Sales Integrity is the deep theft read. You drop a per-server sales report from your POS, Bar Cop benchmarks every server against the rest of the floor, and it lists the ones whose numbers do not add up, with a dollar figure wherever one can honestly be worked out.

It is the one an owner runs every shift, or every week, when something feels off and the takings will not explain it.

What it catches, and what it cannot

It catches the register and cash games that leave a fingerprint in the sales data: no-sale drawer opens, void abuse, abnormal cash mix, low average checks, heavy comps and refunds, plus the drawer shorts and walked tabs Bar Cop already has on file.

It cannot catch product theft. Overpouring, free pours, watering and walking out bottles never reach a sales report, because nothing was rung. Pour cost, inventory variance and Spot Check catch that half, and you need both halves to see the whole picture.

Drop the report

Pull a per-server sales summary for a shift or a week from your POS and drop it in the box at the top. Map the columns once and Bar Cop remembers the format.

• The only column it must have is the server name.
• Net Sales earns its place several times over: the void, comp, refund and check-average reads are all measured against it.
• Include the Date column if your export has one. It is what lets Bar Cop line the report up with the drawer shortages and walked tabs you have already logged, and shortages are one of the strongest tells it has.

Most other columns unlock a signal, so the richer the export the sharper the read. A server with too few checks to judge fairly is set aside rather than flagged.

This wants a richer export than the weekly one. Your basic per-server covers and sales already import at the Shift weekly close and show up in Server Check. Use that for check averages, and this one, with the register and cash columns, for the theft patterns.

Why a server gets listed

One number out of line is usually noise. Bar Cop takes at least two separate readings before anyone appears on the page, which is why a naturally low check average on a slow station never puts somebody there on its own.

There is one exception and it is deliberately narrow. A single reading can list somebody when it runs five times the rest of the floor and the money behind it is real: a meaningful share of that server's own sales in voids, comps, discounts or refunds, or at least ten no-sale drawer opens.

Cash mix, check averages, sales per labor hour, drawer shorts and walked tabs are never enough on their own, however far off they look. A cash-heavy or a slow station explains the first three, and a bartender who runs short once or twice is a training problem, not a case.

How the numbers are worked out

Every rate is measured against the rest of the team with that server left out, so their own numbers never soften the comparison they are being judged by. The dollar figures on voids, comps and refunds are what that server ran above what the rest of the floor would predict.

Drawer shorts and walked tabs work differently. Those dollars come straight from your own Shift Control records rather than from a comparison, so nothing is subtracted from them. They cover only the days that server appears on in this file, so if your report misses a night, the shortages logged that night are not counted here either.

Dollars are shown only where they can be worked out honestly, so a listed server can show none at all. Bar Cop does not tell you what any of it means. It tells you which numbers do not line up and what the gap is worth.

Working a flag

A flag is a lead, not a verdict.

• Open Investigation starts a six-step case in Loss Prevention, so you work it the way you work any variance: watch the drawer, pull the void timestamps, talk to the shift, document what you find.
• Export PDF saves the review for an owner or a partner.

Confirm before you act on anyone. These are patterns in a sales file. They are the reason to go and look, and nothing more than that until you have looked.