Your next A/B test
already has results

Put your campaign in front of 1,000 synthetic customers with real budgets, real skepticism, and real switching costs. They vote. You ship the winner.

Free to start, no card required. Verdict in about a minute.

Run it
LIVE PANEL · 1,000 VOTINGvariant leads
CONTROL 28%
14%
VARIANT 58%

One minute. Whole story.

Watch the panel turn a rough campaign into a verdict.

One thousand opinionated people.
Sixty seconds.

Every dot is a panelist with a job title, a tool budget, a skepticism score, and something better to do. They do not flatter your idea. They tell you whether they would click, buy, or scroll past.

Prefer your variant
Stick with control
Would scroll past both
Browse the panel

Campaign in. Verdict out.

One workflow from rough idea to a decision you can defend in Monday's growth meeting.

1

Brief it

Describe the campaign or paste your control: landing page, email, ad set, or social post.

2

Build variants

AI drafts experiments scored by ICE, then generates close-to-control variants you can download.

3

Face the panel

1,000 synthetic customers, weighted to your audience, react to control and variant side by side.

4

Get the verdict

Run it, iterate first, or kill it, with the frictions and fixes the panel actually named.

Run it
Iterate
Kill it
Three possible verdicts. Zero wasted traffic.

The panel learns your audience.

Upload past campaign results or describe your customers. The panel reweights itself: more skeptics if your audience is skeptical, more budget hawks if they are price-sensitive. Reactions calibrated by your data, not averages.

  • Upload CSVs from ads, email, or analytics
  • Derived weights you can inspect and edit
  • Every simulation reacts the way your audience does
AUDIENCE WEIGHTS · FROM YOUR DATA
Skeptical Evaluator×1.9
Budget Optimizer×1.6
Efficiency Seeker×1.1
Trend Follower×0.5

Derived from churn_export_q2.csv + your notes. Editable before every run.

Every idea, hypothesis, and reaction is grounded in frameworks growth teams already trust: ICE scoring, Cialdini's persuasion principles, the Fogg Behavior Model, Jobs-to-be-Done, and Reforge growth loops.

What is AB Test Plan?

AB Test Plan predicts A/B test outcomes before you spend real traffic. Upload your page, describe your experiment, and let synthetic personas, each with real economic constraints and behavioral patterns, evaluate your control and variant.

Most A/B tests fail. Teams spend 2-4 weeks of traffic only to get inconclusive results. AB Test Plan simulates the test first, giving you a clear verdict (Run, Iterate, or Kill) with specific persona reasoning, so you only invest traffic in tests that are likely to win.

The tool also generates experiment ideas scored with the ICE framework, builds structured hypotheses, calculates statistically valid sample sizes, and generates ready-to-ship variants for landing pages, emails, ad copy, and social posts.

How to Predict Your A/B Test

1

Describe your product and goals

Your landing page, checkout flow, onboarding, pricing page — any conversion point. Upload HTML and supporting data.

2

Get AI-generated experiment ideas

5-8 ideas scored with ICE (Impact, Confidence, Ease) and grounded in behavioral psychology frameworks.

3

Build a structured hypothesis

"If we [change], then [metric] will [improve] by [amount] because [behavioral reason]."

4

Calculate sample size and impact

How many visitors, how many days, projected revenue impact if the experiment wins.

5

Preview control vs variant

AI generates a modified variant to compare side-by-side against your original.

6

Predict with synthetic personas

Personas with fixed budgets and real skepticism evaluate both versions — 6 in the free demo, up to 1,000 on Pro.

Frequently Asked Questions

What is AB Test Plan?
A free AI-powered tool that predicts A/B test outcomes using synthetic persona simulation. Instead of spending weeks of real traffic, you get a prediction in 60 seconds — complete with a Run/Iterate/Kill verdict, persona-by-persona reasoning, and specific iteration suggestions.
How does the prediction work?
Synthetic personas with real economic constraints (fixed budgets, time pressure), specific behavioral patterns (skepticism levels, decision styles), and existing workflow investments (switching costs) each evaluate your control and variant independently. Their responses are synthesized into an actionable prediction.
Why should I trust synthetic persona predictions?
Our personas are not generic AI chatbots saying everything is "interesting." They have rigid constraints that force honest trade-offs, like a real person deciding whether to spend their limited budget on your tool versus keeping their current workflow.
How does ICE scoring work?
ICE rates each experiment on Impact (1-10), Confidence (1-10), and Ease (1-10). Higher total scores indicate better candidates for testing. It helps you prioritize which experiments to run first.
Is it free?
The demo is free with no account: experiment ideas, hypotheses, sample sizes, variants, and 6-persona predictions. A free account adds saved campaigns and 25-persona panels. Pro ($79/mo) unlocks the full 1,000-persona panel, segment analysis, and audience calibration from your own data.
How long should I run an A/B test?
Until it reaches statistical significance (95% confidence) and has run for at least 7-14 days. But first, run it through AB Test Plan to make sure it is worth testing at all.

The panel is waiting.

1,000 customers, zero traffic spent, one verdict.