Clean digital baseline
A native or high-quality digital PDF. It has a clear header and multiple line items.
The baseline. Every invoice parser should handle this before the documents get difficult.
Controlled test · Invaritech owns one tool
We are testing 8 tools to find the best OCR software for invoice processing. Each tool gets the same clean PDF, handwritten invoice, and older scan. Results are pending.
Verdict
No verdict yet
We have locked the test method. We will publish results only after all 8 tools have processed the same 3 invoices and we have checked every output against manually reviewed ground-truth CSV files.
Protocol locked
2026-07-13
Three invoices cannot prove universal accuracy. They can show whether a tool handles the easy case. They can also show how it fails on handwriting and an older scan.
A native or high-quality digital PDF. It has a clear header and multiple line items.
The baseline. Every invoice parser should handle this before the documents get difficult.
A real handwritten invoice. Labels, numbers, spacing, and table structure vary across the page.
The first limit test. It checks whether a tool reads handwriting without inventing fields or rows.
An older invoice with faded print, an unusual layout, and scan artefacts.
The second limit test. It checks what survives when contrast and document structure have degraded.
We have not run the full controlled test. Every row stays blank until we check its output against the same ground-truth files. We will record signup, price barriers, exports, processing time, and published file-retention claims on the test date.
| Tool | Digital | Handwritten | Older | Status |
|---|---|---|---|---|
01 Invaritech Invoice Extractor Owned by publisher | Not tested | Not tested | Not tested | Awaiting controlled test |
| Not tested | Not tested | Not tested | Awaiting controlled test | |
| Not tested | Not tested | Not tested | Awaiting controlled test | |
04 | Not tested | Not tested | Not tested | Awaiting controlled test |
| Not tested | Not tested | Not tested | Awaiting controlled test | |
06 | Not tested | Not tested | Not tested | Awaiting controlled test |
07 | Not tested | Not tested | Not tested | Awaiting controlled test |
| Not tested | Not tested | Not tested | Awaiting controlled test |
We normalise harmless formatting differences. We count wrong values, missing rows, merged items, and invented data as failures. We will not hide those failures inside one weighted winner score.
The 2 published accuracy measures
header accuracy = correct expected fields / expected fields
line-item accuracy = correct expected cells / expected cells
Disclosure: Invaritech owns one of the 8 tools in this comparison. We will score it with the same files and rules. We will publish its misses, product limits, and any access problems.
No vendor pays to appear. We use no affiliate links. We will tie every score and verdict to the published output.
This is a small, difficult test. It covers 3 disclosed documents. It does not cover every language, industry, layout, or paid product configuration.
Use the line-item review checklistWe published the test method, 8-tool list, 3 invoice classes, ownership disclosure, and scoring rules. We have not published extraction results.
Protocol releaseTest the document you care about
Upload 1 single-page PDF, JPG, or PNG. Check the header fields and line items against the source. Download both CSV files without creating an account.