Get started
Web Experiments

A/B testing that shows which version wins, and why

Web Experiments runs A/B, split URL and multivariate tests from the same snippet as your heatmaps. Build variants in a visual editor, read results you can trust, and watch recordings of each variant to see why the winner won.

Get startedSee Journey Insights
No extra codeTests run from the snippet that already powers your heatmaps.
Flicker-safe by defaultOnly the elements a variant changes are hidden, for at most 500 ms.
Plan first, then launchThe main metric and the smallest lift worth detecting are set before a test starts.
Three ways to read resultsBayesian, sequential or fixed-horizon statistics.

What you get

Everything you need to run a test properly

From the first idea to the rolled-out winner, without a second tool or a second snippet.

A/B, multi-page, split URL and multivariate tests

Pick the kind of test that fits the change. Every type runs from the same snippet and uses the same metrics and reports.

  • A/B tests on one page or a URL pattern
  • Multi-page tests that change every step of a funnel together
  • Split URL tests that send a share of traffic to another page
  • Multivariate tests that try every combination of sections and options
  • Multi-armed bandit traffic that sends more visitors to the variant that’s ahead

A visual editor for most changes, and code when you need it

Click any element to change its text, style, link or image. Switch to code for custom CSS and JavaScript in each variant.

  • Click to select, drag to move, double-click to edit text
  • Desktop, tablet and mobile widths
  • Custom CSS and JavaScript for each variant
  • Changes save as you go, and drafts never go live on their own

Metrics that match your goals, and guardrails that catch harm

Choose what success means for this test: a click, a signup, revenue per visitor or average order value. Add guardrails so a variant that hurts another number gets flagged or switched off.

  • A metric library with clicks, form submits, goals, revenue, bounce, scroll depth and more
  • Revenue per visitor, average order value and profit per visitor
  • Guardrails that alert you, or turn off a variant, when a metric gets worse
  • Rage, dead and error clicks as metrics too

Show each test to the right visitors, at the right moment

Run a test for everyone, or only for the visitors it’s meant for. Decide when it appears and how often.

  • Target by device, channel, country, landing page, UTM, visitor type and more
  • Save rules as audiences and reuse them
  • Trigger on page load, scroll depth, time on page or exit intent
  • Show it always, once per visitor, once per session or once a day
  • Keep tests apart with mutually exclusive groups

Statistics you can explain to your team

Cuttl plans each test before it starts and warns you when a result isn’t as solid as it looks.

  • Bayesian, sequential or fixed-horizon statistics
  • A pre-test planner that estimates how many weeks a test needs at your traffic
  • The smallest lift worth detecting is set before launch, and later changes are flagged
  • False-positive risk next to every result, and a warning if you check too early
  • Sample ratio mismatch checks

Results with the recordings behind them

Every variant has its own heatmaps and recordings, so a result comes with the reason behind it.

  • A launch decision: expected lift, chance to beat control, guardrails and a clear recommendation
  • Recordings and heatmaps for each variant
  • Results broken down by traffic channel
  • A history of every change to the test that can’t be edited afterwards

Roll out the winner gradually, with a holdout

Ship a variant to most of your visitors and keep a small group on the original, so you can keep checking that it works.

  • Ship to 50, 80, 90, 95 or 100% of visitors
  • Compare the new version with the holdout after launch
  • Ramp up step by step, and visitors never flip back
  • Ship a change without a test when you just need to launch it

Not enough traffic to test? Measure before and after

Most changes never get an A/B test. Change Impact compares the pages you changed with the rest of your site, before and after, and tells you plainly when it can’t tell.

  • Site updates detected automatically, plus anything you log
  • Compared against the rest of the site, so seasonal swings cancel out
  • A clear verdict: improved, worse, inconclusive or too early
  • Tells you how many weeks an A/B test would need instead

Test ideas that come with evidence

Turn any issue from Journey Insights into a test idea. It arrives with the recordings, the heatmap and the number of visitors it affects, and one click drafts the experiment.

  • Recordings, page insights and feedback quotes attached automatically
  • Draft the test in one click; drafts never start on their own
  • Search past tests before you run a similar one

All Web Experiments features

Included in every plan, on every site.

A/B testsOne page or a URL pattern
Split URL testsSend traffic to another page
Multivariate testsEvery combination, broken down
Multi-armed banditMore traffic to the leader
Visual editorText, style, links and images
Targeting and triggersAudiences, exit intent, frequency
GuardrailsCatch harm early
Revenue metricsRevenue per visitor and AOV
Three stats enginesBayesian, sequential or fixed
Per-variant recordingsSee why a variant won
Rollouts and holdoutsRamp up step by step
Change ImpactBefore and after, without a test

Tests you can run without worrying

Fast pages, fair results and a person in charge of every launch.

Flicker-safe and fast

Only the elements a variant changes are hidden, for at most 500 ms, so the rest of the page renders normally and Largest Contentful Paint isn’t delayed.

Fair results

Exposures are counted only once the tracker is allowed to record, and bots and AI agents are left out of results.

AI that asks first

Claude or ChatGPT can draft a test through MCP, but nothing launches until a person on your team approves it.

Not sure what to test? Start with what visitors do.

Journey Insights shows where visitors get stuck with heatmaps, recordings and frustration signals, and turns each problem into a test idea.

Explore Journey Insights

Web Experiments questions

Web Experiments is Cuttl’s A/B testing product for websites. It runs A/B, multi-page, split URL and multivariate tests from the same snippet as Cuttl’s heatmaps and recordings, with a visual editor, guardrails, Bayesian, sequential or fixed-horizon statistics and gradual rollouts.

A/B testing shows two or more versions of a page to different visitors at random and measures which one gets more people to do what you want, such as sign up or buy. Because visitors are split at random, the difference in results comes from the change, not from who happened to visit.

An A/B test changes elements on the same page. A split URL test sends part of your traffic to a different URL, which suits full redesigns. A multivariate test changes several sections at once and tries every combination, so you can see which option in each section works best.

Not for most changes. The visual editor lets you change text, styles, links and images by clicking on the page. For larger changes you can add custom CSS or JavaScript to a variant.

Cuttl hides only the elements a variant changes, and only until the variant is applied, for at most 500 ms. The rest of the page renders normally, so your Largest Contentful Paint isn’t delayed.

It depends on your traffic, your current conversion rate and the smallest lift you care about. Before launch, Cuttl’s planner uses your last 28 days of traffic to estimate how many weeks the test needs, and tells you when an A/B test isn’t realistic.

Bayesian by default, with sequential and fixed-horizon tests available. Sequential tests are safe to check every day. Every result also shows its false-positive risk, the chance that a win isn’t real.

Use Change Impact. It compares the pages you changed with the rest of your site before and after the change, and says plainly when there isn’t enough data to tell.

Yes. Every variant gets its own session recordings and heatmaps through Journey Insights, so you can see how visitors behaved in each version.