Product feedback and surveys

Instance administrators read submissions and publish surveys under Instance administration → Product feedback. It has two pages:

• Inbox lists ideas, bug reports, and feedback.
• Surveys publishes surveys and shows their responses.

Workspace admin access alone does not grant access to either page. Members see the readers described as your instance admins.

Publish a survey

Members read a survey's introduction before the first question. It must explain why you ask and what the answers will change.

1. On Surveys, select Create survey.
2. Write a title.
3. Write an introduction.
4. Select an audience: one workspace or all workspaces.

Where the instance names a research organization, the composer also asks for a purpose:

• Product surveys are for instance admins, described to members as your instance admins. They are never research.
• Research surveys are part of the research run by the named organization. Only members whose research consent is a current grant for that organization receive them. The screen labels them as research and links to the setting that withdraws consent. Their answers are research data.

A member who withdraws from the research receives no more requests to answer. If you switch the program off or rename it, every research survey for the old name disappears. You cannot change a survey's purpose after publication. Without a research organization, every survey is a product survey and no purpose choice appears.

Questions are optional unless you mark them required. Members see (optional) after every optional question. Required questions have no such label. Each question uses one of these forms:

• Free text.
• Single choice.
• Multiple choice.
• A 1–5 rating with a label at each end of the scale.
• A 0–10 recommendation question scored as a Net Promoter Score.

The recommendation question also disables keyboard answer shortcuts for the whole survey.

1. Keep most questions optional. Someone who cannot answer a required question leaves.
2. Where appropriate, include “Not applicable” as a choice.
3. Alternatively, allow Something else for an answer that you did not list.
4. Ask about what members did, not what they would do.
5. Select Preview to inspect the draft.

Preview opens the same dialog that members receive. It sends nothing. It refuses a draft that could not be published, for the same reasons. The schedule uses your device's timezone. A survey with no end stays open until you end it.

You cannot edit questions after publication. If the text needs a change, publish a new survey. Feedback and surveys explains how members receive invitations and answer or decline.

The survey's actions have these effects:

• Pause makes a survey unavailable.
• Resume restores availability within the original schedule. It does not extend the end time or reset responses or declines. If a member sends an answer after you pause, Hephaestus refuses it and keeps their draft.
• End now closes an open survey at the current time. Members receive no more invitations, but responses stay. A scheduled survey has not started and cannot be ended. To remove it, use Delete.
• Delete removes the survey and every response and decline it collected. You cannot undo this deletion.

Invite by email

Survey creation or edits never start an email campaign. Only people who opted into the matching product or research survey emails are eligible. Research invitations also require current consent for the survey's research organization. Each recipient must still belong to an eligible workspace. They must have an active account with a verified address. One person receives one invitation across workspaces.

1. From the results page, preview the email audience.
2. Explicitly request invitations.
3. For an audience larger than 1,000 recipients, repeat the action.

Each request queues at most 1,000 recipients. It does not request initial invitations that the relay already accepted. Hephaestus checks eligibility again before it sends or retries. A pause cancels pending invitations. Resume does not restore them. These actions prevent queued delivery:

• Closure or deletion of the survey.
• An answer or decline.
• Withdrawal of consent.
• Departure from the audience.

The email audience card distinguishes these counts:

• Eligible now: current subscribers who can still answer the survey.
• Already requested: people for whom an email invitation was requested, including canceled requests. A request does not prove delivery.
• Accepted by relay: initial invitations that SMTP accepted, not invitations that recipients opened or answered. Reminders do not increase this count.
• Remaining: eligible people who have never had a request. This also includes people whose initial invitation was canceled or expired before relay acceptance and can be explicitly requested again.

The send result counts invitations newly queued in that batch. It does not count email still awaiting delivery. Email requests and acceptance are separate from the in-app invitation counts used for completion rates. Emails contain no tracking pixels or answers.

One optional reminder

Send one reminder is off at first for each send action. If selected, it can schedule one reminder for each new recipient. The reminder is at least 72 hours after relay acceptance of the initial invitation. It is not a repeated campaign. An answer, decline, unsubscribe, audience departure, or loss of research consent prevents it. The survey must still be open.

Email requests expire seven days after the request or when the survey closes, whichever occurs first. A pause cancels pending initial invitations and reminders. Resume neither restores those requests nor removes their Already requested history.

After resume, a new send action can explicitly request a canceled or expired initial invitation that the relay never accepted. It cannot repeat an accepted initial invitation. It cannot restore that invitation's canceled reminder.

End-of-survey summaries

Instance admins can separately opt into Survey summaries in their email notification settings. A survey's end summary contains aggregate counts of people shown the in-app invitation, responses, and declines. It also links to the instance admin console. It contains no individual answers. Subscription to survey invitations does not subscribe an instance admin to summaries.

Only active instance admins with verified email addresses receive summaries. Hephaestus checks eligibility again before delivery. It prepares summaries only for surveys that ended within the last seven days. It does not prepare summaries for older surveys after downtime.

Hephaestus prepares a summary once for a survey's end time. It does not prepare one per response or subscription change. A change to the end time invalidates a queued summary for the old schedule. It permits a new summary when the revised end occurs.

Read survey results

1. Open a survey from the table.
2. Select Export CSV to download responses, up to 10,000 rows.

The results show who received an invitation, responded, or declined. They also show a summary per question and every response with its submitting account and workspace. Voluntary responses are not a representative measurement of satisfaction. Declines are not negative ratings. Skipped questions are not zeroes.

Do not reuse product responses for research. A research survey's results page reminds you whose data they are. The default optional email budget allows at most 200 SMTP attempts in one 24-hour refill interval. All optional categories and retries share this budget. A 1,000-person invitation batch needs at least five intervals without other traffic or failures. The batch size does not promise a delivery speed.

1. Before you invite a large audience, plan the survey deadline and approved relay capacity.

Act on feedback

Inbox lists ideas, bug reports, and feedback. Each item identifies its sender, workspace, and the Hephaestus release that received it. It includes the page and browser only if the sender chose to attach them. Mark resolved removes an item from the open list. Reopen returns it to that list. Neither action notifies anyone.

To receive email about new submissions:

1. In your personal settings, opt into product-feedback email.

To handle a report:

1. Sign in.
2. Use Inbox to read the report text or attached context.
3. Follow your team's process for reports.
4. If necessary, follow up with the sender.
5. Resolve or reopen the item.

Emails are sent as soon as the submission commits, subject to capacity and relay availability. They never include report text or attached context. Only currently active instance admins with verified email addresses receive them. Resolution is private administration, not a reply.

An idea or bug that concerns every instance is worth reporting to the Hephaestus project. Nothing reports it automatically.

1. Before you create a public issue, remove private details.

Product feedback can identify questions worth study. Collection for product improvement does not authorize reuse of identifiable reports or quotations as research data. A general research opt-in does not permit reuse of every submission for a different purpose.

1. Use an explicitly research-purpose survey for research responses.
2. Before research collection or reuse, agree the research purpose, lawful basis, participant notice, retention, and safeguards with your institution.

Handle personal data

Submissions link to accounts and stay on your instance. Users can attach the displayed page path and their browser's user agent. Submissions never attach query parameters, page contents, or logs. They always record the Hephaestus release that received the report. Drafts never reach the server. Feedback has a limit of one submission per account per minute.

Each account can respond to a survey once. A survey records when an account first saw its invitation. Thus, the results page can report a completion rate. Research consent covers research survey answers. The answers stay with the survey after a member withdraws. Removal requires account erasure or survey deletion.

Account erasure and workspace purge remove the corresponding submissions. Account erasure also removes that account's survey email-invitation records. It clears that account's requester attribution on other people's invitations. Survey deletion removes its invitation records and responses. Workspace purge removes records that belong to that workspace. These operations cannot recall email that a relay already accepted.

Retained invitation records have no automatic age-based deletion.

1. Include these records and product feedback in your retention and rights-request procedures.

The personal-data map describes erasure coverage.