Loss attribution beats theft prevention.

The wrong goal

Walk into any gas-station owner-operator meeting and you'll hear some version of the same complaint: we're losing inventory and nobody can stop it. The implied solution is always the same too, better cameras, tighter back-room policy, sharper hiring. The implication is that the right answer to shrinkage is prevention.

After working with a dozen independent stations across California, we're convinced this is the wrong framing. Prevention is a goal that loses you money on two sides: you spend on controls that don't fully work, and you fail to capture the operational insight that's actually available when you instrument the shrinkage you couldn't prevent.

The better goal is attribution. Not “stop the theft” but “tell me who, when, and how.”

Why prevention fails

Three reasons. First, the people best positioned to commit shrinkage at an independent station are employees with knowledge of the workflows, the same people you trained to know where the cameras can't see. Second, customer-side shrink (drive-offs, walked beer/wine, scratcher swaps) requires on-site security to stop, and the unit economics rarely justify it. Third, the operational changes required to prevent shrinkage often create more friction than the shrinkage they prevent, a clerk who can't process a refund without manager approval creates customer wait times that cost more in lost trips than the refunds save.

What attribution looks like

An attribution system answers four questions on every shrinkage event:

1. When did it happen? To a specific shift, ideally to a specific hour.
2. Who was on the clock? Tied to the same shift granularity.
3. What pattern does this fit? Same employee? Same day? Same SKU?
4. What's the dollar impact? Both this event and the running total of the pattern.

Note that none of those four require preventing the loss. They require instrumentingthe loss. You get to those four answers with a data foundation that pulls together EOD, lottery serials, inventory counts, audit logs, and (where integrated) camera events.

What you do with attribution

Once you can attribute, your management posture changes. Instead of a security problem you have a management problem, and management problems have solutions. Coach the shift. Move the employee to a different role. Restructure the lottery handling. Tighten the cash drop cadence. Sometimes terminate. Always document.

The data also gives you a market-rate negotiating position. If you can demonstrate that repositioning a clerk reduced shrinkage from $X to $Y, you have a measurable case for operational changes that previously sat at the level of gut feel.

The technical shape of an attribution system

Three components, in order of importance:

1. An immutable audit trail. Every action, every login, void, refund, EOD edit, lottery scan, written to an append-only log with user, timestamp, IP, before/after values. This is the substrate. Without it, nothing else works.
2. Cross-system reconciliation.EOD cash variance ↔ POS tender ↔ lottery settlement ↔ inventory delta ↔ camera events. The reason most operators can't attribute shrinkage isn't lack of data, it's that the data lives in five systems and never gets stitched. Attribution lives in the stitching.
3. Pattern detection. The first shrinkage event is noise. The third one on the same Tuesday closing shift is signal. The detection layer is what turns the log into a manageable inbox instead of an infinite stream.

Closing

Pick attribution as the goal and most of the back-office software questions answer themselves. The vendor you want is the one that lets you read the audit trail in plain English and ask “who was on shift when the till ran short last Tuesday?” in three clicks. Everything else is decoration.