Invoice OCR accuracy: where automation stops
Realistic OCR error rates on crumpled and faded restaurant invoices, line-item pitfalls, and safe human-review fallbacks.
OCR means optical character recognition. Measure accuracy against a manually checked sample of your own invoices. Record the number of correct fields and correct complete lines separately, and name the supplier layout and image condition. The ranges below illustrate a review method; they are not measured vendor performance. Preserve the original image, extracted value and reviewer correction so the result can be checked.
What the data can and cannot tell you
- Treat field accuracy and line accuracy separately; a correct vendor name does not validate a price.
- Expect decimal shifts, merged lines, tax-column confusion, and pack-size loss on poor images.
- Use confidence thresholds plus materiality thresholds, for example review every line over €250.
- Never overwrite the original image when a reviewer corrects extracted data.
A worked example
Illustrative ranges only, invented to compare review needs for three image conditions. They are not measured accuracy, vendor claims or a guaranteed result. Measure your own sample before setting review thresholds.
| Clean PDF | 97–99% fields | Consistent supplier layout |
|---|---|---|
| Phone photo | 90–96% fields | Glare and skew need review |
| Faded/crumpled scan | Below 85% lines | Decimals and quantities are fragile |
| Fallback | Human review | Keep image, confidence, and correction |
Use the calculation only when the source, unit, and exception log are visible. Treat it as illustrative until your own records confirm it.
A repeatable process
- 1Define the question. Write the decision, source, unit, time window, and threshold before opening the spreadsheet.
- 2Reconcile inputs. Match identifiers, quantities, dates, credits, and exceptions. Flag missing evidence instead of guessing.
- 3Calculate the gap. Show formula, baseline, observed result, and illustrative impact. Keep price separate from waste or process variance.
- 4Review and act. Have a manager review material exceptions, agree one action, and set a date to measure it.
FAQ
- What is the most common mistake?
- Comparing numbers that look similar but use different units, dates, specifications, or commercial terms.
- How precise should the result be?
- Precise enough to support a decision, not more precise than the source. Preserve the underlying invoice, image, or count.
- When should a person review it?
- Whenever confidence is low, the amount is material, the item is unusual, or an exception changes the conclusion.
- Are the savings figures guaranteed?
- No. All figures in this guide are illustrative. Validate the formula and opportunity against your own records.
Keep reading
- A 15-minute invoice review that catches what audits miss
Five checks, in order, that surface pricing errors before they reach your books.
- Catch invoice line-price drift before it eats your margin
Compare each invoice line against expected unit prices and act on real variances, not noise.
- Invoice matching issues and how to fix each one
The six ways a three-way match breaks — duplicate invoices, phantom vendors, price drift, quantity gaps, substitution fraud, and statement mismatches.
Browse by category
Want to review your purchasing?
Submit the details and we will confirm which invoices are useful.
Review 3 invoices free