Ask why an invoice needs three approvals and you will usually be told about a thing that happened once, several years ago, to somebody who has since left. The chain is a scar. It is not wrong, exactly — it is just sized for an event that occurs twice a year and paid for on every invoice.
Design for the median, escalate the tail
The useful question is not “who should approve this?” but “what proportion of these does anybody actually change?”. For most payables the answer is under three percent. A chain that stops ninety-seven percent of invoices to confirm they are fine is not a control; it is a tax with a compliance justification.
Route on the things that predict an exception — a new supplier, an amount outside the supplier’s own range, a missing purchase order, a category that does not match the supplier’s history — and let everything else through with a record rather than a stop.
The chain has to finish before the period does
This is the part that actually costs days. An approval sitting in someone’s inbox on the thirty-first is not a governance win, it is an accrual and a conversation. Two changes fix most of it:
- Deadlines relative to the period, not to the invoice. “Before cut-off” is actionable; “within five days” is not, because the fifth day may be the second of next month.
- One reminder that names the consequence. People do not respond to “you have 3 pending items”. They respond to “these two will land in next month’s numbers.”
Fewer approvers, not weaker approval
Every chain I have shortened has ended up with stronger controls, because the approvals that remain are read. Three people rubber-stamping in sequence is not three times the scrutiny of one person who knows the invoice will not be reviewed by anyone else.
What to measure afterwards
Not the number of approvals. Measure the share of invoices that were changed or rejected at each step, and the number still open at cut-off. If a step never changes anything and is never the one that is late, it is a notification wearing a control’s costume.
