The same field is 83% full in one job category and 21% in another
`ats_source` names the applicant tracking system behind a posting. Its fill rate varies by 62 percentage points across role categories, which is why the average is not published.
Measured across 2,229 jobs in 25 roles and all 13 Wellfound categories, `ats_source` — the applicant tracking system a company hires through — is filled on 83% of HR and recruiting postings and 21% of Marketing ones. That 62-percentage-point spread means the overall average describes no category well: a recruiter told to expect the mean would find the field missing four times as often as promised in Marketing. So the range is what gets published, and the same rule applies to `years_experience_min`, which runs from 12.8% to 75% by category.
Key points
`ats_source` is 83% filled in HR and recruiting and 21% in Marketing, measured over 2,229 jobs across 25 roles and all 13 categories.
A 62-point spread is not noise around a mean; it means the mean is a number that describes no category.
The pattern is explicable: categories whose companies hire through a formal pipeline expose an ATS link, and categories that do not, do not.
`years_experience_min` behaves the same way — 31.6% overall, 12.8% to 75% by category — so it is published as a range too.
Publishing a range rather than an average makes the field harder to advertise and makes a run's result predictable, which is the trade the site takes every time.
This is a short article about one column and one number that should not be published: its average.
What the field is
ats_source names the applicant tracking system a company hires through — Greenhouse, Ashby, Lever or Workable. It is useful for segmentation: a vendor selling into recruiting teams cares which system a prospect already runs.
The spread
Role category
ats_source filled
HR and recruiting
Filled83%
Marketing
Filled21%
Measured over 2,229 unique jobs across 25 roles and all 13 categories. Sixty-two percentage points between the highest and the lowest.
Why the average is the wrong number
An average is a useful summary when the spread around it is small. Here it is not a summary of anything: no category sits near the middle in a way that makes the middle informative, and a recruiter planning a Marketing run against a blended figure would find the field missing about four times more often than they were told.
The general version of this is the reason the site publishes sample sizes next to every percentage — why fill rates get published raw.
A plausible explanation, marked as one
The pattern is consistent with formality of hiring process: categories whose companies run a structured pipeline point applicants at an ATS, and categories where hiring is more ad hoc point them at an email address or an internal form. That is inference from the shape of the data, not something measured, and it is flagged as such in this article's caveats.
The same rule applied to experience
years_experience_min is 31.6% overall and runs from 12.8% to 75% by category. Same treatment: the range is published, the average is not offered as a planning figure.
Using a field that varies this much
Treat it as a filter rather than as a dimension. Rows where ats_source is present are a usable segment; rows where it is absent are not a claim that the company uses no ATS. Absence here means the posting did not reveal one, which is a fact about the posting.
Frequently asked questions
▸How often is the ATS identified?
It depends on the category, by a factor of four: 83% in HR and recruiting, 21% in Marketing, measured over 2,229 jobs across 25 roles and all 13 categories. Plan against the figure for your category, not against an average of all of them.
▸Which systems does it detect?
Greenhouse, Ashby, Lever and Workable. It is read from where the posting sends an applicant, so it is present when the company hires through one of those and absent when it does not.
▸Why publish a range instead of an average?
Because a 62-point spread means the average describes no category in the data. A single number would be easier to advertise and would mislead anybody planning a run in a low-coverage category.
Sources
Every URL below was requested and returned a page on the date shown.
A row can be fetched seconds ago and still be a day old. Cloudflare's cache sat in front of Wellfound job pages at up to 23.4 hours, and `data_age_hours` counts it.
All 58 fields grouped by what they describe, with the rate each was filled on across 2,229 jobs — and the three that are published as ranges rather than averages.
Each role paginates until its own corpus runs out. Measured ceilings ranged from 37 pages to 130, which is why a fixed page cap is the wrong way to bound a run.
