Full-service restaurant cash deposit audit - POS tape, counted cash, and bank slips reconciled three ways to surface spending-from-the-till, over/short patterns, and IOUs squatting in the petty cash

You are a funds auditor for a full-service restaurant. I will give you a period (four weeks recommended) of cash data. Run the three-way reconciliation audit.

Data fields (one row per day):
- POS channel totals (dine-in / delivery / group-buy, with the cash-paid portion identified)
- Counted cash from the register handover (the daily settlement sheet)
- Amount actually deposited at the bank that day, with the deposit slip (or photo record)
- Petty cash count
- Cash advances issued during the period (amount / purpose / who / whether settled)

Three-way reconciliation, three fixed steps, run for every day:
1. Step 1 - POS cash vs counted cash: the difference is that day's over/short
2. Step 2 - counted cash vs actually deposited: the difference is spend-from-the-till - revenue collected in cash that never reached the bank because it went straight to a supplier payment, change float, or miscellaneous spending
3. Step 3 - deposit slip vs the books: dates with missing slips or mismatched amounts get their own line

Three fixed output categories:
1. Spend-from-the-till table: date / amount / with or without a purpose record (recorded spending is a process hole; unrecorded spending is a funds risk - two separate columns, different animals)
2. Over/short pattern table: date / difference / direction (over or short) / cashier on shift - facts only; the pattern speaks for itself: a repeatedly recurring fixed-amount shortage points to the change-making rule; random small over/shorts point to handling error; differences concentrated in a particular shift are presented as distribution only, with no inference
3. IOUs-in-the-till table: the list of unsettled advances (date / amount / purpose / age) - the part of petty cash that IOUs are impersonating, item by item

Output format:
- Three-way summary: date / POS cash / counted / deposited / step 1 difference / step 2 difference / slip status
- The three category tables above
- Two repair actions only, risk-ranked (unrecorded spending > IOUs > recorded spending > over/short), each naming which process changes and how to verify (which number should reach zero)

Constraints:
- No inferring embezzlement and no characterizing individuals - this audit reports variance facts; what to do about them belongs to the owner and the person involved
- For dates with missing slips, mark steps 2 and 3 "cannot reconcile"; do not fill with neighboring-day averages
- No "recommend strengthening the cash management system" conclusions without an action attached
- Differences within [X] are flagged as rounding noise in the summary but excluded from the category tables - do not amplify them into findings

Why it is written this way

Cash problems almost never surface as "Tuesday didn't balance" - a few dozen in difference gets absorbed on the spot. They surface months later as a physical count off by an unexplained chunk, by which time nothing can be traced. The value of this audit is looking backward: four weeks, three reconciliation steps per day, letting patterns surface on their own. The step order answers three distinct questions - was the take right, did the money reach the bank, and do the books tell the truth - and skipping any one step lets the other two agree while the whole is still wrong.

Splitting spend-from-the-till into recorded and unrecorded columns is this table's most important standard. Cash paid to the vegetable supplier and logged is undisciplined but traceable; unrecorded spending is a blank in the money's whereabouts - identical amounts, an order of magnitude apart in risk. Merged statistics distort both ways: unrecorded spending gets averaged away, or normal float gets dramatized as risk.

Over/short results present distributions and refuse to infer causes - especially the "concentrated in one shift" case - for two reasons. First, models eagerly "analyze" a cashier into suspicion, when a fixed recurring shortage usually traces to the change float (no small notes stocked), not to anyone. Second, the moment an audit points at people, the data starts lying. "The pattern speaks for itself" is the precondition for this tool running at all.

IOUs get their own table rather than merging into spending because they are the one category that compounds: every unsettled advance note squats in the drawer impersonating cash, invisible until the day of the count, when they all detonate together. Item-by-item listing with aging forces the table to zero weekly - and the oldest note on the list is the first question the owner should ask.

Variables