Petty cash: how to run a float so it reconciles

Petty cash covers the small payments that cannot practically go through a bank transfer: an auto fare, a courier, tea for a meeting, a replacement bulb. Individually trivial, which is exactly why they get handled informally. Collectively they are frequently the least documented part of a business's spending, and they are cash, which means an unexplained difference has only two possible causes and one of them is uncomfortable.

That asymmetry is the argument for the discipline. With a bank payment there is always an independent record: even with no receipt, the statement proves that money moved, to whom and when. Cash has no such witness. If the box is short and there is no paperwork, nothing distinguishes a genuine purchase whose receipt was lost from money that was taken, and nobody can prove either way. A system exists to protect the person handling the cash at least as much as to protect the business, and that is worth saying to whoever is asked to keep it, because the request otherwise reads as suspicion.

The second reason is that undocumented cash spending is generally not deductible, so a shortfall is not only unexplained, it is also money spent that the business gets no tax benefit from.

The imprest or float system is the standard method and its whole point is that a single number stays constant, which makes checking trivial.

Fix a float, say five thousand rupees, at a level that covers the usual spending between top-ups without leaving a large amount of idle cash lying around. Put that amount in the box. Every withdrawal comes out against a voucher that stays in the box in place of the cash. At any moment, therefore, the cash remaining plus the vouchers held should equal the float exactly. Not approximately, exactly.

When the cash runs low, total the vouchers, and reimburse the box with precisely that total. This restores it to the float and no arithmetic about the balance is required. The vouchers then go to the bookkeeper as the batch of transactions to record, each one against its expense head, and the entry in the books is the reimbursement rather than each individual withdrawal.

The elegance is that the test is one subtraction and it is available at any time without notice. A system where cash is simply topped up when it runs out has no such test, because there is no expected balance to compare against, and without an expected balance there is no such thing as a discrepancy.

A voucher is the substitute for the bank statement that cash does not provide, so it has to carry the facts a statement would: the date, the amount, what it was for, who took the money, and who approved it if that is a different person. If a receipt exists it is stapled to the voucher. If no receipt exists, and for an auto fare or a tip none will, the voucher itself is the record and the description has to do more work. A voucher reading transport, two hundred rupees is weak. One reading auto, warehouse to Kalbadevi, delivery of order 4412 is defensible, because it names something checkable.

The signature matters more than it looks. It converts a claim about spending into a statement by a named person, which is what makes the record evidence rather than an assertion by the business about itself. For the same reason the person who holds the cash should not be the only person who ever signs, since a set of vouchers written and approved by one individual with no second party involved is documentation that only tests whether that person is consistent.

Pre-numbered vouchers are worth the small effort. A missing number in a sequence is visible; a missing loose slip is not, because nobody can count what was never numbered.

Reconciling petty cash is a physical count. Someone opens the box, counts the notes and coins, totals the vouchers, adds the two, and compares the result with the float. Doing this on a fixed schedule matters less than doing it unannounced sometimes, because a check that always happens on the last Friday tests only the last Friday.

When it does not match, the language people use gets in the way. A difference in petty cash is almost never a rounding error, because nothing here is rounded: notes are counted in whole rupees and vouchers are written in whole rupees. A shortfall means one of a small number of specific things. Money was spent and no voucher was written. A voucher was written for the wrong amount. Change from a purchase was not returned to the box. Money was taken. An overage is just as informative and is usually a voucher written for more than was actually spent, or a reimbursement that was miscounted.

So the correct response is to look for the missing document rather than to adjust the balance. Writing off a shortfall as a difference and resetting the float removes the evidence that a process step is being skipped, and if the cause is systematic it will recur silently. Record the difference, investigate it while the week is still fresh in memory, and note what was concluded even if the conclusion is that nobody could establish the cause.

The failures are predictable and worth naming, because each is easier to prevent than to detect.

The float is too large. A generous float reduces the number of top-ups and simultaneously reduces the number of times anybody checks anything, because the reimbursement is the moment the vouchers get totalled. A smaller float that is topped up more often is checked more often.

The box becomes a lending facility. Someone borrows from it and intends to repay, and for a few days the cash is short with a mental note in place of a voucher. This is how the most common shortfalls begin, and it is usually not dishonesty. A rule that the box is never a source of personal advances, stated once and applied without exception, is more effective than any reconciliation.

One person owns the whole cycle. When the same individual holds the cash, writes the vouchers, approves them and performs the count, the reconciliation has no independent element and confirms only internal consistency. Splitting any one of those roles away, even to an owner doing an occasional count personally, restores the test. Finally, top-ups from personal pockets: the owner pays for something out of their own money and the box is never involved, so the expense exists with no record in the cash system at all. That is a separate transaction and it needs recording as one, not absorbing into petty cash later.

A petty cash box that balances proves that the cash present, plus the vouchers held, equals the float. That is a real and useful assurance: it means nothing has left the box without a piece of paper being written for it.

It does not prove that the money was spent on what the vouchers say. A voucher describing an auto fare that was never taken balances exactly as well as one describing a fare that was. Reconciliation checks completeness of documentation, not truthfulness of documentation, and no amount of care in the counting closes that gap. What narrows it is corroboration from outside the box: receipts where they can be obtained, a second signature, descriptions specific enough to be checked against something else such as a delivery that either happened or did not.

This is a general limit rather than a quirk of petty cash. Any system, on paper or in software, can verify that records are internally consistent and flag when an expected document is absent. Whether a described transaction occurred is a question about the world, and the only mechanisms available for it are corroborating evidence and the judgement of somebody who knows the business. Treating a balanced box as proof of honest spending is the specific error to avoid, because it is precisely the conclusion a well-kept fraudulent set of vouchers is designed to produce.