We Said the Job Market Was 61% Senior. We Were Counting Office Managers.

For most of this year one of our most-read pages opened with a clean, quotable number: 61% of active job postings want senior or above. It was wrong, in a way worth writing up, because the mistake is one almost every job-market statistic you read is exposed to.

The number reproduced, which is what made it dangerous

The first thing we did when re-measuring on a fresh corpus was check whether the figure had simply drifted. It had not. On 257,534 active postings the classifier still returned 60.6% senior or above. Same answer, five weeks later, on a corpus that had grown by tens of thousands of rows.

A number that stable feels trustworthy. It was stable because it was measuring something stable: our own vocabulary, not the labour market.

What the classifier was actually doing

Seniority in our pipeline is assigned partly from a list of title patterns. The senior band's list contains the bare words manager, specialist, supervisor and lead. Anything matching lands at senior.

That band holds 100,764 postings. Inside it:

• 36.7% contain the word "manager"
• 12.7% contain "specialist"
• 57.2% carry no seniority word in the title at all
• 55.0% state no years-of-experience requirement anywhere in the description

The most common titles in our "senior or above" bucket that carry no seniority word include product manager, business development manager, account manager, customer success manager, store manager, mortgage specialist and office manager.

We were publishing office managers as senior openings.

Product was the clean kill

The old page named product management as the worst field for newcomers, at roughly 69 senior openings for every entry-level one. That claim is entirely an artefact:

82.0% of the product postings our classifier called senior reached that band through a title containing the word "manager."

Product Manager is the standard title for that role at every level, including the most junior. The finding was a readout of a job title, not of the product job market.

The control that proves it is not board mix

The obvious objection is that our corpus over-samples certain job boards, and boards differ. So we computed the same comparison inside each source separately. The classifier exceeds a title-based count in all 17 sources, from 1.33x at one to 3.98x at another, with none going the other way.

When an effect holds inside every single source, it is not composition. It is the instrument.

What replaces it

Counted by what titles actually name, senior or above is 29.9%, not 61%. But we are not swapping one false precision for another, because the title count is a floor: a posting titled simply Engineer frequently demands eight years in the body, and 63.3% of the market names no level at all.

So the honest answer is a range. Somewhere between 29.9% and roughly 61% of active postings target senior-or-above candidates, and the width of that band is the real finding. Anyone quoting a single decimal place for the market's seniority mix, us included until now, is quoting their own classifier.

What survived

The direction did. Among the 104,203 postings that state a years figure, 60.7% want five or more — an independent measure taken from description text rather than titles, landing in the same place by a route that does not touch the broken classifier. Explicitly entry-level titles really are scarce, at 1.7% of the market.

The market does skew experienced. We just could not support the precision we published.

The generalisable lesson

The check that caught this was cheap and we should have run it sooner: measure the same quantity a second way, from a different field, and see if the two agree. Our seniority band came from titles. The years requirement comes from description text and never reads the title. Comparing them costs one query and it is the difference between a number and a claim.

The corrected page, with the full method and the original figures preserved so the error stays auditable, is at the job market seniority pyramid.