29 SEPTEMBER 2026 · 7 MIN READ

We read 583 job postings. Fewer than half say how much experience they want.

Every guide to writing a CV rests on the same instruction: mirror the job description. Match the years, match the degree, match the keywords. It is good advice as far as it goes, and nobody checks how far that is. So we took a random sample of live postings and counted what a job description actually states — not what applicants assume it states.

What we did

We sampled 600 live job postings at random from our index on 29 September 2026, drawn across 70 randomly chosen role-and-location pages so the sample would not be one employer or one job family. 583 of them carried a usable description; the rest were near-empty. The systems behind them, in proportion: Greenhouse 219, SmartRecruiters 146, Workable 76, Lever 63, Ashby 57, and the remainder across Personio, Paylocity, Recruitee, Pinpoint and Breezy.

Then we read the description text — the same text you read — and counted. Everything below is a count of what the posting says, not an inference about what the employer wants.

First, the shape of the thing. The median description runs 633 words. A tenth are under 246 words and a tenth are over 1,157. A "long, detailed job spec" is, in practice, about two and a half pages of double-spaced text, and half of them are shorter than this blog post.

Fewer than half state how much experience they want

266 of 583 — 45.6% — named a number of years. The other 54% asked for experience in words that cannot be matched against anything: "proven", "significant", "demonstrable", "a track record of".

The median figure is 4 years, and the two most common numbers are 5 and 3 — which are round numbers rather than measurements. There is a visible preference for 5 over 4 and for 3 over 4, which is what you would expect if the figure is chosen by feel.

The practical consequence: for more than half of the postings you read, "do I have enough experience?" is a question the posting does not answer, and no amount of CV tailoring extracts an answer from it.

A third mention a degree, and a third of those don't mean it

203 of 583 — 34.8% — mentioned a degree of some kind: a bachelor's, a master's, a doctorate, or a named field of study.

Of those 203, 65 — 32% — also offered a way round it in the same posting: "or equivalent experience", "or equivalent practical experience", "or relevant experience". So roughly one in nine postings overall asks for a degree and simultaneously says a degree is not the only route.

Two-thirds of postings never raise the subject at all. If you do not have the degree, the modal posting is not asking about it.

One in nine wants a language other than English

67 of 583 — 11.5% — named a specific non-English language alongside a proficiency word ("fluent", "native", "business level", "C1"). The languages, in order: French, Japanese, German, Spanish, Dutch, Mandarin, Italian, Portuguese.

This is the single hardest requirement on the list to fake and the easiest to check, and it is also the one most likely to be invisible until you are halfway through a form. It is worth filtering for before you start writing anything.

What this means for how you write a CV

The advice to mirror the posting assumes the posting is specific. Usually it is not. Three things follow:

• Put your years in numbers, because half the postings won't. When the employer writes "proven experience" and you write "6 years", you have supplied the comparison they left out. You cannot mirror a requirement that was never stated, so state yours instead.
• Do not self-reject on a degree line. A third of postings mention one; a third of those say it is optional in the same breath. The degree requirement is softer than the genre of CV advice implies.
• Read for the language requirement first. It appears in one posting in nine, it is binary, and it is the one line that makes the rest of the tailoring pointless.

The broader point is that a large share of the tailoring effort goes into matching requirements the posting never actually made. We wrote about the 408,567 different job titles in the corpus for a related reason: the text of a posting is far less standardised than the advice industry assumes, and strategies that depend on it being standardised do not survive contact with it.

What we are not claiming

These are counts of stated text, nothing more. A posting that does not name a years figure may still have one in the recruiter's head; a posting that says "5 years" may hire someone with three. We measured what is written down, because that is the only part both you and we can see.

The matching is done by pattern, and patterns miss things. "Five years" spelled out, a requirement stated only in a bullet we did not parse, or a degree implied by "graduate scheme" will all be undercounted. Every figure here is therefore a floor rather than an exact rate.

The sample also skews toward the systems that dominate our index — Greenhouse and SmartRecruiters are over a third of it between them, which is roughly their real share of the 661,133 openings we track, but it does mean this is a picture of employers who use modern hiring software, not of the whole labour market.

Method: 600 live postings sampled at random on 29 September 2026 from 70 randomly selected role-and-location pages on ottoroles.com; 583 carried a description of 40 words or more and form the denominator throughout. Requirements were counted by pattern matching over the description text, biased toward undercounting. Years figures above 25 were discarded as parse errors. Any reader can reproduce this from the same public pages.