Skip to content
Sign inRequest access

Insights: voice of customer

Insights is a single screen that answers one question: what are people telling us? It gathers the numbers that normally live in three separate places (the feedback queue, your surveys, and your announcements) and lays them side by side so a product lead can read the pulse of a space in one glance. This screen is read-only: one request goes out, and nothing on it changes anything. Read-only

Open it from Insights in the space navigation. There is also a per-project version: open a project and choose its Insights tab. That tab is a subset, not a copy: it carries three of the four headline counters (Feedback, Open, Resolved), the 14-day trend, and the status mix, all narrowed to that project. Top issues, Survey health and Announcement engagement stay on the space-level screen. On plans that include AI the project tab also carries one panel the space-level screen does not — Weekly themes, described at the end of this page.

Headline counters

Four totals across the top: Feedback, Open, Resolved, and Survey responses.

Feedback: last 14 days

A daily bar chart of new feedback over the trailing two weeks, with empty days shown as zero.

By status

A bar for each of the five statuses, showing the current mix.

Top issues

Up to eight labels ranked by how often they appear across feedback and survey responses.

Survey health

One row per survey that has responses: its headline metric and response rate.

Announcement engagement

Total Seen, total Clicks, and the click-through rate across your announcements.

The four counters are cumulative totals, not a windowed view:

  • Feedback: every report the viewer can see, across all statuses.
  • Open: everything that has not reached a terminal state. It is derived as total minus resolved minus archived, so a report sitting in waiting is counted as open rather than falling out of both columns.
  • Resolved: reports in the resolved status.
  • Survey responses: every response received across the surveys in scope.

These are counts of everything you can see. Only the trend chart below is time-bounded.

This is the one time-bounded panel: a bar per day for the trailing 14 days, with days counted in the business timezone (Riyadh) rather than UTC, so “today” on this chart is the same today as the rest of the product. Days with no new feedback are drawn as zero rather than skipped, so the shape of the last two weeks is honest: a quiet stretch reads as a flat run, not a gap.

On the space-level screen the plot carries a real axis: the peak value at the top of the gutter and zero on the baseline, so the bars are a measurement and not just a shape. Hover a bar for that day’s date and count, and the whole chart carries a spoken summary (total, peak, and the busiest day) for anyone who cannot hover fourteen bars one at a time.

The project tab draws the plainer version of the same chart: the window’s total above it and the first and last dates beneath it, but no value axis, no spoken summary, and a hover that shows the raw date and count (2026-09-23: 4). Use the space-level chart when you need to read values off the bars, or when you are not reading with a mouse.

A horizontal bar for each status in the lifecycle (new, in_progress, waiting, resolved, archived), with the count beside it. The list is driven by the model’s own status list, so a status added later cannot quietly go missing from this panel.

This is the current standing mix (all-time, not windowed), so it complements the trend chart: the trend tells you how much is arriving, the status mix tells you how much is still open. See Status lifecycle for what each status means.

This is where the connective tissue shows. Because a single space-level label taxonomy tags both feedback and survey responses, “top issues” can rank a theme by everything that mentions it, not just tickets. (Labels belong to a space: two spaces in one account have separate label sets, so this panel ranks a theme within the space you have open.) Each row is a label with its colour as a dot, ranked by its combined count, and reads:

Checkout · 12 reports · 4 responses

Up to eight labels appear, highest combined count first; labels with nothing attached are omitted. If no labels are applied yet, the panel tells you so: apply labels on feedback and survey responses to populate it. Labels are managed on the Labels, priority & assignment page.

One row per survey that has at least one response (surveys with none are hidden). Each row shows the survey’s title, its project and type, a headline metric that depends on the type, and a response rate:

Survey type Headline metric
nps Net Promoter Score
csat, ces Average score
form Number of responses

The response rate is the share of times the survey was presented that produced a response, capped at 100%. A metric or rate that can’t be computed yet shows as — rather than a zero or a fabricated number. For how each score is calculated, see Surveys.

A single strip of three numbers summed across the announcements in scope: Seen, Clicks, and CTR: clicks divided by views, as a percentage. When nothing has been seen yet, CTR shows — instead of dividing by zero. More on publishing and tracking announcements: Announcements & changelog.

Every panel is built from the projects the viewer can actually reach, and the same rules apply to all of them: feedback, labels, surveys, responses, and announcements alike. One screen, one scope.

  • Account isolation. Every model in the rollup carries the account scope, so the page can only ever count the current account’s rows. Another tenant’s data is invisible by construction.
  • Space scope. Feedback, labels, surveys, survey responses, and announcements each carry the per-space scope, so the page counts only the space you have open: switch spaces and every number is rebuilt from that one. This is a wrapper, not a per-person boundary — by default everybody in the account reaches every space, and confinement happens one level down, at the project.
  • Project reach. A member attached to one project of two does not see the other’s volume, status mix, trend, labels, surveys, or announcement engagement. A project that answers not-found on its own page does not contribute a number here either.
  • Per-project role overrides. A project role that does not carry feedback.insights removes that project from the rollup (which is how an admin says “not this project’s numbers”), and asking for it by project_id returns not-found rather than the data. The space-level totals simply stop counting it. Its own Insights tab is still drawn, because whether that tab exists is decided by an account-level check, and it reports that it could not load — there is no separate “you were not given this project’s numbers” message today.

Two more rules apply to the feedback figures specifically:

  • Merged duplicates are out, exactly as they are out of the inbox this page summarizes. A report merged into another would otherwise inflate the status mix, the trend, and the label counts against a board that will not show it.
  • A plan history window applies to the feedback figures, for the same reason it applies to the inbox: this page summarizes reports the product will open. It bounds the volume, the Open/Resolved split, the status mix and the 14-day trend. It does not bound the Top-issues label counts: a report older than the window still counts toward its label there. In practice this is a narrow case: no plan in the standard catalogue combines the two, because the one tier that ships a history window (Free) cannot open Insights at all. You meet it only where a history window was put on your plan or on your account deliberately.

What it reads, and the one thing it can spend

Section titled “What it reads, and the one thing it can spend”

The space-level screen has no forms, no buttons that change state, and no bulk actions. It is served by a single GET /app/insights request and does nothing but count and display. You cannot resolve, assign, label, or delete anything from it. Go to the report or the survey itself for that. There is no drill-through from a panel either: a top-issue row and a survey row are readouts, not links.

Weekly themes: the one button, and it is on the project tab

Section titled “Weekly themes: the one button, and it is on the project tab”

The project tab carries one control the space-level screen does not: a Weekly themes panel with a Find this week’s themes button. Pressing it sends that project’s recent reports to an AI provider and asks for at most five themes, each a short name, a one-sentence description, and a count.

  • It runs only from that press — never on opening the tab, never on switching projects.
  • It spends the account’s AI credits, and the call is written to the account’s AI usage ledger. Nothing else on either Insights surface costs anything to look at.
  • The answer is a draft: it is not saved, not attached to any report, and gone once you dismiss it or reload the page.
  • It is off until your account turns AI on, and it says why it could not run — AI switched off for the account, or this month’s credits used up — rather than failing quietly.
  • Two role keys guard it: the panel is drawn only for someone who can see Insights, and the endpoint additionally demands the key that governs spending credits on a multi-report pass.

More on what the model is asked and what it is never allowed to do: AI copilot.

Insights is built to stay honest when there is little or no data. You will never see NaN, a divide-by-zero, or an invented figure:

  • No feedback in the window: the trend chart reads “No feedback in the last 14 days.”
  • No labels applied: Top issues prompts you to start labelling feedback and responses.
  • No survey responses: Survey health reads “No survey responses yet.”
  • Nothing seen yet: Announcement CTR (and any survey metric that can’t be computed) shows —.
  • The request failed: the page says the load failed. An error is never drawn as an empty space, because “we have no data for you” and “we could not reach the server” must not look the same.

When your plan includes automations, the space’s home dashboard carries one more card: Automations handled X% of triage this week. It counts the reports that arrived in the last 7 days in the projects you can see, and how many of them an automation changed without anyone undoing it, next to the number of automated actions and the share that was undone. These are counts, not estimates; a week with no new reports says so instead of showing 0%. The card links to the Automations page.