Data basis
What our job corpus actually says
This page is not someone else's study and not a market forecast. It shows what our own classification finds when it reads every active posting in our index: how many listings keep the promise they make, where information is simply missing, and how fast the stock turns over. Every figure comes out of live operations and moves with them.
As of: Time a posting was last confirmed in a source feed.
The stock
We currently carry 13,516 active job postings. 13,060 of them were seen again in a source feed within the last 24 hours, which is 96.6 percent. That second number says more than the first: it describes how much of the stock is currently confirmed rather than merely sitting in a database.
- Active listings
- 13,516
- Confirmed in 24 hours
- 13,060
- New in 7 days
- 2,092
- New in 30 days
- 11,215
Claim against reality on remote work
Of 5,475 checked postings that present themselves as fully remote, 4,011 hold up to that description. That is 73.3 percent.
Every posting gets a first classification at ingestion, taken from its own wording. A language model then reads the full text a second time, without seeing the first classification, and judges the actual work arrangement. A role only counts as verified remote here when both readings agree.
The gap is the real finding: 1,464 postings advertise remote work and still require regular presence somewhere in the text. 1,069 of them are hybrid, 395 are fully on site. These are the listings that cost applicants time, because the restriction only appears far down the page.
So far 6,029 of 13,516 active listings have been checked, which is 44.6 percent. The rate above applies only to the checked listings that describe themselves as remote. An unchecked posting counts neither as passed nor as failed.
Own claim against verified result
| What the posting claims | Listings | Checked | Verified remote | Verified hybrid | Verified on site |
|---|---|---|---|---|---|
| Calls itself hybrid | 7,016 | 554 | 471 | 72 | 11 |
| Calls itself remote | 6,500 | 5,475 | 4,011 | 1,069 | 395 |
Salary transparency
1,191 of 13,516 active listings state a salary at all, which is 8.8 percent. The far larger remainder leaves open the question applicants ask first.
We keep track of where the figure comes from. For 1,191 of these listings the provenance is on record: for 735 the source supplies its own salary field, and for 456 that field is missing and the figure appears only in the body text, where we read it out ourselves. Nothing is extrapolated or estimated: where a figure is missing, the field stays empty.
- With a salary
- 1,191
- 8.8 %
- From the source field
- 735
- Read from the text
- 456
Employment form
1,711 listings, or 12.7 percent, do not say what employment form they offer. That is not a gap in our data collection but a gap in the postings: we record what the posting itself states, in its text or in a structured feed field, and we deliberately do not read a missing value as full time.
| Employment form | Listings | Share |
|---|---|---|
| Full time | 10,675 | 79 % |
| Part time | 2,091 | 15.5 % |
| Fixed-term contract | 1,203 | 8.9 % |
| Internship | 835 | 6.2 % |
| Working student | 491 | 3.6 % |
| Temporary | 364 | 2.7 % |
| Freelance | 357 | 2.6 % |
| Apprenticeship | 144 | 1.1 % |
| Minijob | 13 | 0.1 % |
Shares are of all 13,516 active listings, including those that state no form. A posting can name more than one form and then counts in several rows, so the rows do not add up to 100 percent.
Where the demand actually sits
The role split comes from matching every posting against our curated role profiles, not from search volume. It therefore answers a different question than a keyword list: where are there really open roles in the current stock, and where is only the interest large.
- Home office4,528 Listings
- Sales2,271 Listings
- HR & People1,962 Listings
- Freelance1,873 Listings
- 100% Remote1,726 Listings
- Full-time1,721 Listings
- Sales (DE)1,537 Listings
- Data1,282 Listings
A posting can match several role profiles. Match computed: September 10, 2026.
Regions
Remote does not mean location independent. Many postings allow work from home but tie the role to one country or to the EU, usually for tax and employment law reasons. We track that restriction separately from the work arrangement so it does not disappear inside the word remote.
| Region | Listings | Share |
|---|---|---|
| EU and Europe | 6,497 | 48.1 % |
| Germany | 6,034 | 44.6 % |
| Not stated | 709 | 5.2 % |
| Worldwide | 244 | 1.8 % |
| USA | 32 | 0.2 % |
How long a listing lives
Listings that expired in the last 30 days were visible for 15 days on average, with a median of 12 days. We measure the span between the first and the last sighting in a source feed. That is a lower bound: a role may well have been open before we first saw it.
This number explains why a job board ages differently from a guide article. A large part of the stock is gone within weeks, and a page that keeps showing a dead listing wastes its readers' time. So we re-check daily and take dead listings out of the index.
- Average
- 15 days
- Median
- 12 days
- Listings evaluated
- 13,376
Where the data comes from
The stock is fed by company career pages, by job boards with an open interface, and by postings published directly with us. We name the origin of every single listing, because a figure whose origin cannot be traced is worth nothing.
| Source | Listings | Share |
|---|---|---|
| Personio | 6,269 | 46.4 % |
| Greenhouse | 1,925 | 14.2 % |
| Ashby | 1,739 | 12.9 % |
| SmartRecruiters | 1,706 | 12.6 % |
| Recruitee | 641 | 4.7 % |
| Workable | 607 | 4.5 % |
| Himalayas | 261 | 1.9 % |
| Jobicy | 158 | 1.2 % |
| We Work Remotely | 90 | 0.7 % |
| d.vinci | 44 | 0.3 % |
| Lever | 42 | 0.3 % |
| RemoteOK | 20 | 0.1 % |
| Remotive | 14 | 0.1 % |
Shares refer to the active listings at the date given above.
How this page is produced
Every value is computed straight from the live index, not from a curated spreadsheet. Where a check has not yet reached every posting, the coverage is printed next to it. Where a value is missing it stays empty instead of being estimated. Anyone who wants to quote one of these numbers may do so with attribution and a date.