This page lists every figure in the job-posting keyword studies that was withdrawn, corrected or qualified after it was published: what the page said, what it says now, and why. Each study links here from its version history.
A figure that cannot be trusted is withdrawn rather than left in place, and the study page says so where the row used to be. If you quoted one of the figures below, the “Published” line is what you saw.
27.3% of postings (251 of 921) named the public sector as a segment.
Now
The row is removed from the table and the CSV. The analyzer now counts "federal" only where it names a buyer, such as a federal agency or federal accounts.
Why
The pattern matched the word "federal", which in most postings sits in the equal-opportunity line ("protected by federal, state or local laws"). On a later sample, 179 of 261 matches named no public-sector buyer, so the published share was about three times too high.
The figure was presented as the share of postings naming the spreadsheet.
Now
The page labels it a ceiling: the true share for the software is at or below the published figure. The analyzer now counts only the product, which applies from the next collection.
Why
The match ignored case, so the verb also counted ("excel in a fast-paced environment"). On a later sample, about one match in nine was the verb.
The methodology said the collector and analyzer were open source.
Now
The claim was removed. The scripts have since been published and are linked from the methodology.
Why
The code was in a private repository at the time, so the claim was false.
Report an error
If a figure looks wrong, write to support@zoevera.com with the page and the row. The methodology links the scripts that produced every table, so a count can be checked against the same public endpoint.