How we know
Every number on this site is a count of real job postings. This page explains exactly how the counting works, including where it is weak.
Where the postings come from
We read the complete text of postings from employers’ own careers pages, through the public job-board APIs of the applicant-tracking systems that host them — Greenhouse, Lever, Ashby, and SmartRecruiters. Those feeds exist so that postings can be syndicated, need no credentials, and return the whole advert rather than a preview. We do not scrape LinkedIn, ZipRecruiter, or any site whose terms prohibit it — not as a matter of caution but because the resulting data would be unusable to publish.
The trade-off is who is in the sample. The corpus is a curated list of employers that run a hosted ATS, which skews toward technology, digital health, hardware, and consumer companies and away from hospitals, trades contractors, and public bodies. Every page states its sample size, and roles the corpus cannot reach are listed as uncovered rather than filled in from somewhere weaker.
No single employer may be more than 25% of any sample. These feeds carry one advert per vacancy, so a large employer hiring the same role in fifty cities can outnumber everyone else combined — and the page would then be reporting that company’s requirements as a market’s. Where trimming to that ceiling leaves too few postings to publish, no page is written. That is why some roles the corpus clearly contains have no page: we can see the adverts, but they are all from one company. Every page states how many employers its sample came from and how much of it the largest was.
The occupation list and its codes come from the O*NET-SOC taxonomy. The credential catalogue — names, issuers, fees, study hours — is maintained by hand against each issuer’s own page, with the month each fee was last checked. Wage bands come from the Bureau of Labor Statistics Occupational Employment and Wage Statistics programme.
- Employer job boards (Greenhouse, Lever, Ashby, SmartRecruiters APIs) — Each employer's own careers page
- O*NET Web Services — U.S. Department of Labor / ETA
- Occupational Employment and Wage Statistics — U.S. Bureau of Labor Statistics
How credentials are counted
We maintain a catalogue of 61 credentials — certifications, licences, clearances, and degrees — each with the names employers actually write. A posting is matched against that list by exact name, not by a language model. That choice is deliberate: a model deciding whether a posting “mentions CCNA” is a black box you would have to take on faith. A name match is something you can check yourself.
A posting counts once per credential no matter how many times it repeats the name. The figure we publish is postings that mentioned it, out of the whole sample — including the postings that mentioned nothing at all, because those are evidence too. Dropping them would inflate every percentage here.
What “lower signal” means
A credential appearing in 10% or more of the sample is listed as something employers ask for. Below 5%, it is listed as lower signal for that specific role in that specific place. That is not a claim the credential is worthless — plenty of them are valuable for a different job, or a different city, or a promotion later.
We will not make the lower-signal call from a sample smaller than 40 postings, because at that size a single posting moves the percentage by more than the threshold does. When the sample is thin, the section is simply absent.
We sell no courses and take no fee from any certifying body, which is the only reason we can print that section at all.
Where this is weak
- Samples are small. Each page states its exact sample size and date. A few dozen postings is a snapshot, not a census. We do not publish a page at all below 15 postings — at that size the percentage is mostly an artefact of the denominator, and printing the denominator underneath does not rescue it.
- Postings are not jobs. They are what employers wrote down. Requirements listed as mandatory are frequently negotiable, and the reverse happens too.
- Not every role and city is covered. Where we have no local sample we show national figures and label them as such rather than passing them off as local.
- Costs drift. Every dollar figure carries the month we last checked it against the issuer’s own page. Confirm before you pay.
The disclaimer
Throughline is informational. It is not career counselling, and nothing here is a guarantee of employment, admission, or licensure. Verify requirements with the certifying body and your state before spending money.