The Cash Room: Reconciliation That Actually Reconciles

Every cash-heavy venue has a number it has learned to live with. "We're usually out by a bit." "A hundred or so, most nights." "It evens out."

It does not even out. A variance that averages near zero is not the same as a variance that is small, and neither is the same as one you understand. A room can be short and over on alternate nights, average to nothing, and still be losing steadily — because the two are not the same money.

This is how to make the count mean something.

Start with what "balanced" means

A till balances when:

Opening float + cash sales + cash tips in − payouts − drops = closing count

Every term in that line needs a definition your staff agree on, and most rooms have at least one they do not:

Term	The question that decides it
Float	Counted at open by whom, and witnessed?
Cash sales	Does the POS know which tabs closed in cash?
Tips in	Do cash tips enter the drawer at all, or go straight to a jar?
Payouts	Taxi money, a supplier at the door, a staff advance — are they slipped or logged?
Drops	Timed and witnessed, or whenever the drawer feels full?

The most common single cause of unexplained variance is payouts. A manager hands a driver cash at 2am and intends to write it up later. The night gets busy. The note never happens, and by 5am the drawer is short by an amount nobody can name.

Two numbers, not one

Recording a single net variance destroys the information you need.

Over  = closing count − expected, when positive
Short = expected − closing count, when positive

Track them separately, per shift, per person. A room whose variance nets to zero because it is +200 one night and −200 the next does not have a small problem — it has two problems that happen to cancel in the arithmetic.

What the shapes mean:

• Consistently short, small amounts → usually process. Unlogged payouts, change errors under pressure, tips handled inconsistently.
• Consistently over, small amounts → also process, usually the mirror: over-collecting change, or tips landing in the drawer that were not counted as sales.
• Occasionally short, large amounts → a specific event. Find the night, not the pattern.
• Short only on specific shifts or people → the conversation nobody wants, and the reason you track per-person.

Only the last requires suspicion. The first two are almost always fixable with a form.

Variance per volume, not per night

An absolute figure is unreadable across different nights.

Variance rate = |variance| ÷ cash sales for that shift

A $150 shortfall on a $2,000 night and on a $20,000 night are not comparable events, and comparing them as raw dollars will point you at your busiest nights every time — not because they are worse, but because they are bigger.

Then set your own threshold from your own trailing weeks, not from a rule of thumb. Whatever your normal band turns out to be, the useful action is the same: investigate outside the band, ignore inside it. A room that investigates every variance stops investigating any of them within a month.

Drops: the control that costs nothing

A drop moves cash from the drawer to the safe mid-shift. It does two things:

One, it caps exposure. Whatever is in the drawer is what is at risk from a single event.

Two — and this is the part rooms miss — it segments the night. If you drop at fixed times and record the count, a shortfall can be located to a window rather than to the whole shift.

Without drops: "we were short $300 tonight"
With drops:    "we were short $300 between 1am and 2am"

The second is actionable. The first is not. You can look at what happened in that hour — who was on, what the queue looked like, whether a payout happened — and often the answer becomes obvious.

Fixed times beat "when it feels full." A drop schedule that varies with the manager's attention produces windows of varying length, and a variance in a four-hour window tells you nothing.

The two-person count

The count at close should be done by two people who both sign it.

This is not primarily about theft, and framing it that way makes it harder to introduce. It is about arithmetic. Counting several thousand in mixed notes at 5am after a ten-hour shift produces errors in both directions, and a second pair of eyes catches most of them.

It also protects the person counting. A single-person count means any discrepancy attaches to one name with no way to distinguish a miscount from anything else. Staff generally welcome the second signature once that is explained.

What a persistent shortfall is usually telling you

In our experience the order of likelihood is:

1. An unlogged payout process — the single most common, and fixable with a numbered pad by the drawer
2. Tips handled inconsistently — sometimes in the drawer, sometimes not, so "expected" is wrong rather than the count
3. Change errors under pressure — real, small, and reduced by removing cash from the door
4. Float not counted at open — you cannot balance against a number you assumed
5. Deliberate loss — real, but rarer than the first four, and you will not find it before eliminating them

Work them in that order. Rooms that start at five damage trust while leaving the actual cause in place.

What to record

float_counted_at_open (and by whom)
drops (time, amount, witnessed by)
payouts (time, amount, reason, authorised by)
closing_count (by whom, countersigned)
cash_sales (from POS)

Five records. Four of them are pieces of paper. Only the last needs a system, and you already have it.

The system we built

tasteck is a booking and analytics system for night venues, built by people who ran them for sixteen years and grew the business from ¥200 million to ¥1.2 billion a year — six-fold, by attacking it with systems rather than by selling harder.

What it does here: settlement and payouts, staff shift records so a variance can be attached to who was actually on, guest records that persist across visits, bookings and dispatch, and analytics across accounting and staff performance.

What it does not do: it does not manage a cash drawer. There is no float count, no drop log, no till reconciliation screen. Those five records above stay on paper or in your POS. If drawer-level cash control is your main pain, we are not the answer to that specific problem, and it is better you know now.

Where it does bear on the money

Reconciliation protects revenue you already took. The other half is what you spend to create it, and there tasteck outputs something nothing else in the nightlife category does:

The maximum you can spend on each marketing channel next month, as an amount in your currency.

Lifetime value of the guests each source actually delivered, times your target margin. Not a chart — a figure you take into the renewal conversation.

Ask it from ChatGPT

tasteck connects to ChatGPT over MCP: ask your numbers as a question and the answer comes back in the chat — settlement by period, which staff worked which shifts, how last month split by source.

Measured against every vendor listed on Japan's principal nightlife-industry directory, this is the first implementation of it in the category, and the same interface is callable from anywhere rather than being tied to one assistant.

Multi-language is built in, the operating surface itself, with your language set put in place during onboarding.

From $34 a month for up to two venues. Thirty days free on every plan, cancel any time. → Pricing

Start tonight

1. Put a numbered pad by the drawer for payouts. Time, amount, reason, signature. This alone resolves most unexplained variance in most rooms.
2. Count the float at open and have it witnessed. You cannot balance against an assumption.
3. Drop at fixed times, and write the count. It turns a nightly shortfall into an hourly one.
4. Record over and short separately, per shift, per person.
5. Compute variance as a rate, then set your band from your own trailing weeks.

The channel-spend ceiling calculation is open on our site with nothing to sign up for: Ad budget calculator. Nothing is transmitted anywhere — use it and close the tab.

Read next

• Four Questions Your POS Cannot Answer
• Nightclub KPI Dashboard: What Belongs On It and What Is Just Noise
• Bottle Service Math: The Table Numbers Most Nightclubs Never Run

---

On benchmarks. No acceptable-variance figure appears in this guide. We do not have a dataset broad enough to publish one, and it varies with volume, staffing, and how much of the night runs on cash. A rule of thumb from another room is worse than four weeks of your own numbers — and four weeks is all it takes.