselines, detects deviations, and delivers a report with findings classified by risk level.
If nothing unusual turns up, you get a message confirming everything is within parameters. If something deserves attention, you get the specific tickets, the employee involved, and the pattern detected. You decide what to do with the information.
It's not a punitive system. It's a visibility system. Because the best defense against fraud isn't cameras: it's the certainty that someone is watching the numbers.