Search for accounts payable best practices and you get the same list fifteen times: automate, centralise, standardise, match, measure, control. All true, none ranked, and a good half of it written for a finance department with a dedicated AP clerk and a procurement system.
A four-person business does not have those problems. It has a different set, and copying the enterprise list wastes effort on controls that cost more than the errors they prevent.
So here is the same territory, ordered by what it saves a small team, with the ones worth skipping named as such.
First principle: remove work before you add control
Almost every AP practice falls into one of two groups. Some remove work — fewer touches, fewer places to look, less re-keying. Others add control — approvals, matching, segregation of duties.
Control is not free. Every approval step is a queue, and a queue is where invoices sit doing nothing. Add control to a process that is already slow and you get a process that is slow and annoying, which is how approval steps quietly turn into someone forwarding a PDF and saying "fine by me".
Do the work-removing practices first. They make the control cheap enough to be worth having.
1. One arrival route
The highest-value change in most small finance functions is also the dullest: every invoice arrives in one place.
The usual state is three or four routes — a shared mailbox, someone's personal inbox, paper that comes with a delivery, a supplier portal nobody remembers the login for. Nothing is lost, exactly, but nothing is findable, and the month-end scramble is mostly reconstructing where things went.
One address, and a rule that anything arriving anywhere else gets forwarded there rather than handled where it landed. This costs nothing and removes an entire category of problem.
2. Record the invoice on arrival, not when you get to it
The gap between an invoice arriving and being entered is invisible, which is what makes it expensive. During that gap the liability exists and your ledger does not know about it, so your aged payables report is confidently wrong and any cash forecast built on it inherits the error.
Recording on arrival separates two things that get conflated: knowing about an invoice and approving it. The first should happen immediately. The second can take as long as it takes.
3. Capture the whole document, not just the total
Header-only capture — supplier, date, total, tax — is fast and covers a surprising amount of routine spend. It also makes line-level questions unanswerable later: which cost centre, which project, which of the four items on this invoice was the one that was wrong.
The honest position is that header capture is enough for most invoices and not enough for the ones that matter. Decide per supplier rather than per policy.
4. Match to a purchase order — where you raise them
Three-way matching (invoice, purchase order, delivery note) is the practice most often copy-pasted from enterprise lists into businesses that raise no purchase orders at all. If you do not raise POs, three-way matching is not a control you are missing; it is a control that does not apply.
Where POs exist, matching them is genuinely the strongest control in AP, because it catches quantity and price errors before payment rather than after. Where they do not, the equivalent control is a named person who knows what was ordered confirming that it arrived.
5. Approve by threshold, not by blanket rule
Requiring approval on everything means approvers stop reading. The signature becomes a formality, which is worse than no approval at all because it produces a record suggesting someone checked.
A threshold — anything above a figure that means something at your size, anything from a new supplier, anything without a PO — concentrates attention on the invoices where attention pays. Everything else clears on the strength of the earlier checks. There is more on the mechanics of this in invoice approval workflows for small businesses.
6. Keep the supplier list clean
Duplicate supplier records are how the same invoice gets paid twice, how spend analysis becomes meaningless, and how a payment goes to a bank account somebody changed on the wrong record.
Two habits cover most of it: one person owns supplier creation, and bank detail changes are verified out of band — a phone call to a number you already had, never the number in the email asking for the change. That second one is the single highest-value fraud control in AP, and it costs a phone call. The rest is ordinary master data hygiene.
7. Have an actual duplicate check
Duplicates are not rare and they are rarely fraud. They are the same invoice arriving twice — once from the supplier, once forwarded by the person who received the goods — or a statement entered as an invoice, or an invoice re-sent with a chaser.
Checking supplier plus invoice number plus amount catches most of them. The awkward cases are the ones where the supplier reused a reference or issued a credit and a replacement.
8. Pay on a schedule
Ad-hoc payment is expensive in a way that does not appear anywhere. Every unscheduled payment is a context switch, a separate authorisation, and a separate reconciliation.
A payment run — weekly or fortnightly, whatever suits your volume — turns that into one batched task. The exception worth making is an early settlement discount, where the arithmetic is usually decisive: 2/10 net 30 annualises to roughly 36%, which beats almost any use of the same cash.
9. Reconcile supplier statements
The one routine check that tells you whether your purchase ledger is complete. Everything else in AP verifies the invoices you have; a statement is the only document that reveals the invoice you never received.
It takes minutes per supplier and does not need doing for all of them — the largest few, monthly, catches most of what there is to catch.
10. Measure one number, and make it cycle time
Most AP metrics measure volume, which tells you how busy you were rather than how well the process works. The number worth tracking is the time from an invoice arriving to it being approved and ready to pay.
It is the one figure that responds to every practice above, which makes it a fair test of whether any of this worked. Days payable outstanding is worth knowing too, but it is a cash decision more than a process measure — a rising DPO can mean you negotiated better terms or that your process fell over, and the figure alone will not say which.
What to skip at small scale
- Segregation of duties as a formal matrix. With three people in finance you cannot separate every function, and pretending otherwise produces a policy document nobody follows. Separate the two that matter — whoever creates a supplier should not be the only person who can approve a payment to it — and accept the rest.
- Dynamic discounting and supply chain finance. Real practices, wrong scale. They need volume and supplier appetite that a small business does not have.
- A vendor portal. Suppliers will not log in to it. Email is the interface they already use.
- Invoice accuracy percentages. No figure of this kind means anything without saying which field, on which document type, at what confidence — which is why we don't quote one.
The short version
One inbox. Record on arrival. Approve by exception. Verify bank changes by phone. Pay in batches. Reconcile the big statements. Watch cycle time.
Seven habits, most of them free, and they remove more work than any amount of policy. The software question only becomes interesting once those are in place — at which point it is worth asking what you are still doing by hand.
Cribble covers the first three: one forwarding address for every document, recorded and read on arrival, with every field scored separately so what needs checking is a short list rather than the whole invoice.