Metrickle

Find the release that’s costing you conversions.

Every release gets a verdict: did it cost you sign-ups, sales or onboarding, and for whom? When a step breaks, you see where, who it hit, what it costs and the recordings that prove it. No waiting for a customer to tell you.

Release v4.12 · Sign-up frictionquillmere.example · Safari, against v4.11(sample data, fictional company)

Drop found Costing about £4.8k a month. Since v4.12, “Create account” does nothing on Safari, so people rage-click it. “Sign up” fell from 52% to 38%, and on Safari from 48% to 7%.

Sign-up page elements by rage and dead clicks on Safari, since release v4.12
ElementRage clicksDead clicksAgainst v4.11
button "Create account"4123889× more
text "14 days free"18143About the same
link "Sign in instead"94About the same
input "Email"611About the same

Case: Stuck at Sign up, Safari users. 38 sessions clicked “Create account” 3 or more times in a second. The 5 clearest recordings are picked out, masked by default. Fixed in v4.12.1: “Sign up” is back to 51%.

Free for three apps, no card · Recordings from Growth · No cookie banner in cookieless mode · Web, iOS, Android, Flutter and React Native

A verdict on every release

Analytics tells you conversion moved. Metrickle tells you which release broke which step, for whom, and what it costs.

  1. Send each deploy

    Call the deploy hook from CI, GitHub, Vercel or Netlify. In your apps, each new app version is picked up on its own.

  2. Every release gets a verdict

    Before against after, for getting in and for getting things done, and for every accessibility setting, browser and device. A drop only counts once it’s real, never on one bad session.

  3. Fix it, and see it recover

    You get the step, what it costs and the recordings. Mark the fix shipped and the case closes only once conversion measurably recovers.

How release verdicts work →

Checked past the sign-up, not just up to it

A release can let people in, then stop them getting anything done. Each release is judged across the whole journey: sign-up, then onboarding, then the things that make people stay.

  • Getting in: each conversion goal, before and after the release.
  • Getting things done: each protected task, like finishing onboarding or inviting a teammate.
  • Who it hit: the browser, device or accessibility setting that fell behind since the release, not gaps that were already there.

How release verdicts work →

Release v4.11quillmere.example · compared with v4.10(sample data, fictional company)

Early signal Could cost about £2.1k a month. “Finish onboarding” fell from 64% to 52%: 12 fewer in every 100 new sign-ups get going.

Each step of the journey before and after release v4.11
StepBeforeAfterVerdict
Bought a plan3.1%3.2% No drop
Sign up51%52% No drop
Finish onboarding64%52% Early signal
Invite a teammate33%32% No drop

Where it leaks: “Add your first client”, the onboarding step v4.11 changed. Most people who leave there u-turn to the dashboard.

Know what to fix first, with the proof attached

Every failed attempt is sorted by what went wrong. The biggest group tells you what to fix first, and the recordings that show it are already picked out.

  • Five recordings, not five hundred: masked, and each linked to the moment it went wrong (recordings from Growth).
  • Counts the sales you keep: refunds and cancels within seven days come off the success rate.
  • Straight to your tracker: the fix drafted in Linear, GitHub or Jira with links to the recordings, and alerts in Slack or Teams.

How protected tasks work →

Case: Checkout, people using keyboard navigationquillmere.example · opened 6 weeks ago(sample data, fictional company)

Gap found About £9.1k lost so far. 212 people didn’t finish.

  1. Open
  2. Fix drafted
  3. Fix shipped, measuring
  4. Fixed and verified
Attempts by what went wrong
What went wrongChartAttempts
Stuck151
Never started35
Wrong path26
Done but expensive18

Stuck at Payment: since v4.12, focus skips button "Pay now", so Tab jumps from the card details straight past it. The fix shipped in v4.13.

  • Recording 1: Tab skips “Pay now”0:48
  • Recording 2: stuck at Payment1:12
  • Recording 3: stuck at Payment0:31

Live in five minutes, cookieless if you like

Drop in the script tag and friction and accessibility settings arrive with the first pageview. Revenue takes one more line, and your deploys one more step.

  • Nothing stored on the device in cookieless mode, and no raw IP addresses kept.
  • Replay waits for consent, with every input and all text masked by default.
  • SDKs for iOS, Android, Flutter and React Native, and a deploy hook so every release gets a verdict.
  • Ask your AI assistant what broke, from any MCP client. AI assistants (MCP) →
<!-- Pageviews, friction, accessibility: automatic -->
<script defer
  src="https://app.metrickle.com/m.js"
  data-key="mk_pub_…"
  data-cookieless></script>

<!-- What a sale is worth (queued until m.js loads) -->
<script>
  window.metrickle ||= { q: [], track() {
    this.q.push(["track", ...arguments]) } };
  metrickle.track("purchase", { revenue: 49 });
</script>

# After each deploy, so it gets a verdict (or point GitHub, Vercel or Netlify at the hook)
curl -X POST "$METRICKLE_DEPLOY_HOOK" -d '{"version": "4.12"}'

Know who your next release breaks for

Free for three apps with one protected task, and recordings from Growth. Add the script tag today and it’s checked from your next release.

Start free

The one-minute tour 63 seconds, with music. Sample data from a fictional company.

Read the video as text
  1. 0:00 The Metrickle logo. “Every release gets a verdict.”
  2. 0:08 When a release breaks a step. “See where, who it hit, and what it cost.” A case for the Quillmere web app scrolls by: keyboard navigation can’t reach “Pay now” at Payment. 212 people didn’t finish, about £9.1k lost so far, 5 recordings.
  3. 0:15 Conversion optimisation. “Find the release that cost you sign-ups.” Each release is compared with the one before it. Conversion rate falls from 6.0% to 5.0%, right after release v4.12.
  4. 0:23 Accessibility. “See who a release broke it for.” Release v4.12 against v4.11, checkout by accessibility setting, before and since: no settings 72% then 71%, reduced motion 70% then 70% (no drop), zoomed in 69% then 68% (no drop), large text 50% then 37% (early signal), forced colours 31% then 30% (no drop), keyboard navigation 69% then 38% (drop found). 212 people didn’t finish.
  5. 0:31 UX research. “Watch where people get stuck.” A replay of someone using keyboard navigation at checkout, with card details masked. Tab jumps from the postcode field straight past “Pay now” to “Change plan”.
  6. 0:38 AI assistants. “Ask Claude where it leaks.” Asked where keyboard navigation fails since v4.12, it answers Payment, explains why, and drafts the fix on the case: put “Pay now” back in the Tab order.
  7. 0:46 Web and mobile. “Every app, one view.” The Today page shows the Quillmere web app and its iOS and Android apps, with what each is losing and what was fixed this week.
  8. 0:54 Verified fixes. “Ship the fix. See that it worked.” Release v4.13 against v4.12: people using keyboard navigation finish checkout 68% of the time, back from 38%. No drop.
  9. 0:57 “Find the release that’s costing you conversions.” Web, iOS, Android, Flutter and React Native. Free for three apps, no card, and no cookie banner in cookieless mode.