A job advert is not the same on every board
The same role can be a two-line teaser on one board and a full brief on another. Which board you read it on decides how much of it you get.
Jobzyl3 min read
Figures on this page are current as of
You read a posting, cannot tell what the job involves, and conclude the employer wrote a vague advert. Often the employer did not. The board you read it on handed you a fragment.
This is the least discussed difference between job boards and the largest one we can measure.
The spread is more than twenty to one
Show the numbers
| Category | Value |
|---|---|
| smartrecruiters | 0 chars |
| careerjet | 258 chars |
| jooble | 278 chars |
| reed | 452 chars |
| adzuna | 500 chars |
| lever | 1107 chars |
| shine | 2329 chars |
| mycareersfuture | 2385 chars |
| indeed | 3938 chars |
| ashby | 5463 chars |
| greenhouse | 5907 chars |
Data as of
A typical Careerjet posting is 258 chars characters. Three sentences. A typical Greenhouse posting is 5,907 chars. These are frequently the same jobs, written once by the same employer.
The pattern is not random. It splits cleanly by what kind of thing the board is.
Aggregators truncate. Adzuna's median lands on 500 chars characters, which is a cap and not a coincidence. Reed's sits at 452 chars, which is its search API cutting the description at a fixed length. Jooble is in the same band. These boards are indexes pointing at postings, and their search endpoints were built to return a preview.
Applicant tracking systems hand over everything. Greenhouse and Ashby are the employer's own system, so there is no preview to serve, and you get the brief the hiring manager wrote.
Show the numbers
| Category | Value |
|---|---|
| jooble | 0.0% |
| adzuna | 0.8% |
| reed | 4.5% |
| smartrecruiters | 24.2% |
| lever | 30.4% |
| careerjet | 31.7% |
| shine | 80.9% |
| mycareersfuture | 84.3% |
| indeed | 93.7% |
| greenhouse | 99.7% |
| ashby | 99.8% |
Data as of
Reading it as a share rather than a median says the same thing more starkly. Greenhouse clears the bar on 99.7% of postings. Reed manages 4.5%.
One entry on that chart is partly about us
The chart does not distinguish this, so it needs saying.
SmartRecruiters shows a median of zero. That does not mean it publishes empty adverts. It means we hold no description text for most of the rows we have from it, for two reasons stacked on top of each other. Its listing endpoint genuinely sends no description, and the full text sits behind a separate per-posting call that we have backfilled for part of the corpus and not all of it. Half their API design, half our unfinished work.
Read the low bars as how much of the advert reaches a reader through that route, which is the question that matters when you are the reader. They are not a judgement on the employer, and the bottom one is not really a judgement on the board either.
What this means when you are searching
A short advert is frequently a good job seen through a narrow window. Discard postings for being thin and you are discarding by board rather than by role.
Click through before you decide. The version on the employer's own site is usually several times longer than the one you just read, and that is where the requirements, the level and the salary band tend to live.
It also explains why CV matching is uneven. Any tool that scores you against a posting can only read what it was given, and a teaser of a couple of hundred characters names almost no skills, so a match score computed from it is close to meaningless. That is why we show how confident a score is rather than always showing a number, and why a posting can be too thin to score at all.
Backend developer roles in Germany and UX designer roles in Australia both pull from every board at once, and each row says where it came from, so the difference above is visible in a single list.
Method
Measured on a three per cent sample of the index rather than all of it. Reading the length of every description means fetching and decompressing several million of them, which does not finish inside a query timeout. The sample is roughly 120,000 postings and leaves every board on the chart several thousand rows deep, which is far more than a median needs. A board with too few rows in the sample is dropped rather than shown with a wobbly number.
Not deduplicated, deliberately, and this is the one measurement in this series where that is correct. The question is what each board hands over when you fetch a posting from it, so two boards carrying the same job are two answers to that question, not a duplicate to be merged away.
The threshold for "a full description" is 1,200 characters, which is roughly where a posting stops being a teaser and starts listing duties. It is a judgement call, and the median chart is there so you can check it against a measure that needs no threshold at all.
The comparison covers the boards we index and name. Rows that reach us through a scraping fan-out rather than a provider integration are excluded, because we do not list those sites as sources and a chart is not the place to start.
Related
- Job boards overlap far less than you thinkThe same posting rarely turns up on more than one board. The repetition you notice is one board relisting, not boards copying each other.
- How old is the job you are looking atMost open roles on a company careers page are more than a week old, and many are past a month. Here is the age of what you are actually seeing.
- Where the salary is actually statedPay appears on a minority of job adverts, and which ones depends far more on the board that carried the advert than on the country the job is in.
Every figure here comes from the jobs Jobzyl has indexed, and carries the date it was pulled. Subscribe by RSS, or run the search yourself.