# Review answers and share feedback

## Purpose
Work through the responses you have been assigned, record a verdict on each answer,
correct what is wrong, keep notes, and send the parts you want the participant to see.

## Who this is for
Auditors, and the managers who oversee them.

## Getting in

**Auditor workspace** in the sidebar. The entry only appears if you either hold a
manager permission or have at least one auditor assignment. If you open it without
either, the page says "You do not have auditor workspace access in this account."

If you expect assignments and see none, ask a campaign manager to assign you (see
[Add people to a campaign](/operate/tenant-admin/add-people-to-a-campaign)).

**If you also answer questionnaires yourself**, the sidebar has a second group, **As a
participant**, with **My assessments**. It appears only when you have assessments of
your own to take, and it opens the same page your invitation email leads to.

The workspace has two screens and a review mode.

## Your assignments

**Auditor workspace** shows one card for each assessment you have been asked to review:
the campaign, the assessment, a line such as *6 responses · 2 reviewed · 1 in progress*,
a progress bar, and one button. The button reads **Review responses** while there is
work left and **Open** once every response is reviewed. If you have exactly one
assignment, the workspace opens its responses straight away; **Auditor workspace** at
the top of that screen brings you back to the card.

A manager also sees the account's other assignments here, with a search and a scope
filter, and the **Shared notes feed** tab (see [below](#managers-the-shared-feed)).

## Responses to review

Choosing an assignment shows its responses, oldest first:

- the campaign, the assessment and the same summary line as the card;
- **Notes on this assessment** (see [The three kinds of note](#the-three-kinds-of-note));
- **Export all responses**, with the file format beside it (see
  [Export your responses](#export-your-responses));
- filters with their counts: **All**, **Not started**, **In progress**, **Reviewed**;
- one row per response: the respondent, their organisation, the day it was submitted,
  how many of its answers are reviewed (*7 of 24 answers*, with a small bar), its
  status, and a button: **Start review**, **Continue** or **Open**. A response that a
  manager has reopened, or that came back after a reopen, says so beside its status.

Dates and times in the workspace are the campaign's own, in Ottawa time, and each screen
says so once.

The list holds **submitted responses only**. A draft that a respondent has saved but
never submitted is not a response, and it is not listed or opened here for any auditor.
A response that a manager has reopened for correction stays listed for ordinary
auditors (it was submitted), but not for a
[restricted assessment auditor](#the-restricted-assessment-auditor).

## Review mode

Opening a response gives it the whole page, one question at a time.

### The two bars

The top bar says who answered: **Responses** (back to the list), the respondent, their
organisation, the assessment, and when it was submitted, in Ottawa time. **Previous
response** and **Next response** step through the list, and the bar says where you are
(*Response 2 of 6*). **More** holds everything about the whole response (see
[More](#more)).

The second bar says how far you have got: *7 of 24 reviewed*, a two-colour track
(verified, then corrected), **Show** (**All questions**, **Not reviewed**,
**Verified** or **Corrected**), **Questions** (shows or hides the outline), and **Next
not reviewed**, which jumps to the next answer nobody has reviewed, wrapping round to
the start. It is unavailable once everything is reviewed.

On a wide screen both bars stay at the top while the question scrolls. On a phone
everything stacks.

### The three statuses

Every answer is **Not reviewed** (an empty ring), **Verified** (a tick) or
**Corrected** (a pencil). The marks differ in shape as well as colour, and the card
always says the status in words. These are the workspace's words; the stored values,
the API and the exports still say `unverified`, `verified` and `modified`.

### The outline

**Questions** shows the questionnaire's own **sections** (its pages, or its panels when
a single-page questionnaire groups its questions that way), each with *n of N reviewed*
and a tick once every answer in it is reviewed. A section with no title of its own is
named by its place (*Section 4*); answers the questionnaire has no record of (a renamed
question, a legacy import) are listed last, under *Other answers*. Every answer is
numbered **Q1**, **Q2** and so on, in the order the respondent met it. The section you
are in lists its questions with their status; choose one to show it.

### One question

The card says where it is (*Leadership · Question 8 of 24*), its status, the
**question** as the respondent saw it, and the **Response**, with choice labels
resolved. When a correction applies, it shows **Respondent's answer (kept)** and
**Answer in effect** side by side, with the **Reason** under them. The answer in effect
is the value the scorecards, exports and published results use; the respondent's own
answer is always kept and always shown.

Under it:

- **Verify and next** records your verdict at once and moves to the next question
  nobody has reviewed. A message at the foot of the screen says what happened
  (*Q8 verified. Now on Q9.*) and offers **Undo**. On a verified answer the button
  reads **Verified · Undo**.
- **Correct** opens the correction panel (see below). Once an answer is corrected, it
  reads **Change correction**.
- **Note on this answer** opens a note on that one answer. Once there is one, the card
  shows it as **Your note on this answer**, and the button reads **Edit note on this
  answer**.
- **More details** opens everything else about the answer (see below).

**Previous question** and **Next question** under the card move through the answers
**Show** leaves, across sections. When nothing in the response is left to review, a
**Review complete** banner says how many were verified and corrected, with **Back to
responses** and **Next response**.

If you have typed something you have not saved (a note, a reason, a hashtag, a change
to sharing) and move to another question, another response or back to the list, the
workspace asks before throwing it away.

### Correcting an answer

The correction panel lists the answers the question offers, each on its own row, with
the respondent's own answer marked and the answer in effect chosen. A long list, a grid,
a multi-select or a free-text question keeps its own control (one selector per row on a
matrix, check boxes on a multi-select, a text box only where the question itself is
free text). On a **maturity item** a correction has to be a point on the item's own
scale, and the panel says which. On a **qualifier** (a province, an employment band, a
sector, a discipline, a job title) you are choosing between the answers the question
offers, because a qualifier records what the respondent *is* and segments the cohort
rather than scoring it.

**A reason is asked for only when something changes.** Choose a different answer and an
amber box appears: **Reason for this change** (required, at least five characters),
with a line such as *b. Initial → c. Defined. The respondent's answer (b. Initial) is
kept.* Choosing the respondent's own answer again **puts it back**: the correction is
cleared, the answer stands as verified, and the line says *This puts the respondent's
own answer back.* **Save correction** applies it.

**Note on this answer** is always there in the panel, and optional. With nothing
changed, the panel says *Pick a different answer to correct this one. A reason is asked
for only when something changes.*, and its button reads **Save note**: it saves the
note alone, with no reason.

The reason is never prefilled. It is kept with the review and written to the account's
audit trail with your name, the time, the value before and the value after. Verifying,
undoing a verification, notes and **More details** change nothing a reader of the
results can see, and ask for no reason.

### More details

**More details** holds, for that one answer:

- **Review level**: your own account's list, if it has one. This is *not* a computed
  score: an account adds, renames and reorders its own levels from **Labels** in the
  dashboard sidebar, under a group named **Review level**. If the account has not
  defined any yet, the panel says so and links there. **Not set** clears a level you
  set earlier.
- **Hashtags**: space-separated, for grouping findings across responses.
- **Include in export**: on by default. Turn it off and your note and hashtags on that
  answer stay inside the tenant: they appear in no export, in any format. The answer
  itself, and your verdict on it, still export (a file that presented your correction as
  the respondent's own answer would be worse than one carrying no review at all), and
  each row reports how many notes were withheld, so the file says what it is not
  carrying.
- **Share with the participant**: see [below](#share-feedback-with-the-participant).
- **Locked**, for managers (see [Locking](#locking-managers-only)).
- the **question key** (what the export columns are named after);
- the answer's **evidence** files, if any;
- its **change history**: the respondent's answer and when it was submitted, then every
  change to the answer in effect, in order, with who made it, when, the value it set and
  the reason given. A change that put the respondent's own answer back says so. Changes
  made before reasons were required show that no reason was recorded.

**Save details** saves the review level, hashtags, export and sharing together. It never
changes the verdict, the answer in effect or the note.

### More

**More** in the top bar opens a panel beside the question, with **Close**:

- **Notes on this response** and **Notes on this assessment**, each with its count;
- **History**: everything recorded on this response (verdicts, corrections with their
  reasons, locks, shares, a reopen and what came back);
- **Answers across responses**: how every submitted response answered the question on
  screen, with counts, this response's answer marked (only for auditors whose role
  carries it);
- **Reopen for correction** (see [Reopen a response](#reopen-a-response));
- **How reviewing works**: the rules in four lines.

**Answers across responses** counts submitted responses only, never drafts, and for a
question answered in the respondent's own words it shows how many answered but never
what they wrote. A grid question (a dropdown or dynamic matrix) with any text, comment
or file column counts the same way, and a choice question lists only its own choices: an
"Other" description someone typed is counted, never shown.

### Attachments on an answer

Where a question asked for an uploaded artefact (a policy document, a scan, a
screenshot), the card names each attachment, and **More details** lists them with a
**Download** button beside each, instead of printing the stored value.

You can download an attachment because you are **assigned to review this response**,
not because you hold a file permission: any auditor or campaign manager who may read
this response's review may download the evidence it cites. The grant is exactly that
narrow: it reaches the artefacts cited by the answers you are reviewing and nothing else
in the account, and a file that this response does not cite is not found from here even
if it exists elsewhere. A cited file must also be that respondent's own upload: because
the respondent writes the answers, a citation that names somebody else's file (another
respondent's evidence, a coordinator's upload) is not found either.

Downloads are recorded. Each one writes an entry to the account's activity log and a
`file.downloaded` event, identical to a download taken from the files browser.

An attachment added before the platform stored evidence files separately shows its name
with *no downloadable copy* beside it. There is no stored artefact to fetch; the
respondent's answer recorded only the file's name.

## The three kinds of note

The workspace keeps three kinds of note, and names them the same way everywhere:

- **Note on this answer**: about one question of one response. Add it from the card's
  own button or inside the correction panel, with or without a correction. The campaign's
  managers and auditors see it; the participant sees it only if you share that answer
  with them.
- **Notes on this response**: about the whole response, not a single answer. From
  **More** in review mode.
- **Notes on this assessment**: about every response to the assessment, not a single
  response. From the **Notes on this assessment** button on the responses screen, or
  from **More** in review mode.

Notes on a response or an assessment are seen by the campaign's managers and auditors
and never by a participant; they are also listed in the managers' **Shared notes
feed**. Each note shows who wrote it and when, and *edited* if it was changed later.
**Your own** notes have **Edit** (change the text in place, then **Save** or **Cancel**)
and **Delete**, which asks first, under the note: *Delete this note? This can't be
undone.* Nobody's notes but your own offer either. Deleting a note is recorded in the
account's audit trail (that a note was deleted, never its text).

## Locking (managers only)

A manager can tick **Locked** under an answer's **More details**. Once locked, other
auditors see the card greyed out and its **Verify and next**, **Correct**, **Note on this
answer** and **Save details** buttons disabled, explaining themselves on hover: *A
manager has locked this answer*. Managers can still
edit it. Use this to freeze a moderated verdict.

## Reopen a response

**Reopen for correction**, under **More**, puts a completed response back into an
answerable state so the participant can add or correct answers. It asks **What needs
correcting?** first, and at least five characters are required. That sentence is not a
note to yourself: it is the whole of what the respondent receives, which is why the
dialog says so. **Reopen and notify** commits it.

If the respondent has no address, or no link can be minted for them, the server refuses
rather than reopening silently, and offers a second dialog (*Nobody can be notified* →
**Reopen for correction here**), so you can still reopen it knowing nobody was told.
Either way the response is marked *Reopened*, and it cannot be undone from here.

## Share feedback with the participant

Tick **Share with the participant** under an answer's **More details** and choose **Save
details**. That answer's verification status, your corrected value (if any) and your
note on the answer become visible to the person who submitted it. Answers you do not
tick stay internal.

What the participant sees, and where:

- On the assessment's completion page, under **Reviewer feedback**, with the question's
  own text rather than its internal key. Statuses are shown to them as **Reviewed**,
  **Verified** or **Adjusted**.
- An email, sent once per response, when the response goes from "nothing shared" to
  "something shared". Sharing more answers afterwards does not send another email; nor
  does re-saving an already-shared answer. Un-sharing everything and sharing again sends
  one more at most once a day per response, so the respondent's inbox cannot be flooded.

Sharing is its own permission, **Share a review with the respondent**. Every built-in
auditor role carries it; a role without it does not show the **Share with the
participant** box, and the save is refused if the field is sent anyway.

**Tell people this:** the link in that email only works in the browser the assessment was
completed in. The response is recognised by a key stored in that browser, not by a
sign-in. The email says so, but it is worth repeating in your own message. If the
participant has changed device, they cannot reach the feedback and you will need to
relay it another way.

## Managers: the shared feed

Managers get a second tab, **Shared notes feed**, listing every shared note across the
account with campaign, scope, text, author and time, plus a search and a **Reload**
button.

## Export your responses

On the responses screen: pick **Excel (.xlsx)**, **CSV (.zip with notes and codebook)** or
**JSON** and choose **Export all responses**. The file downloads immediately and covers
every response assigned to you, across all your assignments (for a manager, every
assignment in the account): it is the same caseload file as before, not only the
assessment on screen.

A CSV file holds a single table, so the CSV choice downloads a `.zip` of three:

- `submissions.csv`, one row per response;
- `notes.csv`, the Notes table described below;
- `codebook.csv`, one line per column of `submissions.csv`, with the question text, the
  row or choice label and what the column holds.

The three open correctly in Excel, accented characters included. A text cell that begins
with `=`, `+`, `-` or `@` and is not a number is written with a leading `'`, so that a
spreadsheet shows what the respondent typed instead of running it as a formula; the Excel
and JSON files carry it unchanged.

Every row names its campaign (`campaignId` and `campaignName`), and `campaignClosedAtIso`
gives the moment that campaign closed, blank while it is open.

It carries your review, not just the raw answers:

- the answer column holds **your corrected value** where you corrected an answer (the same
  value the scorecards use), and `ans:<question>::original` holds what the respondent
  actually wrote, but only where your correction changed it;
- `rev:<question>::status`, `::modifiedApplied`, `::locked`, `::reviewedAtIso`,
  `::reviewerKey`, `::reviewLevel` and `::reviewLevelLabel` describe each review, and a
  question nobody reviewed gets no such columns at all; `::reviewLevel` is your account's
  own label id, `::reviewLevelLabel` its current display text;
- each row counts its reviews, including how many notes were **withheld** by *include
  in export*;
- a **Notes** table (a sheet in XLSX, a `notes` section in JSON, `notes.csv` in the CSV
  download) carries the notes and hashtags you did permit.

The answer columns are named after the questionnaire's own item names:
`ans:<question>` for a single answer, `ans:<question>::<row>` for each matrix row, and
`ans:<question>::<choice>` for each choice. A checkbox choice holds 1 when it was
selected and 0 when not; a **ranking** choice holds its position in the respondent's
order (1 is first), and is blank when the respondent did not rank it.

The columns follow the questionnaire's order, and each question's `::original` and
`rev:` columns sit right after its answer, so a correction is next to the answer it
corrects. When two assessments in the file use one question name for different
questions, each gets its own columns, named after its assessment key: `ans:q1@CORE`
and `ans:q1@SSFPB`. The same question in both keeps a single column.

The respondent, label and metadata columns come from the respondents whose responses are
in the file, never from the rest of the campaign's roster.

Reviewers are identified by a stable key, never by email address.

## The restricted assessment auditor

The **Restricted assessment auditor** is a role for a reviewer from outside the
organisation running the campaign (for example a partner department reviewing the answers
to its own question set), who must see that one assessment and nothing else.

What the role can do, on the **one assessment** it is granted on:

- read that assessment's **submitted** responses in the auditor workspace, with the
  question text and the evidence each response cites;
- record verdicts, review levels, hashtags and notes, and edit or delete her own notes;
- correct answers (each change with a reason, as above);
- download an export of its caseload (**Export all responses**, as above): that
  assessment's submitted responses and nothing else. No draft, reopened response, other assessment,
  other campaign or roster column enters the file, whatever the request asks for.

What it cannot do, by any route, link or edited address: see the campaign's other
assessments, any other campaign, the roster, reports, scorecards or the account's
report definitions, the campaign-wide analytics or **Answers across responses**, the
account's notifications, reopen or lock a response, delete a note somebody else wrote,
or open a respondent's draft. A response that a manager has reopened leaves its list until the
respondent submits it again. Its sidebar shows the Auditor workspace and nothing else
(plus **My assessments**, when she has questionnaires of her own to answer), and any
other page of the account takes it there. It cannot join the account or one of
its campaigns by passcode either, and it cannot upload files: an administrator grants any
wider access explicitly. A participant role (being invited to answer a questionnaire too)
does not widen it: a reviewer whose other roles grant nothing in the console stays
restricted. What it reads of the account itself is what its screens draw: the name, the
branding and the languages.

**Every capability is a permission you can untick.** The role's permissions are listed,
in words, in the role and policy editors: view the assessment's responses, read, write
and edit notes, record a verdict, correct an answer, share a review with the respondent,
and download the answers. To give a different reviewer less, **duplicate** the role
(**Team & Access**, **Member roles**) and untick what they should not have, then grant the
copy on one assessment as below. The copy keeps the restriction: the editor shows
**Restricted reach: one assessment only**, read-only, and it cannot be removed. A
restricted role only ever uses the permissions in that list; adding any other permission
to it, or attaching another policy, gives it nothing.

**Granting it.** On **Team & Access**, open **Member invitations** and choose **Invite
member**. Choose the role, and the dialog moves **Access** to **One assessment**: pick the
campaign, then the assessment. The role is only ever granted on one assessment, so the
whole account and a campaign are not offered for it (and **Add member**, which grants at
account level, does not list it). The invitee may use any email address, including
another organisation's; when they accept, they land in the Auditor workspace. For
somebody who is already a member, use **Assign role** on **Member assignments** with the
scope **Assessment**. Anyone who may grant the Assessment Auditor role may grant this one.

**Removing it.** Revoke the assignment on **Member assignments** (or a campaign's
**Auditor assignments** sub-tab). Access ends on the reviewer's next request. When that
was the reviewer's last role in the account, their membership ends with it, and the
activity log records both the revocation and the end of the membership. A reviewer who
was already a member before the restricted role was granted keeps the membership and
loses only the role. A restricted role, or a copy of it, cannot be deleted while anyone
holds it: revoke its assignments first. Grants and revocations are recorded in the
account's activity log.

## Limits and gotchas
- *Include in export* governs the export only. It hides **your note and your hashtags**
  on that answer, and nothing else. It does not remove an answer from scoring: a
  response you excluded still counts towards every scorecard, the maturity profile and
  the published results, and the answer, your verdict and any correction still export.
- A *modified* verdict with no corrected value (possible on older reviews) changes no
  number, and can still be verified. `rev:<question>::modifiedApplied` in an export reads
  1 only where a corrected value was actually substituted.
- The review level never changes a number either: it does not touch scoring, the
  maturity profile or any published result, the same as *Include in export*.
- Reviewing does not change the participant's stored answers; a correction sits next to
  the original, and the original is always shown beside it.
- The question text and the section titles in the review are read from the assessment's
  template in its default language.
- Sharing is per answer and irreversible in the sense that the participant may already
  have read it before you un-share.
- An attachment download needs a storage backend configured for the deployment. Where
  there is none, the button reports that there is nothing to download rather than
  claiming a download happened.
- The download link is short-lived by design (about two minutes). It is minted when you
  press the button and cannot be copied out and kept.

## Related pages
- [Add people to a campaign](/operate/tenant-admin/add-people-to-a-campaign)
- [Reports and scorecards](/operate/tenant-admin/scorecards)
- [What other people can see about you](/use/participants/participant-data-visibility):
  the participant's side of this page.
