High-value trades already maintain the identification standards structured data requires — reference numbers, serials, provenance, certification. The implementation task is not inventing a taxonomy but mapping an existing one to Product, Offer and identifier properties. Unique inventory needs per-item markup rather than category-level markup, which most e-commerce templates get wrong.
Last updated 25 July 2026
Each item needs its own record rather than a shared model entry: identifier, condition, provenance where publishable, specifications and price or availability. The distinguishing requirement is that condition and provenance attach to the individual object, not to the product line it belongs to.
Per-item records also give each object its own addressable URL, which is what allows a buyer searching a specific reference to arrive directly.
Reference and model numbers map to identifier properties such as mpn or productID; serials and certificate numbers map to additional identifiers. The rule is to keep each identifier in its own field rather than concatenating them into a description, because only separated values are matchable.
Separated identifiers are what let a buyer who knows exactly what they want find it — the most qualified traffic in these categories, and the easiest to lose.
A catalogue product is one model with many identical units, varying by size or colour. One-of-one inventory is the inverse: each record is a distinct object whose value depends on its individual history and state. Applying catalogue-shaped markup describes the model and discards what makes the object valuable.
Most schema documentation and nearly all e-commerce examples assume the catalogue case, which is why generic implementations validate cleanly while describing the wrong thing.
Publish what establishes authenticity and specification: reference, certification, documented history where the owner permits. Withhold what establishes custody: current storage location, security arrangements, prior owner identities and acquisition prices. A reference number is a public fact; a vault location is not.
Four recur. Using category-level markup for unique items. Flattening identifiers into a single text description rather than separate fields. Omitting condition, which in these trades is a primary value determinant. And publishing images without the identifying text, so the object is visible to a person but unclassifiable to a machine.
The condition omission is the costly one, since two items sharing a reference can differ substantially in value on state alone.
From the file the business already maintains. In these trades inventory records are kept rigorously because authentication demands it, so the vocabulary already exists and is already current. Building a parallel taxonomy for the website creates a second source of truth that will drift.
The durable version connects the maintained file to a publishing step rather than replacing it, so the data stays current because the business already depends on it.
The visibility diagnostic scores whether your entity, credentials and catalogue are declared in a form a machine can classify.
Run the visibility diagnostic →