How PoP-Eye works
When you upload a PDF, PoP-Eye runs it through five stages. This page tells you what each stage checks — not what numerical weight each carries (we keep that opaque on purpose, so scammers can't reverse-engineer their fakes against us).
1. FormatGate — is this even a PDF a bank would send?
Cheap, fast checks that reject anything obviously not a bank-issued PDF: an image-only file, a corrupt or password-protected document, or a file too large to scan. If a PoP fails the format gate, we don't even run the full scan — we tell you directly to ask the buyer for the bank's original.
2. PdfForensics — what does the PDF's structure say?
Inspects creation tool metadata, modification timestamps, font fingerprints, and document structure. A bank's PoP has a recognisable structural signature; edits or regenerations leave traces a forensic reader can pick up on without depending on visual inspection.
3. BankTemplate — does the layout match the bank's template?
Cross-references the PDF's layout against the known Proof-of-Payment template for the bank the merchant selected — the heading position, the field labels, the typography. PoP-Eye v1.1 supports FNB, Capitec, Standard Bank, and ABSA; we'll add more banks once we've calibrated each template the same way.
4. ContentExtractor — what does the PDF say?
Pulls the beneficiary name, account number, branch code, and amount out of the PDF text. If you've typed your expected values into the cross-check fields, we compare them against what the PDF claims and flag any that don't match.
5. ScoringEngine — what's the verdict?
Combines the signals from the previous four stages into a single verdict: Pass or Fail. We don't show you the internal score — every result, including Pass, carries the same standing recommendation to verify with the bank before releasing goods.
What PoP-Eye does not do
- We do not integrate with any bank's API.
- We do not move money.
- We do not currently scan images, screenshots, or photos — only PDFs.
- We do not run OCR on image-only PDFs. If a PDF has no text layer, the format gate refuses it.
- We do not retain personally identifying information — see Privacy.