How do I find and fix invoice approval bottlenecks?

For the approval stage specifically, well-run teams clear most invoices in one to three business days; total receipt-to-posting under a week is a strong mid-market result. More useful than any external benchmark is your own distribution: if median approval time is two days but the 90th percentile is three weeks, you don't have a speed problem, you have an exceptions problem - and the long tail is where late fees, missed discounts, and close delays live.

Yes - an SLA converts "please approve promptly" into a measurable expectation and gives escalation a trigger. 48 hours is a common and defensible standard for routine invoices, 24 for urgent ones, with the clock visible to the approver. The SLA only works if three things back it: approvers get full context to act without research, reminders fire automatically, and breaches escalate by rule rather than by AP nagging. Publish per-department SLA performance - visibility does more than enforcement.

Pull aging-by-approver and aging-by-stage. The distribution almost always concentrates: a few approvers, one amount band, or one exception type. Then match the fix to the cause - notification/mobile gaps, too many small invoices routed high, an absent approver with no delegate, or coding disputes disguised as approval delay.

Timestamp every assignment and action, then report dwell time (assignment to action) per approver and aggregate by department. Dwell time is the fair metric - it measures their queue, not upstream delays.

Publish the metric, not the accusation: routine per-department cycle-time reporting, same definition for everyone, visible to department leadership. Pair it with making approval genuinely easy (mobile, full context, reminders) so the remaining slowness is a choice - peer-visible data then does the persuasion.

Use a system that logs notification events, then fix the underlying visibility problem: consolidate notifications into digests if volume is the issue, add mobile and reminder layers, and put pending approvals on a dashboard that doesn't depend on email at all.

Accrue them - that's what accruals are for - and never force-approve to hit a close date; a forced approval is a control failure recorded in your own audit trail. Then fix the process: a pre-close sweep of pending approvals in the final week, with escalation, shrinks the accrual to noise.

Test each step: what does this approver catch that no one else in the chain does? Keep steps with distinct knowledge (budget owner) or distinct authority (DOA threshold); convert visibility-only steps to reporting; let PO-matched, in-tolerance invoices skip redundant re-approval under a governed rule.

In a healthy process, the large majority - think 80%+ - should clear on first assignment with no reminder, reassignment, or question. A high follow-up rate signals routing errors, context gaps, or notification problems rather than diligent review.

Raise the threshold to match today's invoice distribution, delegate a band to a deputy with the executive reviewing a weekly summary, and reserve their live approval for genuinely exceptional spend. Oversight survives - it moves from gate to review, which is usually what the executive actually wants.

Snapshot every pending invoice with current approver, days waiting, amount, and due date; sort by age and flag SLA breaches; send it to finance leadership and each department head with their own slice. The act of distributing it is half the fix.

Make it effortless (mobile, one-tap actions, full context in the request), make it visible (SLA metrics their leadership sees), keep their queues clean (only invoices that genuinely need them), and let them ask questions in the tool instead of opening email threads. Approvers don't resist approving - they resist friction and ambiguity.

Stampli attacks approval latency on both sides: routing assigns the right approver automatically at dispatch, and approvers get everything needed to decide - invoice image, coding, PO context, prior comments - in one workspace, on desktop or mobile, with questions handled in-context instead of email. Reminders and escalation keep queues moving, batch approval clears routine volume, and dashboards show exactly where every invoice sits and how long it has waited, so bottleneck conversations run on data.