Measure whether people can do what they came to do
Define a task and Metrickle reports how many people finish it, how long it takes and where the rest give up. Then use friction, journeys, surveys, replay and studies to find out why, on the same data.
| Last page reached | Chart | Attempts |
|---|---|---|
/search | 1,284 | |
/invoices/* | 590 | |
/login | 402 | |
/help | 157 |
Most failed attempts end on search. Over half of all abandoned attempts stop there, usually after a dead click on the filter chips.
From a hunch to a finding in three steps
A walkthrough on Quillmere, a fictional invoicing app with sample data. Tasks run on the events you already send, so they work on past data from the moment you save one.
Define the task
Pick where it starts and what counts as success: a page, a screen or an event. Set how long people have before an attempt counts as abandoned: 30 minutes by default, up to 30 days for something like finishing onboarding within a week.
See where it fails
Read success rate, median time on task, how long the slowest 10% take, and the last page reached by attempts that never succeeded. Split any of it by platform, app version or accessibility setting.
Find out why
Compare friction in failed and successful attempts, follow journeys from the step people stall on, ask a one-question survey there, and watch consented replays.
Separate the friction that stops people from the noise
The script tag captures rage clicks, dead clicks, slow controls, error clicks, u-turns, form errors and form abandons on its own. The friction report ranks them by element and page, with the number of sessions affected.
Inside a task, friction is split by outcome. A dead click that shows up everywhere is a nuisance. One that shows up mostly in abandoned attempts is a finding.
- Nothing to tag by hand: friction arrives from the first page view.
- Field values are never read. A form error is recorded, not what was typed. Add
data-mk-maskto keep an element’s text out. - Friction rate per page, so you can find the screens where the most sessions struggle.
| Signal | Attempts |
|---|---|
| U-turn left the template picker | 388 |
Dead click div.template-card | 241 |
Form error input#template-name | 133 |
Rage click button "Create" | 64 |
Abandoned attempts hit friction four times as often. The top signal is a u-turn out of the template picker, which rarely shows up in attempts that succeed.
Follow what people actually do, and whether they come back
Journeys draw a Sankey flow of how people move through your site or app, anchored at any page or screen. Start from the step where a task stalls and see where people go instead.
Cohort retention shows whether each day’s or week’s new visitors come back. When a number looks odd, open a single visitor’s session timeline and read their events in order, alongside a live event stream.
- Journeys re-anchor on any page, with the top paths at each step.
- Retention by day or week, keyed on when each visitor was first seen, in each app’s time zone.
- Session timelines per visitor: pages, events, friction and survey answers in one sequence.
| First seen | People | Week 1 | Week 2 | Week 3 |
|---|---|---|---|---|
| 25 Aug | 1,204 | 38% | 27% | 22% |
| 1 Sep | 1,318 | 36% | 26% | 21% |
| 8 Sep | 1,450 | 44% | 34% | 29% |
| 15 Sep | 1,391 | 45% | 35% | Not yet |
Cohorts since 8 September hold about 8 points better in week 1. Journeys show them sending their first invoice in their first session.
Ask at the moment it happens, then watch it
In-app surveys ask NPS, CSAT, effort, star rating, multiple choice or open text. Target them by page, event, app version, platform, sign-in or accessibility setting, with sampling, frequency caps and schedules.
Every answer is an event. “People who found it very hard” becomes a filter on journeys, funnels, friction and replays, so what people say and what they did sit side by side.
- “Report a problem” widget: always on, with categories and an optional screenshot people can highlight or partly hide before sending.
- Heatmaps: clicks and scroll depth per page and device width, with a text summary and data tables.
- Session replay: opt-in and consent-gated, every input masked and all text masked unless you mark a region safe with
data-mk-unmask, anddata-mk-blockleaves an element out entirely, with friction and answers marked on the timeline.
| Answer | Chart | Share |
|---|---|---|
| Very easy (6–7) | 48% | |
| Fine (3–5) | 38% | |
| Very hard (1–2) | 14% |
Answers are events, so “scored 1–2” is a filter. Applied to replays, most of those 86 sessions show the same u-turn out of the template picker.
Tasks, funnels, journeys, retention, session timelines and accessibility segments work for any app that sends events, iOS and Android included. Surveys, feedback, rage taps and u-turns run on web and in the iOS, Android, Flutter and React Native SDKs. Heatmaps and replay are web-only for now.
Talk to the people behind the numbers
When a task’s numbers say where people struggle but not why, run a study. A moderated session is a video call people book from your availability, with a session guide for your notes. An unmoderated test shows people one task at a time on your live website, detects when they succeed and asks how easy it was.
Recruit from the people who just answered a survey, such as everyone who scored a task as very hard, or share a recruiting link. Results sit next to the live task or funnel the study explains, broken down by accessibility setting, and you record whether the hypothesis held up.
- Access needs asked at booking, such as captions, extra time or a sign language interpreter, so the session works for everyone.
- Participants’ details stay private: names, emails and access needs need their own permission, which the Viewer role doesn’t have.
- Recordings are optional and masked like your replay settings, and people can say no and still take part.
Studies are on Growth and above. Unmoderated tests run on websites with the Metrickle script. Metrickle doesn’t supply testers or handle incentive payments.
Quantitative and qualitative, on the same events
Friction report
Rage, dead and error clicks, slow controls, u-turns, form errors and form abandons, ranked by element and page.
In-app surveys
Six question types, precise targeting and answers you can filter by.
Heatmaps and replay
Consent-gated replay with masking by default, and heatmaps per device width, from Growth.
Funnels
When the question is about a sequence of steps rather than one task, with the biggest drop-off called out.
Accessibility segments
Compare any task against people using no accessibility settings.
Findings into the backlog
Turn a hotspot, task or report into an issue in Jira, Linear, GitHub, Asana or another tracker with the evidence attached.
Studies
Moderated sessions and unmoderated tests, recruited from your own surveys, from Growth.
AI assistants (MCP)
Connect an AI assistant over MCP to run tasks, read friction and draft a survey or study for you to review.
See which release is costing you conversions, and who it broke for
Free to start. One script tag on the web, one SDK on mobile, and no cookie banner in cookieless mode.
Start free