Experimentation
A/B testing and experimentation built for revenue
Test headlines, CTAs, paywalls, and server-rendered features, from a simple A/B to a multivariate A/B/n with up to five named variants, then judge winners by net revenue per 1,000 visitors, not vanity clicks. Get confident calls from your first conversions.
7-day free trial. $0 today, cancel anytime.
Test anything, from copy to pricing
Point at a headline or button with a CSS selector and swap it client-side before paint (no flicker), or run a manual experiment your server or code controls. Use it for wording, layout, paywalls, onboarding, and prices.
- Element and CTA tests applied by the tracker, no code
- Manual experiments for server-rendered pages and pricing logic
- Deterministic, sticky assignment across browser and server
Two variants or five, named your way
Run a classic A/B or a multivariate A/B/n test with up to five variants in one experiment. Name each variant something meaningful (hero-blue, price-29, long-copy) instead of a lettered guess, and read the split, uplift, and winner per variant with every variant colour-coded.
- Up to five variants per experiment, control plus four
- Editable variant names that carry through to results
- Weighted traffic split with deterministic, sticky bucketing
Judge winners by money, not clicks
Tie results to revenue with a Stripe, Shopify, or WooCommerce connection and rank variants by net revenue per 1,000 visitors. A click that never pays is not a win, StatsKit makes sure you optimize the number that matters.
- Revenue-tied results across multiple sources
- Custom metrics as primary or guardrail
- Per-variant uplift, significance, and a clear winner
Confident calls, early
Bayesian stats give you a readable “chance to win” from your first conversions, so early-stage sites with modest traffic still get actionable answers instead of waiting months for significance.
- Bayesian or frequentist engines
- Chance-to-win read on your real conversion volume
- AI verdict that explains what the result means
Frequently asked questions
Will experiments slow down or flicker my site?
No. The tracker applies element and CTA changes before first paint, and the script is tiny (under 6KB gzipped).
How do I run an experiment on a server-rendered page?
Use a manual experiment: fetch the assigned variant from the API (or the SDK) on your server and render it yourself.
Can I test more than two variants?
Yes. Every experiment type supports up to five variants (control plus four), and you can name each one, so results read as hero-blue vs long-copy vs control instead of a, b, c.
See why your visitors don't pay
Install one script, and start finding your leaks today.
Start free trial →