Protect the tasks that make you money
Pick the tasks that make you money, like sign-up, checkout or inviting a teammate. After every release, Metrickle checks that everyone still gets through: on every browser and device, and people using keyboard navigation, zoom, large text or a screen reader (in your iOS and Android apps), who your analytics usually lumps in with everyone else. When a release breaks a task, you get one case: the step, the button, the revenue lost and, from Growth, the recordings that prove it.
| Setting | Success | Slowest 10% | Verdict |
|---|---|---|---|
| No accessibility settings | 71% | 3m 10s | Compared with |
| Reduced motion | 70% | 3m 02s | No gap |
| Zoomed in | 68% | 3m 25s | No gap |
| Large text | 37% | 4m 05s | Early signal |
| High contrast | — | — | Too few attempts (9 of 15) |
| Keyboard navigation | 38% | 7m 40s | Gap found |
About £9.1k lost so far: since v4.12, 38% of people using keyboard navigation finish checkout, against 71% of people using no accessibility settings. Most get stuck at Payment.
From a broken release to a verified fix
Protect a task
Say where it starts and where it succeeds, like checkout started and plan paid for, then protect it. An AI assistant can set one up for you, but only a person can protect it.
Get a verdict after every release
Each browser, operating system and device is compared with everyone else, and people using each accessibility setting with people using none: do they finish as often?
Fix it, and see that it worked
One case for each group a task fails, with links to the recordings on paid plans. Ship the fix and the case only closes once the same people measurably get through more often again.
Red only when it’s real
An alert your team learns to ignore is worse than none, so red has to mean something.
- No gap, Early signal, Gap found or Too early to tell. An Early signal is never green: people are falling behind, but there isn’t enough usage yet to be sure. A single group with too little usage shows Too few attempts.
- No alarm from a handful of visits: by default each group needs attempts from 15 different people since the release, so one bad session can’t fail a release.
- Not a fluke: a gap only becomes Gap found once it’s statistically clear it isn’t chance.
- Sensible defaults you can change: Gap found when a group finishes 10 points less often than everyone else, or its slowest 10% take 1.5× as long.
- Checked every hour, shown for every app on your overview, and sent by email and to Slack, Teams or a webhook when a task’s verdict becomes Gap found.
- Part of every release’s verdict: each protected task’s success before and after the release, alongside your conversion goals. See release verdicts.
| Release | Shipped | Verdict |
|---|---|---|
| v4.13 | 2 hours ago | Too early to tell |
| v4.12.1 | 6 weeks ago | Gap found: keyboard navigation |
| v4.12 | 6 weeks ago | Gap found: keyboard navigation |
| v4.11 | 7 weeks ago | No gap |
| v4.10 | 9 weeks ago | No gap |
Large text has been an Early signal since v4.12. People using large text finish 37% of the time, down from 50%, but there aren’t enough of them yet to be sure. It stays amber, never green, until it’s settled.
One case, not a replay library
When a release breaks a task, Metrickle opens one case for it, about the people it broke for, with what it has cost so far. Every failed attempt is sorted by what most likely went wrong.
- Never started: they couldn’t find the way in, or clicked it and nothing happened.
- Wrong path: extra screens, search used as a back button, u-turns.
- Stuck: a form error, a page error, a button that did nothing or took too long, then repeated angry clicks.
- Done but expensive: got there, but far slower than most, after an error, or later taken back.
- No evidence, no blame: attempts with nothing to show for them are counted, never guessed at.
| What went wrong | Chart | Attempts |
|---|---|---|
| Stuck | 151 | |
| Never started | 35 | |
| Wrong path | 26 | |
| Done but expensive | 18 |
About £9.1k lost so far, and 212 people didn’t finish. Stuck at Payment: since v4.12, focus skips button "Pay now", so it can’t be reached with Tab. Fix shipped in v4.13, measuring. 5 recordings, each linked to the moment.
Recordings in a case come with Growth and up. They’re consented and masked, like all replays, and need replay switched on for the app. Native apps don’t detect taps that do nothing yet.
Success that matches the money you keep
A sale or sign-up counts for real after seven days. If the same person refunds, cancels or reports a problem before then, it’s taken back, so the success rate shows the money you kept, not just the clicks.
- The success rate counts what stuck, with the rate before take-backs alongside, plus how many were taken back, how many are still settling, and the revenue for each.
- Accessibility complaints count against the right group: whichever setting was on when the report was sent.
- A forecast: recent checkouts are ranked by how likely they are to be refunded, from signs like trying many coupons, going back to pricing, or fighting form errors.
| Reason | Completions | Revenue |
|---|---|---|
| Refunded | 96 | −£4.3k |
| Cancelled | 38 | −£1.7k |
Likely to settle at 67% once the 380 recent checkouts are past their seven days. The riskiest are the people who went back to the pricing page again and again before paying.
Connect Stripe or RevenueCat and refunds and cancels arrive on their own (they reach a checkout when the customer was signed in). Or send refunds as events with negative revenue, or give a task an undo step, like subscription_canceled.
Common questions
What is a protected task?
Something people come to your site or app to do that makes you money, from where it starts to where it succeeds, like starting checkout to paying for a plan. After every release Metrickle checks that people using an accessibility setting still finish it about as often, and about as fast, as people using no accessibility settings, and that each browser, operating system and device keeps up with everyone else. When a group falls behind, you get a case that says where it broke and what it has cost so far.
Does it catch a release that breaks one browser for everyone?
Yes. Each browser, operating system and device type is compared with everyone else, so if Safari users suddenly finish sign-up far less often than people on other browsers, the verdict becomes Gap found and a case opens about Safari users, with the same minimum of 15 different people and the same test for chance. This is separate from the accessibility comparison, which can’t see a break that hits everyone on one browser equally.
Which settings are checked?
All 11 the SDKs detect: keyboard navigation, zoomed in, large text, high contrast, forced colours, reduced motion, reduced transparency, inverted colours, greyscale, bold text and screen reader (in your iOS and Android apps). Each platform reports the ones it can see: browsers hide whether a screen reader is running, so on the web keyboard navigation stands in for it. Nobody is asked: the setting comes from the device.
Can one bad session fail a release?
No. By default each group needs attempts from 15 different people since the release before it counts. A gap that could still be chance shows as Early signal, and only becomes Gap found once it’s statistically clear it isn’t, with every group tested at once taken into account.
Can an AI assistant protect a task or close a case?
No. An assistant can set up a task for you to review, check verdicts, open a case, draft a fix, mark it shipped and ask for it to be checked. Protecting a task and dismissing a case take a person in the dashboard. A case only closes when the same people are getting through again, whoever asks.
Which plans include protected tasks?
Free protects one task, with its case: the step, the button and the revenue lost so far. Growth and above protect every task, and add the recordings that prove each case.
Does this replace an accessibility audit?
No, it works alongside one. Audits and tools like axe find problems in your code; send axe results from your build and each one shows up next to the trouble it causes real people. Protected tasks show whether people using those settings actually get through, after every release, and which problem costs the most.
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