Running a monthly accuracy audit (15 minutes)

The point of an audit isn’t bureaucracy - it’s spotting regressions before they become invoices you pay twice. Run this on the 1st of every month.

The 4-step audit

1. Pull the report

In OrderPilot, go to Reports › Accuracy and set the window to “Last 30 days.” Export to CSV.

2. Check the headline numbers

You’re looking for three metrics:

• Document-level accuracy - target ≥99 %
• Field-level accuracy - target ≥99.5 %
• Auto-approved rate - target ≥95 %

If any number is more than 0.5 pp below target, dig deeper.

3. Segment by supplier

Sort the report by supplier, descending by error count. The distribution should be long-tailed: one or two suppliers with 3–5 errors, most with zero. If any supplier has >10 errors in the month, that’s your investigation for the month.

Common causes:

• New template (check the supplier page for “Template changed” flags)
• Supplier changed their PO numbering scheme
• Supplier started sending POs via a different email address

4. Segment by field

Still in the report, switch the grouping to “By field.” Look for fields with >2 % error rate. Typical culprits:

• delivery_date - often ambiguous when the supplier writes “week 23”
• unit_price - currency or UoM confusion
• line_notes - unstructured text; consider whether you actually need this

For each field with drift, ask: is this worth a supplier conversation, a rule, or is the current rate acceptable?

Close the loop

Write two sentences in your shared doc:

• What changed: one fix you made
• What to watch: the thing you’ll re-check next month

The whole audit should take 15 minutes. Teams that skip it for 2–3 months let drift compound and end up with a half-day cleanup instead.

Escalation triggers

Escalate to OrderPilot support if you see:

• Document-level accuracy below 97 % for two consecutive weeks
• Any supplier with a sudden step-change in error rate (was 0 %, is now >5 %)
• A field you can’t correct in the UI (rare, but means a schema issue)