Three people, three answers about what the float is. Why an uncertain starting number makes every closing number uncertain, how much change a busy Saturday actually needs, and the five rules that end float arguments.

Ask the manager what the float is. Ask the bar supervisor. Ask the person who opened the till. If you get three answers, then every variance calculation in the venue has been built on a guess, and nobody knows it.

The float is the least interesting thing in a venue and one of the most expensive to get wrong, because it is the baseline every other cash number is measured against.

An uncertain start makes every end uncertain

Float believed to be 200
Float actually 150 (someone took 50 for change last week)
Takings counted at close: 1,340
Variance reported: +50   ← entirely fictional

⚠️ The variance is not a measurement if the baseline is a belief. Every minute spent investigating it is wasted, and worse, a real shortfall of the same size would be invisible.

⇒ The Cash-Up That Takes An Hour

The five rules

One: the float is counted and signed at open

By the person who will be held to it, not by the person handing it over. Thirty seconds.

⭕ A signature turns "I think it was right" into a fact somebody stands behind.

Two: the float is a fixed number, written where people can see it

Not remembered. Not "about two hundred." A specific number, posted.

⇒ The Board In The Dressing Room

Three: any movement between tills is written down at the moment

02:05  50 from till 2 to till 3 — ran out of fives

⚠️ Without the line, two tills are now wrong and each person believes the other is at fault.

Four: the float never funds anything

Not a taxi, not a delivery, not a staff sub, not petty cash. The moment it does, it stops being a baseline.

⭕ If those things need funding, they need their own money and their own record.

Five: the float is reviewed when trade changes, not never

A float set when the venue was quieter runs out of change on the new busiest night, every week, forever, because nobody owns the number.

How much change a busy night actually needs

Most floats are set once and never revisited. A rough way to sanity-check yours:

Cash proportion of takings (%)  × expected takings
  ÷ average cash transaction
  = number of cash transactions

Then ask how many of those need change, and in what denominations. The answer usually shows two things:

- The total is roughly right
- The mix is wrong — too many notes, not enough coin and small notes

⚠️ Running out of fives at midnight is not a float-size problem. It is a float-mix problem, and adding more money does not fix it.

The Saturday test

For two Saturdays, note the time anyone has to go looking for change. If it is the same hour both weeks, the mix is wrong at that hour and you can fix it with the same money, differently arranged.

What actually goes wrong

The float that drifts

Nobody takes anything deliberately. Small roundings, a fiver borrowed and repaid in coin, a night where the count was rushed. Over months the number moves.

⭕ Counting at open catches drift within a day instead of within a quarter.

The float that gets raided at the busy hour

The bar runs out of change at midnight and takes from another till. Both tills are now wrong, and the person who did it went home at two.

⇒ This is rule three, and it is the single most common cause of an unexplainable variance.

The float nobody owns

If the float belongs to everybody, it belongs to nobody. One named person per till per shift.

⇒ Your Busiest Person Is A Single Point Of Failure

The float used as petty cash

A delivery arrives and needs paying. Someone takes it from the till. The record, if it exists at all, is a scrap of paper.

⚠️ This is rule four, and it is also where cash handling starts to become an accounting problem rather than an operational one.

⚠️ The accounting side is not ours to rule on

How cash movements must be recorded, what documentation a payment out of a till requires, and how this interacts with your books and tax obligations are questions for your accountant. This page covers only the operational habits. Take the recording requirements to a qualified accountant before designing anything around them.

Setting the float properly, once

Step one: count what is actually there

Not what it should be. What it is, tonight, in every till and in the safe.

Step two: decide the number and the mix

Total per till
Breakdown by denomination
Who signs for it
Where the number is posted

Step three: run it for two weeks and look at the variance pattern

Variance now small and random   — the baseline is real
Variance still one-directional  — something structural remains

⇒ The Number You Check First

Step four: put a review date on it

A date, in a diary, when someone looks at whether the number and mix still fit the trade.

⚠️ Without a date, "review when trade changes" means never, because nobody notices trade changing.

⇒ Your Crowd Changed Three Months Ago

What this looks like when it is working

The three-people test

Ask three people what the float is. One answer is the pass condition.

The change test

Count how many times in a night someone goes looking for change. Zero or one is fine. Four means the mix is wrong.

The variance test

Working     — variances are small and change direction
Not working — variances are small and always the same direction

⚠️ A consistently-directional small variance is the most ignored signal in venue cash handling, because each individual night looks fine.

Three to hold

1  Counted and signed at open, by the person held to it
2  Movements between tills written at the moment, not remembered
3  The float funds nothing — not taxis, not deliveries, not subs

Where the record has to sit

The float count, the movements, and the closing variance belong in the same record, because the only useful question — why was this night different — needs all three at once.

⇒ The Month End That Takes Two Weeks

The Float You Start The Night With