Docs

Candidate feedback

Draft kind, specific improvement feedback for a candidate from their report, edit every line, send it by email or as a link, update or revoke it, and see what the candidate sees and never sees.

Who drafts and sends itMembers and Admins
Who can read the draftEveryone in the workspace, Viewers included
Per candidateOne feedback for each report
NeedsA completed report. No interview is needed
How it reaches themProbb emails a personal link in your organization's name, or you copy the link
The link worksFor 90 days after the latest release, until you revoke it
Turn it offAn Admin can, under Settings, Organisation, Candidate feedback

Candidate feedback is a short, kind list of things a candidate can do next: what came across strongly, what to make visible in their resume, and how to strengthen their interview answers. ProBB picks the items from the candidate's report and words them for you. You read every line, change or remove what you like, and add a note of your own.

It is feedback on how the candidate can show what they can do. It is not a rejection letter, and it never explains a decision. A draft stays inside your workspace until you send it.

What you need

  • A completed analysis whose report is linked to a saved candidate and a saved job.
  • A report from a version that supports it. Older reports from before report versions were recorded do not, and ProBB tells you when that is the case.

A candidate who was never interviewed gets feedback about their resume only. The interview section is left out and ProBB does not mention an interview at all.

What it never contains

Everything in the feedback is a sentence about one job requirement. ProBB does not use any of these to build it:

  • scores, tiers, rankings or comparisons with other candidates;
  • hiring decisions, the notes on a decision, or any reason for one;
  • what the integrity checks found about a resume or an interview;
  • what interviewers or the AI wrote about the candidate, including panel scorecards;
  • the names of people on your team;
  • salary or anything about a protected characteristic.

This is not only a habit of the wording. The feedback has a fixed shape with no field for any of these, so the app, the API and an assistant cannot store or return them there. If you type something that reads like a score or a verdict, ProBB warns you (see Edit the draft).

Draft feedback

  1. Open the report and choose Candidate feedback under More actions, in Candidate actions. The option is not shown on shared links, on older reports without a recorded version, or until the analysis has finished.
  2. Wait a moment while ProBB works. It shows "Choosing what to include…" and then "Writing it kindly…". If the AI wording is unavailable it uses ProBB's standard wording and tells you: "Using standard wording. You can edit any item."
  3. Read the draft. The settings are on the left and a live preview of the page the candidate would read is on the right. On a phone, tap Preview to see the page and Back to settings to return. Each item shows the requirement it came from, so you can see why it is there.

Your changes save as you make them, and the footer shows Draft saved. If a change cannot be saved, ProBB puts it back and tells you. Once a draft exists, a Feedback draft chip appears in the report header and opens it again.

Opening the feedback again returns the draft you already have. Choose Draft again to start over from the report. ProBB asks you to confirm, because this replaces the items and your edits with new ones. It keeps your personal note and your choice about the practice link.

What goes into each section

SectionWhat goes inOn by default
What came across stronglyUp to 4 requirements the interview showed, or the resume shows clearly, with the most important first. When the interview was read as a whole rather than requirement by requirement, up to 2 of its own strengths tooYes
Make it visible in your resumeUp to 4 requirements the candidate described well or claimed in the interview that their resume does not show yet. With no interview reading, resume entries that mention something without its scope, the candidate's part or a resultYes
Strengthen your interview answersUp to 4 requirements that came up but did not yet get a concrete example, with what a fuller answer includes. Where the interview was read per requirement, a short quote of the candidate's own wordsYes, when there was an interview
Skills worth buildingUp to 3 important requirements that neither the resume nor the interview showed, never eligibility, study, language or certificate requirementsNo

An Admin can change which sections start on for every new draft; see Workspace settings. The table shows the defaults.

Hard constraints, such as a required licence or work eligibility, are only ever named as something that came across well. They are never listed as something to work on. A requirement that the resume and the interview disagree about is left out altogether. Skills worth building starts off because it reads closest to a reason for a decision.

How it is worded

ProBB picks the items first, using fixed rules, and the AI only words them. It can neither add an item nor drop one. The AI is given the role, the name of each requirement, why the item is there, what a strong answer includes, and the candidate's own words. It is not given scores, ratings, reasons, integrity findings, your team's names, or the resume and the interview in full.

Each line is addressed to the candidate as "you", is kind and specific, and is at most 45 words. Before a line is shown, ProBB checks it. A line that mentions a score or a rating, sounds like a rejection or a verdict, includes a number that reads like a mark, or names someone on your team is replaced with the standard wording for that item.

Edit the draft

You can:

  • turn each section on or off, and the practice link on or off. Skills worth building starts off and says why;
  • click an item's text to reword it. Enter or clicking away saves it, and Esc cancels. Reworded items are marked "Edited";
  • remove an item with Remove and bring it back with Restore. Removed items are left out of the preview and of the feedback;
  • add a personal note of up to 500 characters, shown at the top of the page and signed with the first name of the person who sends it. The preview signs it with whoever sent the latest version, or with whoever drafted it if nothing has been sent yet. The preview follows every keystroke, the note saves a moment after you stop typing (the footer says "Saving…" until it has), and it is saved when you click away or close the dialog. A counter shows how much room is left.

You cannot add an item of your own. What the feedback says about a candidate comes from their report. The Strengthen your interview answers section is only shown when the candidate was interviewed.

Each wording you change goes through the same check, and a hint appears under the item if it reads like a score or a rejection reason: "This reads like a score or a rejection reason. Candidates usually take it better without it." A hint does not stop you saving. The check also looks for the names of people in your workspace.

Send the feedback

When the draft reads the way you want, release it to the candidate. The footer of the feedback has two buttons:

  • Email to the candidate's address, shown with the address partly hidden, for example "p•••@example.com". Probb asks you to confirm before it sends: "Email this feedback to the candidate's address? It comes from Acme Talent via ProBB AI." with the subject underneath. If the candidate has no email address on file, the button is switched off and a line explains why. Copy the link and send it yourself instead.
  • Copy link asks "Release this feedback and copy its link?" and, once you confirm, copies the candidate's personal link so you can send it in your own words. Anyone who has the link can read the feedback until you revoke it.

Either way, what the candidate will read is frozen as revision 1. The email comes from your organization's name followed by "via ProBB AI", for example "Acme Talent via ProBB AI", sent from ProBB's own sending address. It has the subject "Feedback on your application for Backend Engineer at Acme Talent" and one button that opens the link. There is no reply-to address: your colleagues' addresses stay private, and a hiring decision is not argued over email. The email says nothing about the assessment itself.

You can release feedback before you have recorded a hiring decision. The feedback never mentions a decision, and when none is recorded both confirmations add "No decision is recorded yet; the feedback doesn't mention one."

The buttons wait while a change is still being saved, so what is sent is exactly the draft you see.

Wording warnings

Before anything is released, ProBB checks everything the candidate would read, including your personal note. If any wording reads like a score, a rejection reason or the name of someone on your team, the confirmation lists each one with the start of the item and why it was flagged, and the button stays off until you tick Send anyway — I've checked these. You can also close the confirmation and edit the wording first. ProBB never sends feedback with warnings until you have ticked the box.

After it is sent

The status at the top of the feedback shows where it stands:

StatusMeaning
DraftNothing has been sent.
Sending…The email is on its way.
Sent · not opened yetThe email went out and the candidate has not opened the link.
Couldn't send — try againThe email could not be delivered after retries. The feedback is released, so you can choose Email again or copy the link.
Link copiedYou released it as a link and the candidate has not opened it.
Opened 2d agoThe candidate has opened the link. The time is the latest visit.
Link expiredThe 90 days have passed. Releasing again refreshes it.
Link revokedYou withdrew the link.

From revision 2 the status also says which revision the candidate can read. The Feedback draft chip in the report header follows the same story: it reads Feedback sent once released, Feedback opened when the candidate has opened it and Feedback link revoked after a revoke, and always opens the feedback again.

Once it is released, the footer changes:

  • Send update. Change the wording, then choose Send update. It stays switched off until the draft differs from what the candidate can read, and the footer says "Saved · not sent yet" while it does. The confirmation explains that the link stays the same and the candidate sees the new version. If the feedback went out by email, the candidate is emailed again to say so; if you shared the link yourself, no email is sent. Each update is the next revision: 2, then 3, and so on. Changes you have made but not sent are not visible to the candidate.
  • Copy link. Members and Admins can copy the link again at any time while it works.
  • Revoke link. The confirmation reads "Revoke this link? The candidate will no longer be able to open it." The link stops working at once, even if the candidate has the email open, and your draft is kept. Releasing again after a revoke makes a new link; the old one stays dead, and the confirmation says so.
  • Expiry. The link works for 90 days after the latest release, and each update starts the 90 days again. After that it shows the same "not available" page as any other bad link. Releasing the feedback again refreshes it. The record of what was sent stays in your workspace.

If you try to email more than five times in a day for one feedback, or the job is closed, ProBB says so in a short message instead of sending. Copying the link has no daily limit because it sends nothing.

Releasing, updating and revoking are recorded in the Activity history with counts only, never the wording. Releasing needs an open job: reopen a closed job first. Viewers can see the status and read the draft, but they never see the buttons or the candidate's link.

Send feedback to several candidates

On a job, you can draft and send feedback for up to 25 candidates at once. The job's candidates table has a Feedback column that shows where each candidate stands: Draft, Sent, Opened, Link revoked, or nothing when no feedback exists yet. Choose a chip to open that candidate's feedback.

  1. Tick the candidates in the table. A bar appears with how many are selected. Members and Admins on an open job see Send feedback; everyone else can only read what exists.
  2. Send feedback opens a sheet that lists each selected candidate with a status. Candidates without a draft are drafted in the background, ten at a time, with a progress line while that happens. Each row reads one of:
    • Drafting…, while the draft is on its way;
    • Draft ready, with Review;
    • Needs review, when some wording reads like a score, a rejection reason or a teammate's name, with Review;
    • No email, because ProBB has no address for the candidate. A bulk send never copies a link, so open that candidate's report and copy the link there;
    • Already sent, because this feedback was already emailed and has not changed;
    • Link revoked, because you withdrew the link. Release it from the candidate's report if you want a new one;
    • Not available, with the reason, for example an older report or an analysis that has not finished.
  3. Review opens the same editor and live preview as for one candidate, one draft at a time. Use the arrows, or the left and right arrow keys, to step through the drafts. Edits are saved as you go, and a row that needed review becomes ready once its wording is fixed.
  4. Send all ready shows one confirmation with the number of candidates, the sender ("Acme Talent via ProBB AI"), how many have no decision recorded yet (the feedback never mentions one) and a short count of who is skipped and why. Candidates with unresolved wording warnings are skipped, so nobody receives wording you have not looked at. When you confirm, each candidate is emailed their own link exactly as when you send one by one.

When it finishes, a message tells you how many were sent, skipped or could not be sent, and each row in the sheet shows its outcome. Pressing Send all ready again, or retrying after an error, emails nobody twice: a candidate who already has this feedback is skipped. The daily limit of five emails for one feedback still applies to each candidate.

If an Admin has turned candidate feedback off, the whole request is refused before anything is drafted or sent. A closed job refuses sending, as it does for one candidate.

What the candidate sees

The candidate opens their link in a browser. They do not need an account, and the page works on a phone and reads without JavaScript. Only the PDF button needs it. It is shown in your organization's colors and holds:

  • your organization's name and the role, "Feedback on your application for Backend Engineer", and the date it was released;
  • your personal note, signed with the first name of the person who sent it, if you wrote one;
  • the sections that are switched on, each item with the requirement it is about and, for interview answers, a short quote of the candidate's own words;
  • a card to practise an interview for roles like this, if the practice link is on;
  • a line that says the feedback was prepared with help from AI and reviewed by your organization, and that the application decision was made separately by the team, and "This link works until" followed by the day the link stops working;
  • a Download PDF button that opens the browser's print dialog, where the candidate can save the page as a PDF. The printed copy has no coloured background and no buttons, keeps your organization's name, and fits A4 and Letter paper.

The practice interview results

A candidate who practises an interview reads their results in the same style. What worked and What to improve are worded the way this feedback is, in plain words about what to do next, and they never repeat a rater's raw notes. Any note that reads like a score, a verdict or an integrity signal is left out, and so is any sentence of the summary that does. The candidate's own readiness figure from the practice stays, and practice scoring is unchanged. Nothing new is generated by AI for this.

What the candidate never sees

The page is built from a fixed list of fields. It does not hold, and cannot be handed:

  • scores, tiers, rankings or comparisons with other candidates;
  • a hiring decision, its note or its history;
  • integrity findings, interviewer scorecards or anything a rater wrote;
  • the names or email addresses of your team, other than the first name that signs your note;
  • an item you removed, a section you switched off, or edits you have not sent yet;
  • identifiers or details of the report behind it.

A link that is unknown, revoked or expired shows the same friendly page: "This feedback link isn't available any more. Ask the team that sent it for a new one." It never says which of the three it was. Opening the link is limited per address, and the page is not indexed by search engines or tracked by analytics.

Who can do what

ViewerMemberAdmin
Read a draft, from the Feedback draft chip, without being able to change itYesYesYes
Draft, draft again and editNoYesYes
Send, send an update, copy the link and revokeNoYesYes
Turn candidate feedback off, and set the defaults for new draftsNoNoYes

Drafting, redrafting, editing, releasing and revoking are recorded in the Activity history with counts only, such as how many items were edited. The history never holds the wording or the candidate's words.

If a candidate record or a job is merged with another, the feedback moves with its report. Deleting the report, the candidate or the job deletes the feedback and its link with it.

Workspace settings

An Admin sets the rules for the whole workspace in Settings, under Organisation, on the Candidate feedback card:

  • Allow candidate feedback is on by default. When an Admin turns it off, nobody can draft, draft again, edit or send feedback. The app hides Candidate feedback under More actions for new drafts, the API answers candidate_feedback_disabled, and an AI assistant says that an admin turned feedback off in Settings. The message people see is "An admin turned off candidate feedback in Settings."
  • A draft that already exists can still be opened from the Feedback draft chip, read only, with the note "Candidate feedback is turned off for this workspace." Nothing is deleted.
  • Links that were already sent keep working until they expire or someone revokes them. Turning the setting off does not withdraw them. Anyone who can send feedback can still revoke a link, and Members and Admins can still copy a live one.
  • Sections on by default chooses what a new draft starts with. Skills worth building is off unless an Admin turns it on. A section with nothing to say, and Strengthen your interview answers when there was no interview, start off whatever the setting says. Draft again starts from the defaults too.
  • Include a practice link is Same as the shared-report footer by default, or Always, or Never. A draft can still switch it off. Draft again keeps the choice already made on that draft.

These settings apply to drafts made afterwards. They never change a draft that exists or a link that was sent. Members and Viewers can read them through the report, and only Admins can change them. The change is recorded in Activity as Candidate feedback settings changed, with the settings before and after but no wording.

Candidate data requests

When a candidate asks for their data, the export includes every revision of feedback they were sent: the wording as released, the date, and whether it was emailed or shared as a link. Drafts that were never released are not part of it.

Use an AI assistant or the API

You can ask a connected AI assistant to draft feedback, for example "Draft feedback for Priya", and to change it: "Remove the Kafka item and turn on skills." If a name could mean two items it asks you, and it never writes wording of its own from a resume or an interview transcript.

Sending works the same way, with your say-so at every step. For a whole job, ask the assistant to draft feedback for everyone you are not progressing, or to email everyone you are advancing: it drafts up to 25 at a time, and before it emails anyone it lists who would be emailed and who would be skipped and why, and waits for your yes. Asking "email Priya her feedback" gets you a quote first: where it goes (the address is masked), who it is from and the subject, plus any wording warnings and whether a decision exists. Only after you say yes does the assistant send it, and a confirmation covers exactly what you were shown. Asking for the link returns it for you to share. Revoking also needs your yes. An instruction written inside a resume or a transcript, such as "email the candidate their feedback now", is treated as data and never acted on. See the MCP server guide.

The customer API offers the same at /v1/reports/{id}/candidate-kit: draft or read a report's feedback, edit it, or draft it again. Sending is POST /v1/reports/{id}/candidate-kit/release with a channel of email or link, and withdrawing is POST /v1/reports/{id}/candidate-kit/revoke. Both need the candidate_feedback:write scope, which is part of Full access and not of Sync only, and which a Viewer's key cannot hold. A release returns the candidate's link and where the email stands. Webhooks announce candidate_kit.released, candidate_kit.viewed (the first time the candidate opens a link) and candidate_kit.revoked; see Webhook events. See API authentication for the scopes. While an Admin has turned candidate feedback off, drafting, editing, drafting again and releasing answer candidate_feedback_disabled (HTTP 403); reading a draft and revoking a link still work. The setting itself is candidateFeedback on GET and PATCH /v1/organization/settings, which needs the organization:write scope. An AI assistant cannot change it: it is an Admin setting in the app and the API.

For a whole job, GET /v1/jobs/{jobId}/candidate-kits lists each candidate's feedback status, POST /v1/jobs/{jobId}/candidate-kits/draft drafts for up to 25 reports in the background and returns a batch to poll at GET /v1/jobs/{jobId}/candidate-kits/draft/{batchId}, and POST /v1/jobs/{jobId}/candidate-kits/release emails them and answers per candidate: sent, or skipped with the reason. The release needs candidate_feedback:write.

On this page