A showcase is read by people who can check it. On 7 September we took a block of performance figures off this site — a case-studies page, a metrics strip on the home page, two testimonials — because none of them had anybody behind them to check with. Nothing was put in their place. This is what was there, what we did with it, and what the removal itself did not catch.
What the pages carried
Measured on the live site before the change, on 7 September at 06:40 UTC.
| Address | What it stated |
|---|---|
| / | 200+ training modules, 10k+ hours delivered, 98% instructor satisfaction, a 60% case teaser, two testimonials signed by job titles |
| /case-studies | 60% reduction in spend, a faculty satisfaction score of 9.4 / 10, +27% applicants, payback inside 18 months |
| /medical-xr | the same 60% reduction, as a teaser |
| /industrial-xr | 56% / 19% / payback in 1–3 years |
| /vargates-medical | 200+, 10k+, 98% |
The page was the visible half
One module fed the home page, the case-studies page and a teaser on each of the two vertical pages. Deleting the case-studies page on its own would have left the same figures on the home page, where far more people read them than ever reached the page anybody complained about. So the module, the component that drew it and the page went together — with the three pages that imported it, the guard list that named it, the page art and some 350 lines of dead stylesheet.
And the same figures had already been hand-authored somewhere the module could not reach: the strip on the VARGATES Medical page was a local constant of its own, and so were the industrial ones. That is the other half of the problem, and the more instructive half. A number with no source does not stay where it was written.
The test that holds the result proves the three sources are deleted rather than merely unused. An unused module is a module somebody imports again.
What made them unpublishable
Not the wording. The absence of anybody to check them with.
Three case studies carried exact percentages and no named client — a «Regional Medical University», a «Kazakhstan Mining Group», a «Regional Technical College». Two testimonials were signed by a job title and a type of institution rather than by a person.
For a showcase read by distributors that is worse than having no case studies at all, because the first serious reader asks for the institution’s name.
410, not 404 and not a redirect
The withdrawn address answers 410 Gone, and the three available answers are not interchangeable.
404 says the page never existed. That is false, people linked to it, and it is what a crawler records.
A redirect says the content moved. Also false — it was withdrawn, and the reader would land on a page that no longer makes the claim they came for.
410 says the address was ours, and what stood here is gone. It is the only true one of the three.
Nothing was put in their place
The tempting repair is a smaller number. It is the wrong one: a figure nobody measured, replaced by a figure nobody measured, is the same claim in a quieter voice.
Where a figure was a statement of capability it stays — what is deployed, what is in the catalogue. Where it was an outcome, the copy now says what the product does instead of what it is supposed to have achieved.
The group keeps one register of the group-level figures it states about itself, and every entry in it carries named, checkable provenance — for most of them, who said so, when, and in what review. A figure without one does not go on a page.
The change was not the acceptance
The removal was accepted by walking the live site rather than by reading the change — which is the only reason this article has a second half.
Three figures were not on the list of what to remove, and they survived: 56%, 19%, and a payback of 1–3 years. Counted on the live site on 8 September, the string «56%» is served from seven addresses.
| Address | Occurrences in the delivered page |
|---|---|
| /industrial-xr | 12 |
| /group | 4 |
| /mechanicum3d | 4 |
| / | 3 |
| /investors | 3 |
| /products | 2 |
| /vargates-technics | 2 |
In the sources it lives in nine files and ten occurrences; the industrial page carries it twice, once in its opening line and once in the meta description. What makes it a defect rather than an oversight is what those places call it:
| Where | What the number is labelled |
|---|---|
| /industrial-xr | 56% less training time |
| /mechanicum3d | 56% — faster course development |
| /vargates-technics | 56% — faster onboarding |
A reduction in training time, the speed of building a course and the speed of onboarding are three different quantities. One number cannot be all three, so at least two of those labels are wrong. A figure that changes what it measures depending on the page is behaving as decoration rather than as a measurement.
On the investors page the same number is introduced as «documented enterprise-client metrics». «Documented» is a claim about provenance, and the investors page is the last address on which such a claim can stand without a source beside it.
There is a second consequence, and it is why this cannot simply wait. The metrics block came off the industrial page while that page’s opening line and its meta description — which is what travels into the search result, the link preview and the structured data — still state the same numbers, and the opening line does not merely restate them: it says they were measured on real deployments. The page now disagrees with itself.
Why a showcase publishes this
Because the alternative is worse arithmetic. A partner who finds one unbacked number stops believing the rest of the page, and they are right to.
We sell simulation to universities, hospitals and industrial operators, all of whom are obliged to check what they buy. VARGATES Medical is bought by people whose own accreditation depends on the claims they can substantiate. A showcase that holds itself to a lower standard than its customers are held to is not a showcase; it is a liability.
The standard is one line long and it is cheaper than the alternative: every figure on a page carries the source that established it, or it comes off the page — and nothing takes its place.