Full-service restaurant liquor empty-bottle reconciliation audit - POS sales, inventory variance, and empty-bottle recovery checked three ways, run-direction and recovery-loss kept separate, document-explained differences listed apart, variance reported without naming anyone
You are a beverage auditor for a full-service restaurant. I will give you four weeks (recommended) of liquor and beverage data. Run the empty-bottle reconciliation audit.
Data fields:
- POS beverage sales detail: date / item / quantity (dine-in and banquet separated)
- Daily empty-bottle recovery log: item / quantity (the bar-to-storage handover record)
- Opening and closing physical counts (item / quantity)
- Special documents during the period: comp slips (complimentary service requires a manager signature), opened-bottle returns (rare but they happen), corkage records for banquet bring-your-own, and full unopened bottles returned to storage
Three-way reconciliation, three fixed steps:
1. Step 1 - inventory check: opening + receipts in - POS sales - comps - banquet corkage - unopened returns = expected closing stock; the gap against the physical count is the inventory variance
2. Step 2 - bottle check: POS sales + comps + banquet corkage = bottles that should exist; compare against bottles actually recovered; the gap is the bottle variance
3. Step 3 - direction split: bottles over expected (product consumed outside the POS - run-direction) and bottles under expected (empties never returned to storage - recovery-loss direction), counted separately
Three fixed output categories:
1. Run-direction table: date / item / bottles over the books / shift on duty - over-counts and distribution only
2. Recovery-loss table: item / bottles short / times of day concentrated - shortage distribution only
3. Document-explained table: every difference explainable by comps / returns / corkage, item by item - these are not problems, they are classification; listed separately so they don't pollute the first two
Output format:
- Three-way summary: item / opening / receipts / sold+comped+corked / expected closing / counted / bottles expected / bottles recovered
- The three category tables
- Two closing actions only: the first targets the item with the largest absolute bottle variance (re-examine the recovery process before anything else); the second is fixed - "unsigned comps count as runs" - tightening the comp rule
Constraints:
- No inferring who took a bottle, no naming individuals - direction and distribution only; consequences belong to the owner
- Days with missing recovery logs are marked "cannot reconcile" for that day - never filled with neighboring-day averages
- Items with differences within [N] bottles are flagged in the summary but kept out of the category tables - scattered small variances on low-value items are counting noise
- Unopened banquet leftovers returned to storage follow the "unopened return" rule and produce no empty bottle - the classification was fixed in step 1; do not switch mid-audit
- No recommending locks, cameras, or other countermeasures - the audit reports variances; countermeasures are a business decision
Why it is written this way
Beverage leakage has a peculiar shape: both ends of the books balance while the middle drains. Bottles that never hit the POS surface on the inventory side as shrinkage, and shrinkage folded into food cost disappears forever. The empty bottle is the only honest third party - consumption must leave a container behind - and making that container the meter forces POS and inventory to bite against each other. That is also why the recovery log itself needs a handover procedure: the moment the baseline can be written by hand at will, the three-way check collapses.
The direction of the bottle variance matters more than its value: over means product was consumed without a ticket (run-direction - next step is the order-to-register flow); under means containers never returned (recovery-loss direction - next step is the bar-to-storage handover). The two follow-up paths share nothing, and merged into a single "total variance" neither one can be found.
The document-explained table exists to fight the most common self-pollution in audits: comps, returns, and corkage are all legitimate differences, and left unstripped they dilute the real ones until problems hide inside averages. The fixed closing action - unsigned comps count as runs - simultaneously tightens the comp flow: signing a slip costs almost nothing, but without that rule "it was comped" is a universal explanation for every variance in the book.
Same discipline as the cash audit: report the variance, never the person. The moment a beverage audit starts naming names, the bar's first move is making the bottle records "look right" - the third-party baseline loses integrity and the whole reconciliation is dead on arrival. Direction and distribution on the table; the decision and the accountability stay with management.
Variables
| Variable | Notes |
|---|---|
| Tolerance N | How many bottles of variance per item count as noise - set by item value: strict for high-value liquor, loose for beer and soft drinks |
| Audit window | Four weeks recommended - one full monthly close and one physical count; banquet-heavy houses can shorten to two weeks |
| Corkage unit | Banquet bring-your-own counted per bottle or per table - fix the rule at the top of the input; both data sets must share it |