LearnProduct data

Structured data that shows where a fact came from

By Published 3 min read

Explore with AI

The short answer

Structured data that shows a fact's source means using schema.org markup to carry identifiers, units, revisions and modification dates alongside each value, so a machine can read not only what a product fact is but where it came from and when it was last confirmed. The markup states the claim. It does not make the claim true.

In this article · 6
  1. Markup states, it does not verify
  2. What Product markup can carry
  3. An illustrative block
  4. Where the source still has no home
  5. Organizations and locations
  6. Questions

Markup states, it does not verify

Structured data is a way to say what a value on a page is: a name, a part number, a dimension, a price. A machine that reads it does not have to guess from the layout. That is useful and limited. Markup that says a width is 36 inches is still wrong if the width is 35. Whether a value is trustworthy is a separate question, and the vocabulary is wider than most implementations use.

What Product markup can carry

  • Identity: name, brand, manufacturer, and the identifiers sku, mpn and gtin, kept distinct.
  • Measured properties: additionalProperty with a name, a value and a unit code, so a dimension has a unit instead of living in a sentence.
  • Relations: a part that is an accessory or spare for another, or a consumable for it, stated rather than implied.
  • Offers: price and availability, with the date through which a price is valid.
  • Dates: when the page or the record was last modified.

An illustrative block

The values are invented to show shape. They describe no real product.

{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "Casement window unit, series ABC",
  "brand": { "@type": "Brand", "name": "Example Windows" },
  "mpn": "ABC-3040-R2",
  "sku": "D-ABC30R2",
  "additionalProperty": [
    { "@type": "PropertyValue", "name": "Unit width (frame basis)", "value": 36, "unitCode": "INH" },
    { "@type": "PropertyValue", "name": "Rough opening width", "value": 36.5, "unitCode": "INH" }
  ],
  "isRelatedTo": { "@type": "Product", "mpn": "ABC-3040-R1", "description": "Superseded by this revision" },
  "offers": {
    "@type": "Offer",
    "availability": "https://schema.org/InStock",
    "priceValidUntil": "2026-12-31"
  },
  "dateModified": "2026-10-16"
}

Where the source still has no home

The vocabulary has no standard field for who confirmed a value, or which source system it came from. A page can state the modification date and an identifier. The rest has to live in a record the page points to, and the page and the record have to be generated from the same place so they cannot drift.

Generating the markup from the checked record, rather than writing it by hand per page, is the practical rule. The values, units and dates in the markup are then the record's, and a comparison against every public copy shows where a page, a feed or a directory differs from it.

Organizations and locations

The same applies to the company. Organization and location markup can carry legal name, identifiers, departments, parent and subsidiary relationships and the lines carried. In a dealer network, those relationships tell a machine which record is the parent and which is a branch, which is part of resolving duplicate entities.

The record behind it is described in Business Truth.

Questions

Does structured data make a value correct?

No. It tells a machine what the value is. Correctness depends on the underlying record.

Should mpn, sku and gtin be combined?

No. They are different identifiers assigned by different parties, and markup keeps them separate.

Why generate markup from a record?

So the values, units and dates in the markup are the record's and cannot drift from the source page by page.

Is there a field for who confirmed a fact?

Not in the standard vocabulary. That information lives in the checked record that the page points to.

Explore with AI

Where does your own information disagree?

Get Upstream

One email a week: what we asked the AI, what it got wrong, and how we checked.

Our research reaches 17,000+ followers across our channels. Upstream is the weekly version.

Weekly. Unsubscribe anytime. Privacy