Bar shrinkage: spotting real loss without accusing anyone
By Jack Hawkins · Updated
Bar shrinkage is the gap between the stock you bought and the stock you sold, priced at cost, and its cause cannot be established from the gap alone. The working definition venues should use is broader and more useful than the true-crime version: over-pour, unring comps, unlogged waste, spec drift and theft all land in the same number, and the records alone cannot tell you the relative share of those causes, which is exactly why the number is worth reading before the instinct is.
The reason this topic needs a careful page rather than a detective page is that venues get it backwards. They start at the accusation, fix nothing, and teach the team that inventory week is a hunt. The sequence below checks the records before drawing a conclusion, in an investigation sequence that checks records before people.
First, rule out your own paperwork
Before anyone did anything wrong, three boring things are usually true. A delivery got signed for without being checked, and six cases became four on the invoice-to-shelf path. A count was done by eye on partial bottles, and the same shelf read ±2 bottles depending on who squinted. A spec changed in September and the costing sheet never heard, so theoretical usage has been quietly wrong for months, attributed to everyone.
The check costs an hour: reconciliation of deliveries against invoices for the period, recount of the top ten value lines with a scale, and a spec audit against the costing sheet. As an operational investigation principle, an apparent loss can be explained at this stage, which is not a reason to stop looking. It's a reason to look in this order, because a theft investigation that begins with bad paperwork ends with a fired innocent bartender and the real problem still on the rail.
Then read the shape of the gap
Both theft and process errors can be concentrated or spread out. Patterns help choose the next check; they cannot identify a cause without supporting evidence. Run the variance calculator per product and look for:
- One product, one station, one shift. A useful hypothesis to check against counts and transfer records. If matching station counts and transfer records isolate a Friday discrepancy, investigate that workflow; a Friday POS report alone does not locate missing stock.
- High-value, low-viscosity targets. The bottles that fit a pocket and sell at a premium: aged spirits, the good tequila of the moment. A flat pattern may reflect counts, recording, pouring or deliberate loss; it does not rule theft in or out.
- Variance that starts on a date. A sudden change can follow a new recipe, a delivery error, a counting method or a shift pattern. It is not evidence against a new hire or anyone's personal finances; establish the operational facts first.
- Void and comp patterns with names. A reason-coded void report can reveal recording patterns; it is not proof of misconduct or a substitute for a fair investigation. The POS pulls guide covers which reports to run weekly so this arrives by habit, not by crisis.
None of these is proof. All of them narrow a search, which is the honest word for what this stage is.
The interventions that close most of the gap, in order
Pour to spec, measurably. Jigger culture, measured rails, spout flow checks weekly. Measure drift is one possible source of excess usage; quantify it with a repeatable pour check rather than assuming it is the largest cause.
Ring everything, including generosity. Comps need a reason code; staff drinks need a line item; "I'll sort you" needs a ban. The rule that works is that giving a drink is fine and hiding a drink is not, and the variance number is simply the enforcement mechanism.
Waste log at the ice well. Clipboard, pen, three lines a week is normal, and a logged bottle is a cost instead of a mystery.
Till and comp audits monthly. Variance review by product, void report by person, CCTV only for the narrow window the data already points at. Pulling camera footage for a whole month "to be thorough" tells your team you don't trust them, which costs more than the gin.
If the narrowed evidence still says theft, you're in a HR process, not a blog-post topic, and the HR process is what your employment lawyer and the relevant insurance arrangements are for.
The culture test
Here's the line that separates venues that shrink less over time: when the number is bad, do they ask "what in our system allowed this" or "who did this"? The first question finds the jammy speed pourer, the unring culture, the delivery bay where nobody counts. The second finds a scapegoat once and teaches two hundred people to hide the next one. Real theft is real, and it's found far more reliably by venues whose default question is the systems one.
Better Bartender's weekly review workspace brings sales and cost checks into a structured review. Use the inventory workspace to review per-product unexplained depletion, then record the next operational check with an owner and date. Automatic data connections are agreed separately through the venue pilot.
An investigation needs matched evidence
Write down the discrepancy, the product, the time window and the counting method. Recount before escalating. Check whether the sales user identifies the person who poured, the person who took payment or a shared terminal login; those are not interchangeable. Compare deliveries, waste records, authorised transfers and recipe changes in the same window. If a later investigation concerns a member of staff, follow a fair employment process and applicable data-protection rules. Keep the operational review focused on the records needed to explain the stock difference. Better measurement should improve the venue and its working relationships, not create confident accusations from incomplete data.
FAQ
What is a normal shrinkage percentage for a bar? A well-run UK venue targets total variance under about 5% of theoretical cost of goods, with 3% excellent. No measured industry-wide loss rate is established here; a venue should calibrate its own trigger from consistent counts. Trade estimates, indicative.
How do you tell if a bartender is stealing? Narrow by shape: one product, one station, one shift, starting on a date, with a matching void/comp pattern. Process causes (paperwork, over-pour, unring sales, waste) must be ruled out first because they can produce the same apparent discrepancy and must be checked before drawing conclusions.
What's the difference between shrinkage and variance? In practice, synonyms in licensed-trade usage. Some operators reserve "shrinkage" for loss with intent or mystery and call the rest process variance; Better Bartender reports both under variance and lets the shape of the number distinguish them.
Does inventory software stop theft? No software stops anything; it shortens detection from quarters to weeks and replaces suspicion with per-product evidence. Any deterrent effect would need measuring; reliable records and a fair investigation process are the practical starting point.
Method and assumptions
Cause categories and investigation thresholds are operational guidance, not a measured incidence ranking, labelled as indicative rather than measured statistics; this guide does not establish an audited UK bar-theft incidence rate, and none is implied here. People-related decisions should follow employment law advice, not web content. Method standards: data standards.
Last reviewed 19 October 2026.