Glossary

Every label and score on GrailWatch, in plain English. If anything here is unclear, that's a bug — let us know via the acceptable-use channel.

Landed cost: Total cost to get the watch in your hand: sticker + duties + brokerage + VAT + any buyer-paid third-party verification fee. GrailWatch always scores against landed cost, never the ad price, because the ad price isn't what you pay.
Fair-value band: The 20th- to 80th-percentile of comparable verified sales, normalized to excellent / full-set baseline and then scaled to the listing's actual condition and box & papers. The median sits in the middle.
Deal tier: Where the listing's landed cost falls relative to the band median: great (≥10% below), good (3–10% below), fair (within ±7%), or overpriced (>7% above). We never give precise dollar percentages — bands and qualitative tiers carry more honest signal.
Confidence: How much we trust the band itself. High = many recent verified comps with tight spread. Low = thin comp set or wide spread; the band is shown but not trusted enough to call deal tier confidently.
Trust (platform / listing / seller): A three-layer composite. Platform: how protective the marketplace is. Listing: integrity signals about the ad itself (stock photos, reference consistency, etc.). Seller: account age, feedback %, off-platform-payment requests. Each layer is graded independently; the composite never overrides a fundamentally failing layer.
Risk level (clear / caution / elevated / high): A graded read of how cautious to be. Built from factual, non-accusatory signals (e.g. 'seller account is 12 days old'). GrailWatch never labels anyone a fraud or scammer — the risk scale stays graded so a single weak signal doesn't get amplified into an accusation.
Service state: Recent-service / service-needed / undisclosed / unserviced-vintage. Material to the buy decision and to the valuation factor (a recent professional service is worth real money on top, a service-needed disclosure discounts the band).
Movement tier: How the caliber sits in the supplier chain: manufacture (in-house), top-supplier (Eta 2892, Valjoux 7750), mid-supplier (Eta 2824, Sellita SW200), entry (Miyota 8215, NH35), or unknown. Feeds the Reliability rating as a small lift, never overrules owner-verified data.
Era (vintage / neo-vintage / modern): Era split picks the right valuation factor table. Vintage (<1990) treats box & papers and unworn condition very differently from modern; neo-vintage (1990–2009) has its own market logic, especially for Daytona / GMT / Speedmaster references.
Brand tier (top-luxury / luxury / enthusiast): Heuristic tiering used by the valuation factor tables. Top-luxury includes the holy trinity, the haute-horlogerie independents, plus Rolex. Luxury includes Omega, Tudor, Grand Seiko, Glashütte Original, etc. Enthusiast is everything else and is honest, not pejorative.
Best-offer flex: When a listing accepts best-offers, the engine estimates the likely accepted discount from days-on-market, seller type, and brand tier. Used to score a 'true' landed cost.
Auction outlook: For auction listings, the engine forecasts a likely final hammer + buyer's-premium total from bid velocity, reserve status, and time-remaining. Heavy bidding well below band median doesn't mean 'bargain' — it means 'price will rise'.
Lifecycle (active / appears sold / appears delisted / stale / unknown): Hedged status of a listing. We never assert 'sold' or 'delisted' definitively from a single observation — the language stays 'appears X (last checked Y)' because mis-stating a listing's outcome is a tortious-interference vector.
Auction outcome (hammer-sold / aftersale-sold / passed-in / withdrawn / etc.): For auction archives, the actual fate of a lot. 'Withdrawn' (pulled pre-sale, often for authenticity concerns) and 'passed-in' (failed reserve) are real provenance signals; collapsing them into 'delisted' would lose information that matters.
Signal: One evidenced claim about a reference, caliber, brand, or seller. Carries provenance (source URL + author + published date) and progresses through a lifecycle (osint-unvalidated → community-validated or disputed → expert-reviewed).
Reputation gating: Counterfeit-detection details on the Inspection Map are visible only to high-reputation contributors so the knowledge base doesn't train counterfeiters. Most everything else is public.
Score audit trail: Every published score writes an immutable record of its inputs (canonical-JSON SHA-256 hash) and engine version. Lets us defend specific published statements against challenge. The trail survives erasure requests as a documented legitimate-interest override.