Datsnip

Product · Experiments

Know the fix worked. Don't hope.

Build variants in a visual editor, split traffic, track a goal, and let the numbers pick the winner. Then roll the winner out as a live patch without a ticket.

Updated November 11, 2026

From idea to winner in four steps

  1. SetupName it, write the hypothesis, choose which URLs it runs on and how much traffic to include.
  2. VariantsCreate A/B/n variants. Each variant is a set of changes (text, style, link, hide, redirect) built in the visual editor, with weights that rebalance automatically.
  3. GoalPick what counts as a win: a click, a page view or a custom event.
  4. LaunchReview and start. Visitors are assigned a stable variant on their first visit.

What you get in the results

  • Exposures, conversions, conversion rate and uplift per variant.
  • Confidence from a Bayesian model, and a clear "winner found" status.
  • Sample-ratio mismatch (SRM) warnings, so a broken split does not fool you.
  • A sample-size estimate before launch, so you know roughly how long a test needs.
  • Funnels split by variant, to see where the winner wins and not just that it does.
  • Revenue tracking with one line: window.datsnip.track('purchase', { revenue: 49.99 }).
  • Roll out the winner as a live patch in one click.

Honest expectations

A test needs traffic. Small effects on low-traffic pages take a long time to prove, so start with changes you expect to matter, and use the sample-size estimate to decide whether a test is worth running.

Go deeper in the docs

Frequently asked questions

Can I run A/B tests without developers?

Yes. Variants are built in a visual editor over your live page. You only need a developer once, to paste the snippet.

How long should a test run?

Until it reaches a trustworthy result. Use the sample-size estimate before you launch; low-traffic pages need larger effects to reach confidence in a reasonable time.

Will variants flicker?

Datsnip applies changes as the page loads, and offers an optional anti-flicker snippet that briefly holds the page while variants apply. See the anti-flicker guide for the trade-offs.

Can I test more than two versions?

Yes. Experiments support A/B/n with any number of variants and adjustable traffic weights.

Run your first test this afternoon.

The visual editor and experiments are ready on the free plan to try.

Keep exploring