Methodology

Inside the WatchRadar dossier.

A walk-through of what happens between the moment you finish the five-angle scan and the moment you read the verdict. None of this is magic. All of it is auditable. The point of publishing the methodology is so the verdict is something you can defend, not something you have to trust on faith.

Step 1 — Reference identification

The first thing the dossier does is figure out what the watch is. Five photos go in; a candidate reference comes out, with a confidence score and the runner-up references that were close in the matching. The reference catalogue covers the major luxury brands — Rolex, Audemars Piguet, Patek Philippe, Omega, Cartier, Jaeger-LeCoultre, IWC, Vacheron Constantin, Tudor — at the production-year level for high-volume references. Reference identification is the foundation: every subsequent check uses the candidate reference's expected pattern as the comparison.

Step 2 — Region-by-region scoring

Each watch is divided into inspection regions: dial typography, applied logo, hour markers, hands, bezel and pearl, crown profile, case finishing, lug-to-bracelet transition, bracelet construction (if applicable), and movement (where visible through display backs). Each region is scored independently against the expected pattern for the candidate reference. The scores are surfaced in the inspection diagram with the watch photo annotated — green checkmarks where the region passed cleanly, amber where the score was borderline, red where it failed.

Step 3 — OCR cross-checks

Engraved text on the watch — dial signatures, rehaut engraving (Rolex), case-back engraving, bezel numerals, movement bridges — gets OCR'd and cross-checked against the format expected for the production year. A 2024 Submariner with a 2010-format serial is suspicious; a Speedmaster Master Chronometer with a missing METAS marking is suspicious; a Royal Oak with a misspelled "AUDEMARS PIGUET" rotor engraving is unambiguously fake. The OCR module returns confidence scores per character and flags formats that do not match the catalogue.

Step 4 — Market-value estimate

In parallel with the verdict, the dossier produces a market-value band: low, typical, high, and an asking-price fairness verdict if an asking price was provided. The valuation pulls from current secondary-market data across major dealers and auction houses, weighted toward recent sold-listings. The fairness verdict is computed at the time the dossier runs — Fair (asking inside the typical band), High (above typical), Low (below low). The market value is rendered alongside the authenticity verdict because authentic-but-overpriced is the third major buyer risk and the only one most authentication tools ignore.

Step 5 — Final verdict

The region scores, OCR cross-checks, and reference confidence are combined into a final verdict: Likely Authentic (high confidence in the candidate reference and uniformly clean region scores), Uncertain (mixed signals — some regions clean, some borderline), or High Risk (region scores include unambiguous failures, or the reference identification itself was low-confidence). The verdict is paired with a recommended action: proceed, ask for additional photos, or escalate to a watchmaker.

What the dossier does not do

Run your first scan in under two minutes.

Free on the App Store. iPhone only.

Download on the App Store