# Publish results to a public link

## Purpose
Expose a campaign's completed results at a read-only web address that needs no sign-in,
and decide how much of them to show. The control is called **Public JSON link
(advanced)** and sits at the foot of the campaign's Reports page.

## Who this is for
Account administrators and campaign managers.

## Before you start

This is not how respondents see their results. Releasing each organisation's own result
is a different control on the same Reports page, and a public link is the more public of
the two: anyone with the link can read what you include, with no sign-in, and search
engines are not the only way a link travels.

To let respondents see their own result, use
[Releasing results to respondents](/operate/tenant-admin/releasing-results): that shows
every organisation only its own result. Most programmes never need a public link. Read the
disclosure choice below before you publish anything; the product asks you to confirm it
for that reason.

## Publish

1. Open the campaign and choose **Scorecards** in the header to reach its Reports page.
2. Scroll to the last card, **Public JSON link (advanced)**. (It was a **Publish
   results** button in the campaign header before; it is not there any more.)
3. Choose **Set up public link** (or **Manage public link** if one already exists).
4. Set the three switches (see below).
5. Choose **Publish public link**.
6. Read the confirmation. It says that anyone with the link can read the campaign's
   results without signing in, that this is different from releasing results to
   respondents, and what the page will include. Choose **Publish publicly** to go
   ahead, or **Cancel** to publish nothing.

The dialog then shows the public URL with a **Copy** button and an open-in-new-tab
button. The page lives at `/results/<key>`. The card on the Reports page also says
**Published: anyone with the link can read these results** while a link exists.

## The three switches

The dialog opens the block by saying what it means to save it: *"These switches are the
whole configuration. Publishing replaces the previous settings rather than merging with
them, so anything switched off here is switched off on the public page as soon as you
save."* Re-publishing is therefore a rewrite, not an amendment. Read all three before
you press **Update public link**.

**Include per-participant answers**

- Off (default). The published page carries aggregate figures only: assessment count,
  completed submission count, participant count, and completed submissions per
  assessment. No individual answers.
- On. The published page also carries each completed submission's answers.

**Anonymize participants**

Only meaningful when answers are included. On (default) it does two things: it replaces
names and email addresses with "Participant 1", "Participant 2" and so on, *and* it
withholds every answer somebody typed or uploaded. Off shows the real names, email
addresses and every answer verbatim.

The second half is the one worth reading twice. An anonymised publication carries an
answer only when your questionnaire declared every value that answer could take: a
choice list, a matrix row against its declared columns, a rating, a yes/no. Those are
published with their labels. Everything else is withheld and reported as a count instead:

| What you asked | What the public page shows |
| --- | --- |
| A choice, a rating, a yes/no, a matrix | The answer, by its label |
| A number box (a text question set to accept numbers) | No one organisation's figure. Instead: how many answered, the cohort total, the cohort average and a spread (see *An aggregate instead of a number* below) |
| A text or comment box | "Written answer, withheld to keep the report anonymous", and how many people answered |
| A grid where each column is its own question | Column by column: the scale columns publish, the written ones are named as withheld (see *A grid published column by column* below) |
| A file upload | "Uploaded evidence, withheld to keep the report anonymous", and how many people attached one |
| An answer this app cannot recognise | "Answer format not recognised, withheld to keep the report anonymous", and the count |

Nothing is partially masked. A half-hidden sentence still gives away its length and its
shape, and in a cohort of forty organisations that is often enough to identify the
author, which is the whole reason the label said "Participant 7" in the first place.

The counts are there so the withheld questions do not read as unanswered. "41 of 46
organisations described the evidence they hold" is a fact a stakeholder needs and one
that names nobody.

### An aggregate instead of a number

"How many staff work with models?" is a text box, so the answer itself stays withheld:
a staff count against one organisation in a cohort of forty *is* the organisation. But
the totals are not disclosive, and they are usually what the question was asked for. So
where your questionnaire declared a question numeric, the published page carries an
aggregate beside the response count:

- **how many people answered** it,
- the **total** across them,
- the **average**, and
- a **spread** across fixed size bands: under 10, 10 to under 100, 100 to under 1,000,
  and so on up.

The bands are powers of ten, the same for every question in every campaign, and they
are fixed in the product rather than calculated from your data. A band edge chosen from
the answers would leak the answers.

Three things it deliberately does **not** show:

- **No highest and no lowest.** Each of those is one organisation's exact answer with an
  aggregate's name on it.
- **Nothing at all below five responses.** A question fewer than five people answered
  keeps its response count and gets no aggregate. The same floor as the maturity
  profile (see *Small groups are withheld* below).
- **No band with fewer than five people in it.** "One organisation is between 10,000 and
  100,000" beside a published total points at that organisation. So the bands you see
  need not add up to the response count, and the page says so.

A question that asks for several numbers at once (a headcount *and* a budget line, say)
is aggregated field by field, so the headcount eight people answered can publish while
the budget line three people answered does not.

Two conditions have to hold, and both are deliberate. The questionnaire must have
**declared** the question numeric: a comment box everyone happened to answer with a
number is still a comment box, and its "average" would be meaningless. And **every**
stored answer must really be a number: one "about a dozen" and the aggregate is dropped
for the whole question rather than computed over the answers that happened to parse.

### A grid published column by column

A grid question (one row per discipline, one column per thing you asked about it) is
usually a mix. Four columns are scales and the fifth is "anything else?". Publishing the
whole question would publish the comments; withholding it would lose four columns of
scores to protect one column of prose.

So each column is judged on its own. The scale, rating and yes/no columns publish with
their labels, row by row. The written and uploaded columns are withheld, and each one is
named in the withheld list with its own reason and its own count, so a reader can tell
"this column was withheld" from "this column was never asked".

One shape is still withheld whole: a grid where **respondents add their own rows**
stores no row labels to publish against, so there is nothing to disclose a column
against and the question keeps its response count only.

If you need the written answers themselves, they are in the **export** (a spreadsheet
you control who receives), not on a public link.

Turning answers on *and* anonymisation off makes the dialog display a warning in amber:
"Participant names, emails and answer content will be publicly visible." Take it
literally: that publication carries the free text and the file names as typed.

The published page states which of the three modes it is in, in plain words, so a reader
knows what they are looking at.

**Include maturity profile**

The third switch, and the independent one. It publishes the campaign's aggregate
Maturity Profile: domain levels against the target. That figure carries no answer
content and names nobody, so it does not depend on either switch above: you can publish
a profile from a wave whose answers stay private. Segments with too few respondents are
withheld from it regardless (see *Small groups are withheld* below). When no profile is
available for the campaign the switch is disabled and the dialog says why.

## What the public page shows

- The campaign name and description, when the wave **closed**, and when the page was
  published. Those last two are different facts: the close date is what the numbers are
  as at, and a wave still collecting says so instead.
- Three summary cards: Assessments, Completed submissions, Participants.
- A per-assessment completion table, naming your **Active** and **Completed**
  assessments. A Draft assessment is not named and its submissions are not counted
  (see the gotchas below).
- The aggregate maturity profile, if you included it, and the public page shows **less
  of it than your own report does**. See *Small groups are withheld* below.
- The questions that were answered but not published, with a count each, if the
  publication is anonymised.
- The responses, if you included them: each answer under its own question title, with
  choices and matrix rows spelled out rather than left as codes.
- A **Print / Save as PDF** button. It prints the report without the app chrome, keeps
  the charts and tables whole across pages, and carries the campaign name and close date
  at the top. This is the document to circulate to stakeholders.
- A **Download JSON** button, which hands the visitor the whole published dataset as a
  file.

That last point matters: publishing with answers included means publishing a downloadable
dataset, not just a web page.

## Small groups are withheld

The maturity profile can be broken down by segment: by province, by practitioner role,
and by whatever else the questionnaire asks. A breakdown of a small group is not
anonymous in practice: "the four people in this territory scored 2.1" identifies four
people to anyone who knows who they are.

So there are **two thresholds**, and they are deliberately different:

| Where | Threshold | What it means |
| --- | --- | --- |
| Your own report, workbook and PDF | **3 respondents** | A segment (or a single domain inside one) with fewer than three people behind it shows as withheld rather than as a score. |
| The published public page | **5 respondents** | A segment or domain with fewer than five is **not on the page at all**: not the score, not the name, not the headcount. |

The lower figure is for you, reading your own respondents' data, where a report with
most of its segments withheld is a report you cannot use. The higher one is the
conventional floor for a published aggregate, and it applies to anyone with the link.

What follows from it:

- A segment you can see on your own screen may simply be absent from the public page.
  That is the control working, not a bug.
- The **campaign-wide** figures are never withheld. Suppression is about segments:
  a number for the whole cohort attributes nothing to anybody.
- A withheld segment leaves no trace publicly. There is no "withheld" row to click, and
  no headcount, because a headcount beside a segment name is most of the disclosure.
- Neither threshold is a setting you can change from the app. They live in the platform's
  report definition; ask your administrator if a wave needs different ones.

## Change or withdraw

Reopen **Manage public link** on the Reports page at any time. A link published before
the control moved from the header keeps working and shows as published here.

- Change a switch and choose **Update public link** to republish with the new
  disclosure level. You are asked to confirm, as for the first publish.
- Choose **Unpublish** to take it down. This needs no confirmation, because it only ever
  narrows who can read the results. The dialog then reads "Not published: results are
  private."

Withdrawing stops the link resolving, but it cannot recall copies already downloaded.

## Limits and gotchas
- Only **completed** submissions are published. Drafts never appear.
- Only assessments you have put to respondents are named: **Active** or **Completed**. A
  **Draft** assessment is withheld completely: not its title, not its id, and not its
  submissions, which are also absent from the three summary cards. So the per-assessment
  counts always add up to the campaign total. If a set you expected is missing from the
  page, check its Publish State.
- An assessment with no status set at all (the shape the legacy import left behind) is
  treated as publishable.
- Withheld answers cannot be unlocked switch by switch. The choice is anonymised (labels
  and counts) or not anonymised (everything verbatim, with names); there is no
  per-question setting.
- The public link is the only protection: there is no password on the published page.
- The two suppression thresholds mean your own report and the public page can disagree
  about which segments exist. Check the public page itself before telling a stakeholder
  a breakdown is there.
- This is a campaign-level action. There is no per-assessment or per-participant
  publishing.
- Publishing results is separate from publishing an *assessment*. One shares outcomes;
  the other opens a questionnaire. See
  [Publish and preview an assessment](/operate/tenant-admin/publish-and-preview-an-assessment).

## Related pages
- [Reports and scorecards](/operate/tenant-admin/scorecards)
- [Review campaign analytics and logs](/operate/tenant-admin/review-campaign-analytics-and-logs)
- [What other people can see about you](/use/participants/participant-data-visibility)
