What 33,640 new job postings in one week look like (and why raw counts lie)

We maintain a small index of public job postings pulled from company career boards. This post shares what one week of it looks like, how the counting works, and where the numbers are weak. All figures describe the companies we track, not the whole job market.

What is in the index

As of 2026-11-12 the data API covers 1,277,259 open postings from 11,754 companies (live counts are at https://career-index-api.printfilestudio.workers.dev/v1/meta). That is two groups of sources. The core group is four applicant tracking systems with public job feeds, Greenhouse, Lever, Ashby and SmartRecruiters: 380,361 postings from 8,186 companies, refreshed daily, which is also what the website shows. The second group is Workday: 896,898 postings from 3,568 companies, refreshed on the 1st and 15th of each month, available in the API only. Postings older than 365 days are dropped (55,892 were excluded in the 2026-11-09 build), and known staffing or aggregator boards are excluded so we do not count the same job twice under a middleman.

The weekly figures in this post are computed from the four daily systems only, not from Workday, so they lean toward tech companies and European enterprises on SmartRecruiters. Other systems such as Taleo and iCIMS are not included.

Method: how a "new role" is counted

A role is a unique company plus a normalized job title, first published in the 7 days up to the build (window 2026-11-02 to 2026-11-09).

Normalization lowercases the title and strips location and requisition-id fragments. So "Sales Associate - Berlin" and "Sales Associate - Munich" at the same company are one role.

This matters a lot. In that window there were 50,253 raw new postings, but only 33,640 distinct roles. Without deduplication, the weekly numbers are inflated by about 49 percent.

Two further caveats:

• Some boards are very large. Our fetcher stops at 5,000 postings per board. Boards that hit the cap are marked with a plus sign, for example AccorHotel (739+ new roles), Domino's (619+) and AECOM (430+). Their true count may be higher.
• "Remote" means most postings of a role carry the remote flag or the word remote in the location. Companies set that flag in different ways, so treat it as approximate.

What the week looked like

Top companies by distinct new roles, all job types:

The ten biggest posters account for 12.8 percent of all new roles. The other 87 percent is spread across 4,230 companies that posted anything at all that week, out of the roughly 8,200 tracked that week. So most of the market is long tail, which is a useful reminder if you build job-board features around "top employers" only.

Remote share: 11.3 percent of new roles overall (3,793 of 33,640). By role family: Software Engineer 19 percent, Sales 18 percent, Finance and Accounting 14 percent, Healthcare and Clinical 10 percent.

AI and ML: titles that mention AI, machine learning, ML, LLM, GenAI, generative, deep learning, computer vision, NLP or data scientist made up 2.9 percent of new roles (988). One company, Bjak, posted 83 of them. Next were Bosch Group (22+), Sia (18), Clera (16) and OpenAI (16). The headline is partly an outlier effect: one employer posting a lot of near-duplicate titles can dominate a small category, so look at the spread and not only the leader.

A note on the keyword match: "data scientist" is counted as AI/ML in this definition, which is a judgment call. If you narrow it to titles with AI or ML in them, the share drops.

Try the data without signing up

The API has a free tier that needs no key. You pass exactly one filter per request, a company, a role category or a location:

curl "https://career-index-api.printfilestudio.workers.dev/v1/jobs?company=aecom&limit=5"
curl "https://career-index-api.printfilestudio.workers.dev/v1/jobs?role=software-engineer&remote=true&limit=5"

Free tier limits, verified against the live API: 20 rows per request, the first 5 pages only, and about 120 requests per minute per IP. Each row includes title, company, ATS, location, remote flag, role category, posted date and the apply URL. There is also a browsable version of the same data with one page per company and per role family.

If you would rather just download something, there is a free snapshot of 370,046 postings as of 2026-11-10 (CC BY 4.0, gzipped JSON Lines plus a 1,000-row CSV sample): https://open-career-index.printfilestudio.workers.dev/data/#snapshot

Who this is for

Three groups seem to be the natural users of this kind of data:

1. Small job boards that want to backfill listings so a new board does not look empty.
2. Recruiting and sales tools that want hiring signals ("this company just opened 12 data roles").
3. Labor-market researchers who want a transparent, reproducible source.

For groups who want the whole dataset instead of slices, there is a paid Starter plan at US$29 per month. The main thing it gives you is a daily full export of all open postings as JSON Lines or CSV, so you can load it into your own database and stop paging an API. You can cancel anytime from Polar's customer portal, and we refund the first payment on request within 7 days. To ask for a refund or contact us: open an issue at https://github.com/twdatatools/careersite-jobs/issues (or use the private contact link on that page if you prefer). If you are on the free tier and it is enough, stay there.

Things we would like feedback on

• Coverage: which ATS or large employers are you missing most? Workday is in now but only refreshed twice a month; Taleo and iCIMS are the obvious gaps.
• Fields: would you want salary ranges, department or seniority normalized? Today salary is not included.
• Dedup rules: is company plus normalized title the right unit for your use? For some uses you want every location as a row, which the raw export keeps.

Limits, stated plainly

• Counts cover 11,754 tracked companies (8,186 on four daily ATS systems, 3,568 on Workday). They are not a census, and the weekly new-role figures use the four daily systems only.
• Boards over 5,000 postings are truncated, so their numbers are lower bounds.
• Postings are as published by the employer. We do not verify that a role is still open between daily refreshes.
• Everything shown is publicly available on the employers' own career pages. Employers can ask for removal of a posting or company (see /about/ on the site).

Details, the plan, and the OpenAPI file: https://open-career-index.printfilestudio.workers.dev/data/?utm_source=devto