The lottery shrinkage playbook.

Why lottery is the canary

Most inventory at a c-store moves anonymously, a clerk rings up a candy bar and the SKU decrements, but nobody can prove a particular bar was sold rather than walked. Lottery is different. Every scratcher carries a serial number. Every book is logged by serial range when the state ships it to you. Every sale, refund, and settlement reconciles back to those serials. If a serial from book 837 was activated but never sold and never returned, that ticket either fell off a counter (rare) or left in someone's pocket.

The four-week sprint

Week 1, instrument

Wire scratcher serial scanning into the receipt + activate workflow. Every book gets barcode-scanned in at receipt. Every activation gets scanned at the terminal. Returns at end-of-book get scanned. This is not new behavior, most stations already do it. The change is that the scans now write to a queryable log that ties to shift, employee, and timestamp.

Week 2, baseline

Run the system for one full week before reading any signal. The baseline is what tells you whether week-3 anomalies are real anomalies or just normal operational noise. Most stations have a low background shrink rate (~0.05–0.10%) from torn tickets, miscount returns, and the occasional honest mistake. That's the floor.

Week 3, first signal

By the end of week 3, the first real signal usually surfaces. The most common pattern: 4–7 missing serials clustered on the same shift type (e.g., Tuesday and Wednesday closing shifts). Loss Radar flags the cluster automatically; the operator gets the alert. The reflex is to immediately accuse. Don't. Patterns of 4–7 missing tickets can have non-theft explanations: a manager comping a regular customer, a return that didn't get logged, a torn ticket the clerk threw away. Document the pattern, but wait.

Week 4, corroborate and act

If the pattern repeats in week 4 with the same shift, the case is real. At that point you have:

• Specific serials missing
• Specific shifts where they went missing
• Specific employee on the clock during those shifts
• Dollar value of the loss (book value × missing count)

That's a documented case. What you do with it is a management decision: coach, reassign, document for HR, or terminate. Whatever you choose, you're deciding from data, not intuition.

The most common patterns we see

1. Single-shift theft. One employee, one shift, repeating weekly. ~60% of cases. Highest dollar impact per case.
2. Returns-without-record. Clerk processes a customer return but doesn't log it; the missing serial is real, but the cash came back. ~20% of cases. Resolved with training, not termination.
3. Book-rotation drift. Books are activated out of sequence and the tracking gets confused. ~10% of cases. Process problem, not theft.
4. Genuine torn-ticket loss. A torn scratcher is technically a serial gap. ~10% of cases. Noise floor.

The instrumentation surfaces all four. The hard work is correctly classifying which is which. A good back office helps by showing you the cash-side reconciliation alongside the serial-side gap, if the cash matches, it's category 2 or 3, not 1.

What good operators do next

The best operators we work with don't stop at attribution. They use the instrumentation to:

• Set per-shift accountability targets. Each shift gets a tolerance band; consistent overshoot triggers a manager conversation.
• Run lottery-specific shift bonuses. Tie a small monthly bonus to zero unexplained shrinkage. Costs less than the shrinkage you eliminate.
• Use the data in hiring. A clean lottery attribution record from a former employer is a meaningful signal during background checks for new hires.