A team ships structured data across a few thousand pages. The validator reports no errors. The report says the schema project is complete, and nobody ever asks again whether it did anything.
No errors means the markup parses. It says nothing about whether a single additional person clicked a single additional result.
What would it mean for structured data to be working?
One of three things, and they are worth separating because they fail independently. Markup can be valid, eligible and earning — and it is common to have the first without the second and the second without the third.
| Level | Question it answers | What proves it |
|---|---|---|
| Valid | Does the markup parse and use the right properties? | A validator, or the enhancement reports in Search Console |
| Eligible | Is this page type capable of a rich result at all? | Google's documented requirements for that type |
| Earning | Is the enhanced appearance getting clicks? | Search appearance data over time, plus what you see in the results |
Most schema reporting lives entirely at the first level, because that is the level a tool can check in a second. The other two take longer and are the only ones connected to revenue.
A validator answers "is this well-formed?" Nobody is paid for well-formed. The question worth reporting is whether the enhanced listing is being shown and clicked, and no validator has ever been able to answer it.
Why is valid markup such weak evidence?
Because eligibility is Google's decision, not your markup's. Meeting the documented requirements makes a page a candidate for a rich result; it does not oblige anyone to show one. Whether it appears depends on the query, the device, the competition on that page of results, and Google's own judgment about whether the enhancement helps.
There is a blunter reason too. Google has withdrawn rich result types before — FAQ and HowTo results were both heavily reduced after having been widely implemented. Sites that had built reporting around "our FAQ markup is valid" carried on reporting green long after the results themselves had largely disappeared. Validity survived; the result did not.
What can you actually measure?
Four things, in descending order of how much they tell you.
Search appearance data. Search Console's performance report can be filtered by search appearance, which splits out impressions and clicks for enhanced result types your site qualifies for. This is the closest thing to a direct answer, because it reports on results that were actually shown to people.
The enhancement reports. These count valid and invalid items per type and, more usefully, show when the valid count changes. A drop in valid items is usually a template regression that nobody would otherwise have noticed for months.
Click-through rate on affected pages, against a control. Compare the pages that got the markup to a comparable set that did not, over the same period. It is crude and it is the only way to see an effect that search appearance filtering does not isolate.
What the results look like. Search your own queries and look. It is unfashionable as measurement and it catches things no report surfaces — a competitor whose listing now carries a rating yours does not, a stale price showing in your own result.
How do you run a before-and-after that means anything?
By deciding what you are comparing before you deploy, which almost nobody does because schema tends to ship everywhere at once.
- Roll out to a subset. One category, one template, half the product range — enough pages to see a pattern.
- Record the baseline for those pages and for the holdout: impressions, clicks and position over a period long enough to cover normal variation.
- Deploy, then wait. Rich results depend on recrawling, and a large section is not reprocessed in a weekend.
- Compare the two sets, not the treated set to its own past. Seasonality and algorithm updates hit both; only the difference between them is attributable to the markup.
If the treated pages move and the holdout does not, you have something. If both move together, you have a season.
Marking up something the page does not visibly contain — review schema for reviews that are not displayed, prices that differ from the ones on the page. That is not an aggressive tactic with a measurement problem; it is a guidelines violation that can cost the site its rich results entirely.
Which types are worth measuring at all?
The ones that change what the listing looks like for a query with commercial intent. Product schema that puts a price and availability into the result is doing visible work. Organization markup that feeds a knowledge panel is doing slow, mostly invisible work. Both may be worth having; only one of them should be in a monthly report.
The other reason to be selective is maintenance. Every schema markup type you deploy is a template that can silently break the next time the site is touched. Types that never earned anything are pure liability, and reviewing them is a legitimate annual task.
What does good schema reporting look like?
Three lines, per type that matters: valid items, impressions in that search appearance, and clicks. Trended monthly. That is enough to catch a template regression, to see an enhancement being withdrawn, and to answer whether the type is earning its maintenance.
Compare that to what most sites report — a screenshot of a validator with a green tick — and the difference is a project you can defend versus one you can only claim to have finished.
Go and check one thing today: filter your organic search performance by search appearance and see whether the types you deployed are listed at all. If a type you shipped a year ago does not appear, you have your answer about it, and a decision to make about whether to fix it or delete it.
Shipped schema and never saw a result?
We will audit what your markup is eligible for, what it is actually earning in search, and which types are worth the maintenance.
Get in Touch →Does this apply to your site?
Reading about it is one thing. Point the scan at your own site and see whether this applies to you, and what it is worth fixing.
Free and unlimited. No account, no card, and you get every finding rather than a teaser.