Structured data
One business entity emitted once across the site, referenced by identifier everywhere else.
Structured data is markup a reader never sees and a search engine treats as authoritative, which is an unusual combination and the reason it deserves more care than its visibility suggests. The most common failure is not a missing property; it is the same organization declared repeatedly, slightly differently, on every page of a site.
One module builds all structured data for the site. The business entity is emitted a single time with a stable identifier, and every other page references that identifier instead of restating the organization. A site that declares its business on forty pages has told a search engine about forty businesses that happen to share a name.
Properties come from the same configuration that renders the visible page, so the markup and the text cannot disagree. Structured data that contradicts the page it describes is worse than none, because it is the version the machine believes.
Aggregate rating and review markup are not emitted, ever, in the absence of real attributable reviews with a public source. This is the single highest-risk piece of markup available, precisely because it is invisible on the page and carries weight with the engine reading it. There is no configuration option to switch it on.
Breadcrumb markup mirrors the actual route hierarchy, not a hierarchy invented for the markup, and page-type declarations describe what the page genuinely is.
Every emitted graph is validated against the published vocabulary before release, and the validation is part of verification instead of something done once during development.
Choosing the entity type is a decision with consequences, not a formality. A business with a location customers visit, a business that travels to customers, and a business with no public premises at all are three different declarations, and picking the wrong one produces results that look correct in testing and misrepresent the business in the places it actually appears.
Only facts already present on the page are marked up. Structured data is not a place to state things the site does not say — an address that appears in markup but nowhere in the visible content is either an omission from the page or an assertion nobody has reviewed, and both are worth catching before release.
Where structured data would describe something the business cannot substantiate, it is omitted rather than approximated. The absence of a property costs a rich result; a wrong property costs credibility with the one reader that cannot be argued with.
What this does not cover.
- Aggregate rating, review, or star markup without real attributable reviews carrying a public source.
- Per-location business entities. One organization node exists sitewide and is referenced by identifier.
- Markup describing content that is not present on the page it appears on.
Web & Digital
Content in version control
Why the words on a site live under version control instead of in a content management system, and what that trade actually costs.
Information architecture
The route table is written and agreed before anything is designed, because page counts later become assertions the build enforces.
Campaign and landing pages
Single-purpose pages wired to measurement that reports what actually happened, not what a model inferred.
Accessibility
Contrast computed, not estimated, focus visible on every control, and keyboard navigation tested by using it.
Performance
Work concentrated on what a first-time visitor on a slow connection experiences, not on a score.
Indexing and canonicals
Four independent layers decide whether a site is indexable, and they have to agree.