Invoice processing: from arrival to approved in a fixed number of steps

Invoice processing is the sequence a supplier invoice goes through between arriving and being ready to pay: capture, coding, matching, query resolution and approval. It is distinct from paying — the payment run is the easy part — and it is where the time actually goes, mostly in chasing people rather than in doing anything.

The characteristic small-company version is that invoices arrive at three different addresses, sit in personal inboxes, get forwarded for approval, and are found again when the supplier chases. Every step of that is avoidable, and none of the fixes require software.

The steps worth defining

• Intake: one address or one form, published to suppliers, so every invoice enters the same way.
• Capture: supplier, invoice number, date, net and tax amounts, due date, PO reference, and what it relates to.
• Coding: which account and which cost centre or project. Doing this at capture rather than at month end is what makes the books usable.
• Matching: against the purchase order and the record of receipt, above your threshold.
• Query handling: a named owner and a state, because a queried invoice with no owner simply stops.
• Approval: by role, with limits, and by someone other than whoever ordered the goods.
• A status per invoice that anyone can see — received, coded, awaiting approval, queried, approved, paid.
• A target turnaround, so "how long does approval take" has an answer rather than an impression.

The bottleneck is almost never data entry. It is approval — one person travelling, one invoice queried and forgotten, one email nobody replied to. Measure the elapsed time from arrival to approval for a month and the shape of your problem becomes obvious; most companies discover that half the total time sits in two or three approvers' inboxes.

Making it fast

1. Publish one intake route and tell suppliers, on the purchase order and in the terms.
2. Capture the same fields every time, at arrival rather than in a monthly batch.
3. Code at capture, using the same account list you report on.
4. Set an approval threshold so most invoices need one approval, not three.
5. Give every approver a deputy, and use them. A single named approver is a single point of failure with a holiday allowance.
6. Batch approvals — a twice-weekly review of a list is faster and more accurate than an interrupt per invoice.
7. Give queried invoices an owner and a date, and review the queried list weekly. This list is where money and goodwill both go missing.
8. Publish a target — for example, approved within five working days of arrival — and measure against it.
9. Reconcile supplier statements monthly to catch invoices that never entered the process at all.

Automation, honestly

There is a real threshold at which AP automation earns its cost — invoice capture, automatic coding, routing by value, and matching without a person. It is not at thirty invoices a month, and the sales material rarely says so. Below that volume the wins come from a single intake point, coding at capture, approval batching and a named owner for queries, all of which are free. Above a few hundred invoices a month, buy the software; the arithmetic changes decisively.

In Ettex, the intake can be a form in Ettex Forms — supplier, invoice number, amount, dates, PO reference and a file upload for the invoice itself — with every submission landing timestamped in one searchable inbox, a live summary of what has arrived, and export to CSV or XLS. Because the inbox is shared, an invoice does not stall because one person is away. The coding and posting happen in Ettex Books with a chart of accounts, categories and auto-categorisation rules, attachments on any entry and bank reconciliation; supplier records and terms sit in Ettex Contacts; approvals and queries can be tracked as cards in Ettex Board; and the policy with its thresholds belongs in Ettex Docs.

Said plainly: Ettex does not process invoices for you. There is no OCR or data extraction from an uploaded PDF, no automatic coding, no approval routing by amount, no three-way matching, no supplier portal and no reminders when an invoice has been sitting unapproved. The form collects, Books records, and people do the approving. Below a hundred invoices a month that is proportionate; above it, AP automation is the right purchase.

Where invoices get stuck

• Multiple intake addresses, so nobody can say what has arrived.
• Coding deferred to month end, which makes the ledger unusable in the meantime and slows the close.
• One approver with no deputy.
• Approvals requested by message, where they compete with everything else in the thread.
• Queried invoices with no owner — the single largest cause of very old payables.
• No visible status, so the answer to "where is that invoice" is a search.
• No turnaround target, so slow approval is invisible.
• Supplier statements never reconciled, so invoices that never arrived are discovered by a chase.

Invoice processing is a queue problem. One intake route, code at capture, one approver with a deputy, an owner for every query, and a published turnaround — automation only pays once volume makes the manual version the bottleneck.