How this data is collected
Everything on these pages comes from application forms that employers publish themselves. Nothing is guessed, nothing is submitted, and no candidate's answers are ever collected. This page says exactly how it works, so you can decide how much to trust it.
Where it comes from
Employers who hire through an applicant tracking system publish their open roles through that system's public API — the same endpoint their own careers page uses to render the list you'd see in a browser. Those APIs also return the schema of the application form: the wording of each question, whether an answer is required, and the options offered where the answer is a fixed choice.
We read those endpoints, once per board per run, and store the schema. Currently 212 companies across Greenhouse and Recruitee, covering 31,470 questions. The endpoints are listed at the bottom of every page.
What is never collected
| Collected | Never collected |
|---|---|
| The question's wording | Anybody's answer to it |
| Whether an answer is required | Names, emails, CVs, cover letters |
| The options offered on a fixed-choice question | Any personal data of any applicant |
| When we last checked | Anything behind a login |
This is a property of the database, not a promise about our behaviour: the
questions table has no column capable of holding an answer. We also never
submit an application, so we never see what happens after one is sent — which is why
these pages describe what a form asks and never predict what an employer will
do with your answer.
How current it is
Boards are re-read twice a day and every question carries the timestamp of the last run that saw it. A page whose underlying data is more than 14 days old is removed rather than published — a stale form is worse than no page, because someone would prepare for questions that are no longer asked. The verification date is printed at the bottom of every page.
Forms still change between our checks. If you open a form and it differs from what we show, we would rather hear about it than not.
Which pages exist, and which don't
A page is only generated where the data supports one. A company needs at least 5 questions beyond the contact fields and 3 open roles before it qualifies at all, and qualifying is not enough: every page is then checked for whether it contains anything the other pages don't. Pages that are mostly boilerplate are rejected and never published. In the most recent build that removed 6% of the candidates.
This is deliberate and it costs us coverage. A thin page is easy to generate and drags down every other page on the site, so an absent page is the better trade.
Getting something corrected
If a page is wrong — a question we show that isn't asked, wording that's out of date, or a company that would rather not appear — write to corrections@dream-a-job.com and we will fix it within 24 hours. Corrections from the employer are applied without argument.
Politeness
The crawler identifies itself, holds to a maximum of 8 concurrent requests across all sources, backs off on errors, and reads each board twice a day at most. If an applicant tracking system or an employer asks us to stop, we stop first and discuss afterwards.