Experiments without a stats degree
Test the live version of a quiz against one variant, pick the metric that matters, and read a plain summary of who is winning. Phoenix handles the math and the guardrails.
Start with a question, not a spreadsheet
Ask the assistant what to improve. It finds the page losing the most shoppers, explains why, and builds the variant for you. You approve the test.
The live version against one variant
Split traffic 50/50 or set your own split. Each visitor sticks to the version they first saw. Pick the metric that decides the test and the metrics a variant must not harm.
Experiments are on Plus and Enterprise.
A variant cannot win if it harms these, even when it leads on the decision metric.
Email sign-upsGuardrails, so a lucky week cannot fool you
The summary tells you who is winning, or that it is too early to say, in plain words.
Every test runs a full week minimum, so weekday and weekend traffic both count.
A single unusually large order does not decide a revenue test on its own.
Who is winning, by how much, or that it is too early. No p-values to decode.
End the test your way, publish in one step
When the test ends, the winner publishes and the test closes in a single step. No copying changes across by hand.
Start with one quiz
You don't need a perfect plan to begin. Start small, and let Phoenix help you improve from there.