Skip to content
CandidateBrief
Menu
Sign in
Start free trial
Help

Help

The five questions that generate the most support mail, answered from the same rules the product runs on.

Questions agencies ask

Does it rewrite or embellish the candidate’s CV?

No. It keeps the candidate’s wording, dates and numbers exactly as written. Anything you add is marked as your contribution, and a model can only propose facts that a person then confirms.

What counts as one source document?

The first successful export of an uploaded file in a billing period. Exporting it again, in another layout or format, or with an edited note, does not count again.

What does the de-identified version remove?

Name, email, phone, full address, personal links, photo, date of birth and ID numbers — from the text, headers and footers, link targets, file names and document metadata. Employers, schools and exact dates follow your client preset.

Can I edit the Word file afterwards?

Yes. It is a genuine DOCX with real paragraphs, headings and tables — not an image or HTML in disguise. The PDF is rendered from the same approved snapshot.

Where is candidate data kept, and for how long?

In private object storage tied to your workspace. Uploaded files and generated packs are deleted automatically after 7, 30 or 90 days — you choose — and can be deleted on request at any time.

Can we cancel at any time?

Yes. Cancelling stops the next renewal; your workspace keeps its plan until the end of the period you paid for, and existing files stay downloadable.

What counts as one source document

One source document is counted the first time it produces a downloadable file. Re-using the same uploaded file with another template, exporting an extra format, or fixing a recruiter note does not use another one. Uploading a different version of the same person’s resume does.

Monthly allowances: Studio 50, Team 150, Agency 400. Uploading and reviewing are never counted; only producing a downloadable file is.

Why an export is blocked

"All source documents in this period are used"
Existing profiles, files and links keep working. A file you have already exported this period can be exported again at no cost.
A source paragraph was never accounted for
Every paragraph must be included, omitted or redacted before a profile can be approved. The review screen lists the ones still open.
"No unsourced claim may be exported"
A fact either cites a paragraph of the uploaded file or is attributed to a person who added it. Model output can propose a fact; a person has to confirm it before it can be exported.
"The converter is unavailable"
The Word file was produced and verified; the PDF step failed. Retrying is safe and does not use another unit.

What de-identification removes

A de-identified pack is generated again from the filtered data. Nothing is drawn over, masked or hidden — the identifying values are simply not in the model that produces the file.

  • Names, email addresses, phone numbers, postal addresses and profile URLs.
  • Your own contact block, unless you chose to keep it on Brand.
  • Filenames, document properties, headers and footers, and link targets too — not only the visible text.
  • A current employer is kept by default, because a history without one reads as a gap. Each export says which choice it was made under.

How long files are kept

The default window is 90 days, changeable to 7, 30, 90. When it expires the original upload and the generated files are deleted; the confirmed facts and the review history are kept, so a profile still reads correctly after its file is gone.

Deletion also covers encrypted backups, at the date shown on the deletion record — which is later than the deletion itself, and is stated rather than implied.

Who can see what

Permissions are enforced by the server on every request, not by which buttons a screen shows. A member removed from a workspace loses access on their next request, whatever their browser still has open.

Still stuck

Write to the support address on the contact page with the correlation id from the error message. It identifies the exact failed request in the logs, which is usually faster than describing the steps.