Check your test every day. Ship only real winners.

A/B, multivariate and split-URL tests on your store, with a report built to be read while the test runs. Every variant shows its chance of being the best and the risk of picking it — and nothing is called a winner until the data has earned it.

Start freeBook a demo

The Mtrix experiment report for lumière.co's test "Lifestyle photo on the PDP", day 16. The All variants table lists, for Control, Lifestyle photo and Scent notes first: allocation, confidence, visitors, conversion rate, next page visits, add to cart, checkout start and complete, and revenue. Lifestyle photo has a 98% chance of being the best and converts 24.5% better than control. Switch to Statistics for the engine's verdict: Variant 2 wins.

Experiments / lumière.co M

From idea to live test
in about five minutes.

Pick the page and the part of it to change, add the variants, write down why the change should win, and start it. Once Mtrix is on your store, that’s the whole job.

  • Four kinds of test A/B, multivariate, split URL — traffic sent to separate pages — and mask URL.
  • Variants built in the editor Make what each variant shows in the Mtrix editor. No hand-written code.
  • Assigned on the server Mtrix picks each shopper’s variant on its servers and keeps returning shoppers on the one they saw.
  • No cap on tests Run as many experiments as your traffic can carry, on any paid plan.

Chart any event,
already split by variant.

Right above the table sits a full chart builder scoped to the test. Pick any event your store tracks and it’s broken down by variant from the start — then segment it further, filter it, or turn it into a funnel.

A verdict you can check
every single day.

One click from the table, the statistics view says in plain words whether to keep waiting or roll out. It’s designed to be watched while the test runs — so looking early doesn’t fool you.

  • What a winner has to clear A 95% chance of being best, little risk of being wrong, 100 visitors and 25 conversions per variant, seven days, and a verdict that holds up to peeking.
  • Change the split mid-test Move traffic between variants while the test runs. The report notes the change and accounts for it.
  • Filter by another test Running two tests on the same pages? Filter one report by the variant shoppers saw in the other.
  • Watch a variant’s sessions Open the session recordings of shoppers who saw any variant — the fastest way to see why it won.

Ship the winner.
Take it back exactly.

When the call is made, send the winning variant to every shopper in one step. It becomes the page’s new default and goes live with your next publish — and if you change your mind, rolling back restores precisely what was there before.

Run a testing program,
not one test at a time.

Every experiment sits on one list with its leader and confidence, sorted by what’s ready to call. The program board lines up the uplift of every running test on one scale, so you can see at a glance which results have settled.

Test it. Read it honestly. Ship the winner.

Run A/B, multivariate, split and mask-URL tests on your store, check the report every day without fooling yourself, and ship the winner as your page's new default — then start the next one.

The rigor of an experimentation team. Without the team.

Start freeBook a demo
Switching from another stack? Talk to us. →
Build with Mtrix