Petty Cash on a Construction Site: Surviving an Audit

Most contractors can tell you their bank balance to the halala. Ask the same company how much of its cash is sitting in staff pockets, glove boxes and site office drawers right now, and there is usually a pause.

That second number is what this article is about.

Petty cash is bigger than the name suggests

The word "petty" does the damage. It suggests small, and small suggests unimportant.

The individual amounts are small; the flow is not. Diesel for the generator, a taxi to fetch a part, water and ice for the crew, a grinder disc, a municipality fee paid over a counter. Across three or four sites over a month that is a real cost line, and it usually reaches the accounts as one lump labelled "site expenses".

Two things make it awkward. The money leaves the bank in round advances, so the bank statement says nothing about what was bought. And the spending is done by people whose job is building, not bookkeeping, at the moment they are busiest.

None of that is a fraud problem. It is a records problem, and those are solvable.

The imprest system, explained plainly

You decide a fixed float for each holder — say 5,000 riyals for a site supervisor. That figure is the imprest amount. The holder spends from it and keeps the receipts. When the cash runs low they present the receipts, finance reimburses exactly what was spent, and the float returns to 5,000.

The important property: at any moment, cash in hand plus receipts not yet reimbursed should equal the imprest amount. Always. That one equation is the entire control. If the supervisor has 1,850 riyals and 3,150 in receipts, the float is intact. If he has 1,850 and 2,900 in receipts, 250 is unaccounted for — and you know this week rather than at year end.

Compare that with ad-hoc issuing, which is what most companies actually do.

Ad-hoc issuing is not dishonest, it is untestable. There is no figure the balance is supposed to return to, so every advance is a fresh open-ended trust exercise. Moving to imprest costs nothing: set the float, agree it in writing with the holder, and stop issuing cash except as reimbursement of documented spending.

Size the float at two to three weeks of normal spending. Too small and you spend your life processing reimbursements; too large and you are financing a private bank account. Review it when the site changes phase — a float sized for excavation is wrong for fit-out.

Custody: one named person per float

A float belongs to one named human being. Not "the site office". Not "the engineers". One person, who signs for it.

The common failure is a drawer with three keys. Everyone can take cash, everyone can put a receipt back, and when it is short nobody is responsible because nobody ever was. Shared custody is the same as no custody.

Which leads to the number most companies cannot produce on request: how much cash is out with staff right now. The figure exists — issued, minus documented spending, minus returned, per holder, added up. Nobody maintains it because doing so by hand is tedious and never quite urgent. Maintained automatically it becomes an ordinary management number, and it becomes obvious when one holder's balance has been creeping upward since April.

Handovers deserve the same discipline. When a holder goes on leave or leaves, the float is counted, reconciled, signed for and formally transferred. Many petty cash problems begin with a float that changed hands informally.

Segregation of duties when there are four of you

The textbook says the person who authorises, the person who holds the cash and the person who records the transaction should be three different people. In a company with a finance manager, one accountant and two site staff, those three people do not exist.

The practical compromise is to stop pretending and protect the controls that carry weight:

• Separate custody from recording where you can. The holder spends; someone in the office enters it. Where one person does both, the review must be by somebody else.
• Make the review substantive. Someone other than the holder counts the cash, unannounced, at least sometimes. A count that never happens is not a control, it is a policy document.
• Approve by exception. Above a set limit, prior approval; below it, the receipt is the approval. Requiring a signature on every 30-riyal taxi fare teaches everyone to sign without reading.
• Give the owner the summary. In a small company the strongest control available is that the general manager sees the custody balances monthly and asks about anything odd.

One boundary is worth defending. If an item is large enough to warrant a purchase order, a delivery note and a supplier account, it belongs in a project control system, not a cash drawer. Petty cash creep is how a working control system quietly stops working.

Evidence: what to keep for each expense

For every payment out of the float, record the date, the amount, what it was for in words a stranger could understand ("diesel, generator, Site B" rather than "fuel"), the site or cost code, who spent it, and the supporting document.

Supporting document means the original receipt or invoice. Thermal receipts fade to blank paper within months, so scan or photograph them the day they arrive rather than the day the auditor asks. Where a supplier will issue a proper tax invoice, take it — reclaimable input VAT needs an invoice that meets the requirements, not a handwritten slip. Our guide to Saudi VAT invoice requirements sets out what must appear on one.

Where no receipt genuinely exists — a car park with no machine, a porter — use a voucher signed by the spender and countersigned by the supervisor. Keep these rare. A float where a fifth of the spending has no third-party evidence has stopped being a record of anything.

How long to keep it all is a question for your accountant. Retention periods in Saudi Arabia depend on the record type and your tax position, and it is not a number worth guessing at.

Why the ledger must not allow deletion

The instinct when setting up a cash record is to let people fix mistakes by deleting the wrong line. It feels helpful. It also removes the one property that makes a cash record worth trusting: that nothing quietly disappears.

Correct by reversing entry instead. The original stays. A second entry cancels it, dated today, with a reason. Then the correct entry goes in. Three lines where a spreadsheet user would have edited one — and the history survives. Administrators may still need full control; the ordinary daily user should not have it.

Month-end reconciliation as a routine

Count the cash. List the receipts not yet reimbursed and total them. Add the two, compare against the imprest amount, explain any difference in writing. Reimburse, sign the statement, file it. Twenty minutes per float, monthly at worst and fortnightly on an active site.

The value is in it being dull and unmissable. Done every month, it finds a 200-riyal discrepancy while people still remember the week. Done once a year, it finds a discrepancy of several thousand that nobody can reconstruct, and the only response left is to write it off.

Warning signs

None of these proves anything alone. Each is a reason to count the cash this week.

• A balance that only ever goes up. A healthy float oscillates. One that ratchets upward means advances are being issued without reimbursement discipline.
• Receipts arriving weeks late, in batches. Documentation collected as it happens looks different from documentation reconstructed afterwards.
• Round numbers. Real spending is 87.50 and 213.75. A run of clean 500s and 1,000s is estimates, not receipts.
• Requests that sit consistently just below the approval threshold.
• A float never counted by anyone except the person holding it.
• Reimbursements growing month on month while site activity does not.

If you would rather not maintain this in a spreadsheet, our petty cash management software tracks issues, custody balances per person, spending by site and printable statements for signature, and staff accounts cannot delete entries. It is free for 7 days without a card, then a paid subscription.