Our Methodology
Here's exactly how we test every tool and arrive at our scores.
Our Test Document Set
We run every tool through the same set of 100 real documents:
- 50 invoices — from 30+ different vendors, varying formats, languages, and quality levels
- 20 receipts — retail, restaurant, and service receipts (digital and photographed)
- 10 bills of lading — shipping documents with complex table structures
- 10 bank statements — multiple banks, varying layouts
- 10 miscellaneous — contracts, forms, medical documents
The mix includes clean digital PDFs, scanned documents at various DPI levels, and phone photos. Basically, the kind of stuff you actually deal with at work.
Scoring Rubric
Each tool is scored on six dimensions with the following weights:
- Accuracy (25%) — Field-level extraction accuracy on our test set.
- Ease of Use (20%) — Time to first extraction, learning curve, documentation quality.
- Pricing Value (20%) — Cost per page at three volume tiers (100, 1,000, 10,000 pages/month).
- Integration Depth (15%) — API quality, native integrations, webhook support.
- Document Versatility (10%) — Range of document types and multi-language capability.
- Support & Docs (10%) — Response time, support channels, knowledge base quality.
Review Updates
We re-test every tool every quarter. If a tool ships a big update, we re-test it sooner. Each review page shows when it was last updated.
Affiliate Disclosure
Some links on this site are affiliate links, meaning we get a small commission if you sign up through them. This has zero effect on our scores. Our top-rated tool (Rossum) isn't even an affiliate partner.
Editorial Independence
We base our reviews on our own testing. We don't accept payment for reviews and we don't guarantee placements.