1. Establish the piece identity
The system starts with the most defensible identity it can establish: brand, product form, model or reference, material, color or variation, and—when it changes the market—size. A Rolex reference number carries more weight than a broad label such as “watch.” A baseball cap should not be compared with a beanie simply because both titles contain the same brand.
2. Retrieve broadly, filter carefully
Search retrieval is intentionally broader than the final evidence set. Candidate listings can help confirm an identity even when they are not suitable for pricing. Pieces then applies deterministic checks for obvious mismatches such as the wrong product form, incompatible model, accessory-only listing, duplicate URL, or a listing that does not expose a usable price.
3. Keep exact retail and resale evidence
An exact current retail listing and an exact resale listing can both contribute to an estimate. They answer different questions, so Pieces preserves their source type. Retail is especially useful for current or scarce products; active resale asks describe seller expectations; completed sales describe observed transactions. eBay’s own product-research documentation likewise separates active listings, sold-price ranges, condition, color, model, and other filters when researching a market.
4. Deduplicate before calculating
The same listing may arrive through more than one search provider. Pieces merges candidates using merchant, URL, listing identity, and provenance so a duplicated listing cannot vote twice. Distinct sellers or marketplaces remain distinct evidence.
5. Match the precision to the category
Variant strictness is not identical for every category. Reference, material, and size can materially change a watch or jewelry comparison. Shoe and apparel sizes are used when supplied, then the search can broaden when exact-size evidence is too thin. For cars or aircraft used as collection records, model-family evidence may be more useful than rejecting every listing over paint, mileage, or a pictured option.
6. Build a defensible range
The displayed market value is a central estimate and the market range describes the compatible evidence around it. Pieces does not average every number it finds. Obvious mismatches and duplicate evidence are removed first; extreme asks should not stretch an otherwise coherent range. The IRS definition of fair market value also emphasizes an open-market agreement between informed buyers and sellers and notes that condition, scarcity, demand, and recent transactions can matter.
7. Express uncertainty instead of hiding it
One exact listing can be useful, but it supports low confidence. Several compatible sources that agree support a stronger result. When evidence is too thin or contradictory, Pieces can withhold the estimate and let the collector add their own source rather than manufacture precision.
What Pieces does not claim
A Pieces market estimate is not authentication, a professional appraisal, tax guidance, financial advice, or a promise that an item will sell within the displayed range. Sellers may accept offers, charge fees, bundle accessories, or change a listing after it was observed.