Findings · 2026-10-22 · By VSNARY | Emmanuel Orta
Three sections said “not scored”. Our grade counted them.
For about twelve hours on 22 October, three new sections carried a “measured, not scored” label while the grade weighed them in full. Found by checking every row of one report against the site by hand, along with three rows that passed when they should not have.
Between 5:43 AM and 6:02 PM Mountain Time on 22 October 2026, three new CrawlCheck report sections (review integrity, price transparency and third-party dependencies) were labelled “measured, not scored” but were counted in the grade. A site that fell short in one of them lost up to about four points, and up to about eleven in all three. They are now excluded, the scoring version moved from 33 to 34, and the same row-by-row check fixed three rows that had passed when they should have failed.
Every CrawlCheck report splits its sections into two kinds. Scored sections add up to the grade. Unscored sections are measured and shown, but they say so in their title, “measured, not scored”, and the grade ignores them. That promise is printed on every report, and for about twelve hours on 22 October it was false for three sections.
What happened #
Three sections shipped on the morning of 22 October: Review integrity (whether a rating shown on the page matches what the markup declares), Price transparency (whether the page states figures an assistant can quote) and Third-party dependencies (which outside services the page loads and which ones hold it up). Each was built as a reading, not a judgement, and each was titled “measured, not scored”.
The grade does not read section titles. It reads a separate list that says how much each section counts, and a section missing from that list is counted at full weight rather than left out. The three new sections were never added to the list. So the report told the reader they were excluded, while the arithmetic underneath included them.
The first section went live at 5:43 AM Mountain Time and the fix at 6:02 PM, so every scan in those twelve hours and nineteen minutes was graded with them in.
Who it affected, and by how much #
A site lost points only where one of the three sections scored below 100. A site that passed all three was unaffected, because adding a perfect section to an average of perfect sections changes nothing. For a typical local business site, one of these sections scoring zero would have taken about four points off the grade, and all three at zero about eleven. A section that misses one row in five scores 80, and that cost the grade a little under one point.
Reports keep the scoring version they were graded under. A report from inside the window says version 33; a rescan now says 34. The change alerts that watched sites receive never compare scores across two versions, so no watched site was told its grade moved when only the rules did.
How it was found #
Not by a test. By reading one full report from top to bottom, every row of every section including the unscored ones, and checking each row against the site itself: the live homepage, its sitemap, its feed and its DNS records. The score composition table at the foot of the report gives each section's weight. Three sections whose own titles said “not scored” appeared in it with a weight beside them.
We already run a check across stored reports for exactly this mistake, a section that is graded without a declared weight. It would have caught this. Nothing acted on what it reported. A check that runs and is not read does no work, and that is the part of this correction that matters beyond the three sections.
Three more rows that passed when they should not have #
The same row-by-row pass found three readings that were measured correctly but judged wrongly. None of them is in the grade, which is why nothing outside the report changed; all three told a reader something reassuring that was not true.
| Row | What the report said | What was true | Now |
|---|---|---|---|
| RSS or Atom feed | passed as “declared, valid, recent” | the newest item in the feed was 882 days old | a feed whose newest item is that old no longer passes |
| Prices in the markup | passed on a price range of “$$” | “$$” gives an assistant no figure to quote, which the row's own explanation says | a symbols-only price range is shown as such and does not pass |
| Areas with a page of their own | 0 of 19 declared areas had a page | areas declared as “Arvada, CO” were matched against paths ending “-arvada-co”, which no site uses | the state suffix is ignored when matching; the same site now reads 1 of 19 |
| Three “not scored” sections | excluded from the grade | counted at full weight | excluded, scoring version 34 |
The area row is the instructive one. The site has a page called /arvada-tree-service/ and a homepage that is plainly about Arvada. The check was looking for an address shape the site never used, and reported the absence of something that was there.
What a reader should take from it #
If you read a CrawlCheck report from 22 October, rescan the site. If the grade moves, the report says which version it was graded under, so you can see whether the site changed or the rules did. We keep the two apart on purpose, and we have had to correct our own copy before.
More generally: a label is a claim, and the code behind it is a second claim. When they disagree, nothing in the report itself will tell you. The only check that works is the one that opens the thing and compares, which is the same check we ask readers to run on their own sites. Our accuracy page lists which findings have been checked against live sites by hand, and an earlier correction describes the last time a row passed for the wrong reason.
What changed in the product #
The three sections are listed as excluded on every report from version 34. The feed, price and area rows judge what they measure. And the check that catches a section graded without a declared weight is now read on every audit pass, rather than left to run unread.
The main product
Found this on your own site? We fix it for $500.
Scan free to see where you stand. The fix is one site, every finding implemented and re-measured, with a sealed before and after.
Questions this post answers
Did my CrawlCheck grade change because of this?
Only if your site was scanned between 5:43 AM and 6:02 PM Mountain Time on 22 October 2026, and only if it scored below 100 in review integrity, price transparency or third-party dependencies. Rescan the site to get a grade under scoring version 34.
How large was the effect?
For a typical local business site, about four points for one of the three sections at zero and about eleven for all three. A section missing one row in five cost a little under one point.
What does “measured, not scored” mean on a CrawlCheck report?
The section is measured and shown with its readings, but it does not count toward the grade. Every report lists which sections were excluded under the score composition table.
Will a watched site get a change alert because its grade moved?
No. Change alerts do not compare scores across two scoring versions, so a grade that moved only because the rules changed does not trigger an alert.
How was the problem found?
By reading one full report row by row, including the unscored sections, and checking each row against the live site, its sitemap, its feed and its DNS. The score composition table showed weights beside three sections titled “not scored”.
Which other rows were fixed?
Three rows outside the grade: an RSS feed whose newest item was 882 days old no longer passes as recent, a price range of “$$” no longer passes as a stated price, and declared areas written as “City, ST” now match location pages that use the city name alone.
Related findings
Comments
Comments are read before they appear. Nothing is published automatically, and no account is needed.
Writing about this? Facts, live figures and marks — every number on that page is dated and traceable to a scan.