Metrickle

Surveys, feedback and studies

Last updated

Numbers tell you where people drop off. Surveys, feedback and studies tell you why. Metrickle keeps all three next to your analytics, so every answer and every report links to the journey behind it, and you can see whether people using an accessibility setting are the ones struggling.

Each app has three research pages in the sidebar: Surveys, Feedback and Studies. The switches for the feedback widget live in the app’s Settings, under Research.

Surveys

A survey is a short set of questions shown inside your app or site. Answers are stored as events, so you can filter journeys, replays and friction by what people said.

Question types

TypeScaleHeadline result
NPS (0–10)0 to 10NPS, from −100 to +100
CSAT (1–5)1 to 5Share satisfied (4–5)
Effort / CES (1–7)1 to 7Average
Star rating (1–5)1 to 5Average
Multiple choiceUp to 12 choices, one or several answersCount per choice
Open textFree textLatest written answers

A survey has between 1 and 10 questions. Each question can be required or optional, and scale questions can have labels for the lowest and highest score.

Create a survey

  1. Open your app and choose Surveys.
  2. Choose New survey and give it a name.
  3. Work through the steps: Purpose, Moment, People, Questions and Review.
  4. Use Preview at any point to see the real survey on the page. Preview answers go nowhere.
  5. On Review, look at anything flagged under Checks, add a thank-you message if you want one, choose a Status and save.

Keep it short. One score and one “why” question gets the most answers.

Start with its purpose

On the Purpose step, choose the closest Reason for asking:

ReasonWhat it sets up
Whether a release made something harderAn effort question and a “why”, asked of half of the people who match. Choose the release under What it’s about to ask only people on its version and its patches (on the web, only if your app sets appVersion). A release from the last week is chosen for you.
Why people give up at a step“What stopped you from finishing?”, asked when someone leaves a form unfinished.
How hard a task feltAn effort question and “What was the hardest part?”, asked right after the task’s success step. Choose the task under What it’s about, or choose the event on the Moment step.
What people are missing“Did you find what you came here for?”, asked on the third page view in a visit. Set the page on the Moment step to ask only there.
Something elseNothing. You set every step yourself.

Then write what you want to find out, and the decision the answers will inform. The results open with them. If the answers can’t change a decision, it’s usually too early to ask.

Under What it’s about, you can choose a release, an open case, a protected task or a goal. On a new survey, it fills in what fits the reason: the moment for a task, case or goal (right after its success step, when you’re asking how hard it felt or about a release), and who to ask for a release (people on its version). Its link shows above the results.

Each reason fills in the next steps for a new survey. Change anything that doesn’t fit. Once you’ve edited the questions or the timing, choosing another reason keeps your changes; choose Use the suggested setup to replace them.

Every survey you start in the dashboard ends after 14 days unless you change the end date. Drafts made by an AI assistant or the API run until paused unless they set one.

The purpose is for your team. It’s never shown to the people answering.

Checks before it goes live

The Questions and Review steps check the survey as you write it, and flag:

  • no purpose, or no end date,
  • more than three questions, or no “why” question,
  • a question that suggests an answer, seems to ask two things, or is over 15 words,
  • a required open-text question,
  • a long list of choices without “Something else”,
  • a survey that asks as soon as someone is eligible, or a page or event that hasn’t been chosen.

Checks are advice, and you can save with any of them showing, except two: a blank question, and a page or event that hasn’t been chosen.

Choose who sees it

On the People step, leave everything open to ask everyone, or narrow it:

  • Platforms: none selected means every platform.
  • Only people using: show the survey only to people using an accessibility setting, such as a screen reader, keyboard navigation or large text. Use this for research with people who use a screen reader or other assistive technology.
  • Only signed-in people: people your app has identified.
  • App versions (optional): comma separated. 2.4* matches every 2.4.x version.
  • Share of eligible people (%): lower it to ask fewer people on busy pages. A given person is either always in the share or always out of it.

Choose when it shows

On the Moment step:

  • Show: As soon as eligible, On a page or screen (use * to match a path prefix), After an event or After something happens.
  • Delay (seconds): wait so you don’t interrupt the moment you’re asking about.
  • Frequency: Once per person, Until answered (asks again after a dismissal, every so many days) or Recurring (asks again every so many days).
  • Start (optional) and End: leave Start empty to start when the survey goes live. Leave End empty to run until you pause it.

Ask right after something happens

Ask right after the moment you want to understand, while it’s fresh. After something happens shows the survey after someone:

  • sends a custom event, such as billing_toggle,
  • hits a friction signal, such as a dead click ($dead_click), a rage click ($rage_click), a u-turn ($u_turn), up-and-down scrolling ($scroll_thrash) or a form abandon ($form_abandon),
  • or views a page ($pageview).

You can narrow it to:

  • one page, such as /pricing (use * to match a path prefix),
  • one element, form or destination, named the way friction hotspots name it,
  • the Nth time in a visit, such as the third time someone flips the billing toggle. Without it, the survey shows the first time.

For example: right after a dead click on the “Web” tab on /, ask “What did you expect to happen?”

This needs the web or React Native SDK 0.3 or later. Older SDKs and the iOS, Android and Flutter SDKs skip these surveys. See Install Metrickle.

You rarely need to set this up by hand. On Friction, Converters, Funnels and Goals, each finding has a Survey them link that opens a draft with the right moment already filled in. See ask them why.

Draft, Live and Paused

StatusWhat it means
DraftSaved, but nobody sees it yet.
LiveShows to people who match the audience and timing.
PausedStops showing. Answers so far are kept.

Putting a survey live is a decision a person makes in the dashboard. It needs the permission “Put surveys live, change or delete them”, which Owners and Admins have. Someone who can only draft surveys sees Save draft instead, and can’t change a survey while it’s live. A connected AI assistant can draft a survey for you, but never put it live.

Each plan limits how many surveys can be live at once. See plans and billing.

Surveys need to remember who has already answered, so they never show to visitors in cookieless mode. If most of your visitors are cookieless, plan a study instead: recruit through a link you share yourself.

Read the results

Select a survey to open its Results. If it has a purpose, it opens with what you wanted to find out, the decision, and a link to what it’s about. The top row shows Shown, Responded, Response rate, Completed (answered every question) and Dismissed. Answers appear within a minute.

Each question has its own card with the distribution of answers. Scored questions are split into bands, such as NPS detractors (0–6), passives (7–8) and promoters (9–10). Next to each band are links to Journeys, Replays and Friction, filtered to the people who gave that answer. Open-text answers each link to the person’s journey.

Feedback

Feedback is what people send you when something goes wrong. Every report carries the page, device, app version, the accessibility settings in use and a link to what the person did before sending it.

The feedback widget

People choose what kind of feedback they’re giving. Each category shows in the inbox with a short label:

What the person choosesLabel in the inbox
Something’s brokenBroken
Something’s confusingConfusing
An idea or requestIdea
An accessibility barrierAccessibility
Something elseOther

To set up the widget:

  1. Open your app, choose Settings and find Research.
  2. Under Accept feedback from, choose the platforms that can send feedback.
  3. On web, tick Show a feedback button on web to add a floating button. Set its Button label and Position. Without the button, open the dialog from your own interface with metrickle.feedback.open().
  4. Leave Allow screenshots on to let people attach a screenshot. The browser asks permission every time, and people can highlight or hide parts of the image before sending.
  5. Choose Save research settings. Changes reach your app within a minute.

On iOS, Android and Flutter, feedback is available through the SDK. See Install Metrickle.

A problem report counts against a task. If someone finishes a task and then reports a problem within 7 days, the completion is taken back, so a task’s success rate reflects what people actually got done. An idea never takes anything back. A report tagged as an accessibility barrier counts against the setting the person was using. See protected tasks.

Triage the inbox

  1. Open your app and choose Feedback.
  2. Filter by status (New, Triaged, Resolved or Any status) and by category.
  3. For each report, set the status with the New, Triaged and Resolved buttons.

Each report links to the person’s Journey. If they were recorded, a Replay link opens the session, and a Screenshot link opens what they attached. Replays and screenshots need the permission “Watch session replays and see feedback screenshots”, which the Viewer role doesn’t have.

Members, Admins and Owners can change a report’s status. Viewers see the status but can’t change it.

Support tickets in the inbox

Connect Zendesk or Intercom and new tickets and conversations come into Feedback alongside in-app reports, each linking back to the original in your support tool. Metrickle sorts them into the same categories using tags and keywords, and can match the person who wrote in to someone in your analytics. A matched ticket can take back a task the person finished in the 7 days before, in the same way as an in-app report. You connect a support tool on the app’s Integrations page. See integrations.

Create an issue

With an issue tracker connected, each report has a Create issue link. Metrickle drafts the issue from the report, you edit anything you want, and it goes to the tracker connected to the app. Metrickle measures the result 7 days after the issue is closed.

Studies

A study is the deeper dive after a survey: you talk to the people who answered, or watch them try a task. Open your app and choose Studies, then New study.

Moderated or unmoderated

KindHow it works
Moderated sessionPeople book a time from your availability, and you run a video call with a session guide.
Unmoderated testPeople try tasks on your live site on their own time. Metrickle detects success, asks how easy it was and records the session.

The study setup has these steps: Goal, Tasks, Participants, Availability (moderated only) and Review.

  • Goal: pick the kind of study, write a Hypothesis and, if you want, link the Live task it explains or Live funnel it explains. Results then put the live numbers next to the study’s.
  • Tasks: for an unmoderated test, each task shows one at a time in a panel on your site. Describe the goal in the participant’s words, without naming the buttons or pages to use. Add a success step (a page or event) so Metrickle can tell when it’s done. After each task, people are asked how easy it was on a 1 to 7 scale. For a moderated session, the tasks are your session guide.
  • Participants: the Intro for participants, an optional Incentive (Metrickle never handles payment), the number of Participants to stop at and, for an unmoderated test, the Start URL on a site running the Metrickle web SDK. You can also ask to record their session, masked like your replay settings. People can say no and still take part.
  • Availability: your time zone, weekly hours, session length, break between sessions, minimum notice, how far ahead people can book, sessions per day, and the meeting link and host name participants see.

A study is Draft, Recruiting or Closed. Only a Recruiting study takes new people. Closing it keeps the results.

Recruit participants

There are three ways in:

  • From a survey: on a survey’s Questions step, under Follow-up, choose the study and write the invitation. After their last answer, people are asked if they’d take part. You can invite everyone who finishes, or only people with a matching answer, such as NPS detractors. Nobody is invited while the study isn’t Recruiting.
  • Recruiting link: copy it from the study’s setup and share it. Anyone with the link can join while the study is recruiting.
  • By hand: on the Participants (or Sessions) tab, choose Invite someone to create a personal link for one person. Metrickle doesn’t email it, so copy it and send it yourself.

When someone books a moderated session, they’re asked whether there’s anything you should do to make it work for them, such as captions, a screen reader friendly setup, extra time, breaks or a sign language interpreter. Their answer shows under Access needs in the participants list. Names, emails, access needs and survey answers are personal data, so they need the permission “See study participants' details”. The Viewer role doesn’t have it.

For a moderated session, open Session guide next to a participant to take notes on each task and write a summary while you run the call.

Read the results

The Results tab shows how recruitment went (Invited, Opened, Booked or Started, Completed, Cancelled and, for moderated studies, No-shows) and a card for each task with its success rate.

Each task is broken down by accessibility setting. A setting is marked Clearly worse when its success rate is 15 points or more below the task’s overall rate, once it has at least 3 attempts. Below that, it says Too few attempts to say.

Record a verdict

Under Hypothesis, the study’s results sit next to what happens in production for the task or funnel you linked. When you’ve seen enough, answer Is the hypothesis supported? with Supported, Refuted or Inconclusive, add why, and choose Record verdict. Your team sees the verdict and who recorded it.

Who can do what

ActionPermissionRoles that have it by default
See surveys, feedback and heatmapsSee surveys, feedback and heatmapsOwner, Admin, Member, Viewer
Draft surveysDraft surveys, and change draftsOwner, Admin
Put surveys livePut surveys live, change or delete themOwner, Admin
Change feedback statusChange feedback statusOwner, Admin, Member
Draft studiesDraft studiesOwner, Admin
Recruit for, close and record verdicts on studiesChange, recruit for and close studies, record verdictsOwner, Admin
Change the widget and other research settingsChange research settingsOwner, Admin

A custom role can include any of these. See team and access.