Surveys
Surveys are short, focused questions the uMarker widget renders in-page, alongside bug capture. Ask a returning visitor one thing: how likely they are to recommend you, how satisfied they were, how easy something felt. Then read the answer as a score you can act on. A survey collector and the bug-capture widget coexist on the same page: separate requests, separate shadow roots, neither one waiting on or disabling the other.
Surveys are scoped to the project’s account: responses never cross tenants. They live in the left navigation under Surveys, inside a space.
Start from a template
Section titled “Start from a template”Creating a survey starts in the template gallery, not with a blank editor. Each template is a ready-made bundle: the question wording, the thank-you line, an accent colour, and — depending on the engine — either the two scale labels or the form’s questions. A picked template is a survey you could publish as-is.
| Template | Category | Engine |
|---|---|---|
| Net Promoter (NPS) | Loyalty | nps |
| Satisfaction (CSAT) | Satisfaction | csat |
| Effort (CES) | Effort | ces |
| Website feedback | Website | csat |
| Emoji satisfaction | Satisfaction | form |
| Star rating | Satisfaction | form |
| Product feedback | Product | form |
| Feature request | Product | form |
| Feature vote | Product | form |
| Product-market fit | Product | form |
| Onboarding check-in | Growth | form |
| Exit / churn | Retention | form |
| Bug report | Support | form |
Start blank is always there too: it seeds only the plain preset for the engine you pick.
The catalogue is curated by the uMarker operator, so the gallery you see is the live list rather than a hard-coded one. A template seeds its text in every language uMarker has it in, not just the one your dashboard is set to. Your visitors read that text, and half of them would otherwise get it in the wrong language.
The four engines
Section titled “The four engines”Every survey runs on one of four engines. The template picks it for you; Start blank asks. The engine fixes the answer shape and the headline metric. The wording is always yours to edit.
| Engine | Scale | Headline metric |
|---|---|---|
nps |
0–10 |
NPS score |
csat |
1–5 |
Average |
ces |
1–5 agreement |
Average |
form |
Up to several questions | Per-question results |
The single-question engines (nps, csat, ces) also carry two endpoint labels for the scale. For NPS these default to Not at all likely and Extremely likely.
Form question types
Section titled “Form question types”A form survey is a short list of questions, each individually required or optional:
| Type | What the visitor sees | How it is counted |
|---|---|---|
rating |
A 1–5 row of numbers |
Average + distribution |
stars |
A 1–5 row of stars |
Average + distribution |
emoji |
A 1–5 row of faces |
Average + distribution |
choice |
Single-select from your options (up to 12) | Count per option |
text |
A short free-text box | The answers, listed |
rating, stars, and emoji are the same 1–5 answer wearing three different faces, and they aggregate identically. Branching, matrix, ranking, and multi-select questions are not part of this version.
Create and publish
Section titled “Create and publish”-
Go to Surveys → Create. Pick the project the survey runs on, then a template (or Start blank and choose an engine). The survey is created as a
draft, seeded from that bundle. -
Edit the question (up to 255 characters per language) and the thank-you message shown after a visitor answers. For a single-question engine, set the low and high scale labels (up to 60 characters each). For a form, add, reword, and reorder questions, set each one required or optional, and fill in the options of any
choicequestion. -
Set an optional title (up to 120 characters) for your own reference. The list and the results page fall back to the question text when there is none.
-
Style the card under appearance, and choose who sees it under targeting.
-
Select Publish. Only a
publishedsurvey is eligible for the widget to render.
A survey moves through draft → published → paused. Pause unpublishes it: the widget stops showing it on subsequent loads without deleting the survey or its responses. Re-publish at any time. Deleting a survey removes it and every response it collected, for good — with one exception: a response you had already opened as a report stays in the inbox, carrying the answer’s text, because it became a ticket of its own the moment you bridged it.
Editing a live survey
Section titled “Editing a live survey”Questions carry a stable internal key, so deleting or reordering a question does not shift the answers already collected onto the wrong question. A choice question saved with no options is refused with an inline message rather than silently dropped.
Appearance
Section titled “Appearance”The editor has a live preview beside the controls: the accent, the position, the layout, dark mode, the logo, and the questions themselves all redraw as you set them. The preview draws the card, not the whole flow — it does not render the thank-you state a visitor sees after submitting, nor the privacy-notice link the widget adds when the project has a privacy URL, and it shows the top rung of a scale pre-selected where the live card starts with nothing selected.
- Accent colour: a hex colour for the buttons and the selected state. A template brings its own.
- Position is which corner the card sits in: bottom-right (the default), bottom-left, top-right, top-left. A bottom-corner card measures the feedback launcher and lifts itself clear when the two land in the same corner. A top-corner card does not, and the launcher draws above it — so if your launcher sits in the same top corner it will cover the card’s dismiss button. Put one of the two somewhere else.
- Layout:
card(a corner card),modal(centred, over a dimmed page), orinline(embedded in the page itself). - Dark: renders the card dark instead of light.
- Logo: shows the project’s widget logo on the card. The switch is hidden until you have uploaded one on the project’s widget tab.
Targeting
Section titled “Targeting”The survey editor exposes a small slice of uMarker’s targeting rules: who to show it to, and where.
- Audience:
everyone, or onlyidentifiedvisitors (those the widget knows an identity for). - URL contains: a substring the page URL must contain for the survey to appear (up to 300 characters). Leave it blank to run everywhere.
Leave both at their defaults and the survey targets every visitor on the project’s site. Targeting is evaluated in the browser by the widget; the backend serves the eligible survey and the widget decides whether the current page and visitor match.
What the visitor experiences
Section titled “What the visitor experiences”The widget requests the eligible survey and presents it if the page and visitor match. The visitor either:
- Answers it. A single tap on a score (NPS/CSAT/CES) or a completed form is submitted, and the thank-you message is shown. A form blocks submission if a required question is blank, with an inline prompt to complete it. Free-text answers are capped at 2,000 characters; a forged rating outside the rendered scale never reaches your results — on a single-question survey the whole submission is refused, and on a form that one answer is discarded while the rest of the submission is kept (unless the question was required, which refuses the lot).
- Dismisses it. Closing the survey without answering is recorded as a dismissal (counted separately from a response), and it is not shown again in that session.
A presentation is recorded once per visitor per survey, so the response-rate denominator is visitors shown rather than page views — a reload or a second tab does not inflate it. If a survey is paused or deleted while a visitor has it open, an in-flight submit is refused without disturbing the host page: nothing is stored, and the visitor sees a short error line inside the card rather than a broken page.
Reading results
Section titled “Reading results”Open a survey’s Results to see its headline metric, the distribution of answers, the response count, and the response rate. Metrics are divide-by-zero safe: a survey with no responses shows an honest empty state rather than a fabricated or NaN value.
The NPS score is the percentage of promoters minus the percentage of detractors, rounded to a whole number:
- Promoters: scores
9–10 - Passives: scores
7–8 - Detractors: scores
0–6
You also see the full 0–10 distribution and the promoter / passive / detractor counts. With zero scored responses the NPS score reads as not-yet-available.
The headline is the average score across the 1–5 scale, to two decimals, with the distribution across each rung of the scale. With zero responses the average is not-yet-available.
Each question is summarized on its own:
- A rating, stars, or emoji question shows its average and the distribution across all five rungs, including the ones nobody picked, so “nobody rated this 1” is distinguishable from “1 was never on offer”.
- A choice question shows a count per option, including options at zero. Votes are tallied by the option’s identity, so translating an option later keeps its history instead of splitting it in two.
- A text question shows the answers visitors typed: up to 200 of them on screen, with the honest total beside it. The export carries all of them.
The response rate is responses ÷ visitors the survey was shown to, to one decimal place and capped at 100% (a blocked presentation beacon can otherwise push responses over the count that was recorded). Until a survey has been presented at least once, the rate is shown as not-yet-applicable rather than 0%.
Below the aggregate, responses are listed individually, newest first and paged, so you can read the raw answers, apply labels to a response, and open one as a report.
Export responses
Section titled “Export responses”Select Export to download every response as a CSV, streamed so a large response set never times out. The file is named survey-<id>-responses.csv, one row per stored response.
id, submitted_at, page_url, scoreid, submitted_at, page_url, q0: <prompt>, q1: <prompt>, …Open a detractor response as a report
Section titled “Open a detractor response as a report”A low-score response is a lead, not just a number. From a survey response you can Create report to turn it into a ticket in the main queue. That is useful for detractors you want to chase down.
uMarker flags likely detractors for you: an NPS score of 6 or below, or a CSAT/CES/form rating of 2 or below out of 5.
The bridge is deliberately quiet and safe:
- It is dashboard-initiated: never reachable from the public survey endpoint, so submitting a response can never auto-create a report.
- It copies the account and project from the response itself, never from request input, so it can’t be pointed at another tenant.
- It is idempotent: the response records the report it was opened as, and opening it again reuses that same report instead of creating a duplicate.
- The new report is stamped
source='survey'and fires no new-feedback notification, so bridging a response never amplifies it into an alert.
The report’s comment is built from the visitor’s free-text answers, or, for a pure score with no text, a short summary line such as NPS 3/10. Survey responses are anonymous, so a bridged report is an internal record with no reply channel back to the respondent.
Who can do what
Section titled “Who can do what”Surveys do not sit behind one blanket admin switch. Each action asks for its own permission, and on an individual survey a per-project role overrides the space role in both directions — it can grant a permission the space role withholds, and withhold one the space role grants:
| Action | Permission |
|---|---|
| See the list and open a survey | surveys.view |
| Create one | surveys.create |
| Edit questions, appearance, targeting | surveys.update |
| Publish or pause | surveys.publish |
| Read results and responses | surveys.results |
| Export the CSV | surveys.export |
| Delete a survey and its responses | surveys.delete |
The Surveys list is the one screen that does not work that way: it asks for surveys.view at the space or account level before it opens at all, and only then narrows the rows to the projects where you also hold it. So a per-project grant is enough to open, edit, publish, read, export, or delete an individual survey, but not on its own enough to reach the list that links to them. Give someone the space-level surveys.view as well if they need to browse.
Labelling a response uses feedback.label — the same key the inbox uses to label a report, because it is the same taxonomy on the same queue. Opening a response as a report uses feedback.create, which exists for this bridge alone: it is permission to write a new row into the queue, and no inbox action asks for it.