How it works

Paste a link.Get a fix list.

Six steps between your link and a list of what to fix, all done by us. Your part is the first step and the fixing. Here's what happens in between, with nothing left out.

Link inEvery tester on their own
STEP 01 · YOU

Paste a link and say what to try.

Give us the address and, if you like, the job: "buy any tea as a guest and pay with the test card." Leave it out and they'll wander like a first-time visitor.

  • A test login or a test card if they need one. It's used, never shown in a report.
  • Which countries your customers are in, so the testers live there too.
  • Before the first run, you prove the site is yours.

Takes about a minute.

kettle.teaThe link, the job, the test card
STEP 02 · US

We build testers for your product.

Each tester is dealt a way of using websites (looks everywhere, tries to break it, does as told, notices every pixel, has no time), plus a patience level and a phone or a laptop. Then they're given a name, a job, a home and a reason to be on your site.

  • Some get the awkward details forms forget: apostrophes, plus-sign emails, long names.
  • Each gets a working email inbox of their own.
  • Their limits are numbers we enforce: patience, time, how much they read.

Meet some of them →

Five testers for a UK tea shopDealt, then written
STEP 03 · THE TESTERS

Each one uses it alone, in a real browser.

Every tester gets their own Chrome, at phone or laptop size, and nobody else in it. They see the page as you would: the screen and the words on it. They click, type, fill in forms, scroll, go back, wait and check their email. One action at a time, and every step is recorded with a screenshot and what they were thinking.

  • When something stops them, slows them down or just looks wrong, they report it on the spot.
  • Every slow page and every dead end costs them patience. When it runs out, they leave, like a real customer.
5 browsers, 5 people, nobody sharingPatience bars under each
STEP 04 · A SECOND CHECK

Every report is checked before you see it.

Testers can be wrong. A second model reads each report against the recording and the screenshot. It keeps what's real and throws out what isn't.

  • Real problems go on to the fix list.
  • False alarms go in the bin, and we count them.
  • "Not for me" (taste, not a fault) goes in its own pile, off your list.
Real · false alarm · not for meRead against the evidence
STEP 05 · YOU GET

One fix list, worst first.

Ten testers hitting the same bug is one item, not ten. We merge them and rank them: first what stopped someone, then how many people hit it, then how bad it was.

  • Steps to repeat, rebuilt from the recording, never made up, with no password in them.
  • The screenshot, the page, and what the testers said.
  • Where to look and how to fix it, written so a person or an AI agent can work through it item by item.
An example built from Kettle's bugsRanked: stopped someone, then how many
STEP 06 · TOGETHER

Fix it. They come back to check.

Send the same testers back in. Each one remembers what went wrong for them last time, goes back to the same place and tries again. You get three lists: fixed, still there, and new.

  • An item only counts as fixed if a tester went back to that page and it worked.
  • If nobody reached it this time, it says so instead of guessing.
Example retestFixed · still there · new

House rules

Rules the testerscan't talk their way out of.

These are written into the code that runs the testers. The model plays the person, but it doesn't make these calls. Our self-checks test every one of them after each change to the engine.

Today and next

What works today,and what's still being built.

We'd rather tell you than have a tester find out.

Your part

You paste the link.We'll do the rest.

Ten testers on one task is ten credits, and a retest is ten more. See pricing.