Squoosh documentation

Squoosh runs an A/B test without live traffic. You give it two versions of a page — your current Control page (A) and a Variant page (B) you've built — and Squoosh sends a sample of AI synthetic shoppers to each one. Each shopper browses and decides whether to convert, the same way a real customer would. A Standard run returns a verdict in about 20 minutes on which version converts better.

The synthetic shoppers are modeled on your real traffic, so the result reflects your real customers.

How it works

  1. Connect a site and calibrate synthetic shoppers. Add the site you want to test and link Google Analytics or Shopify, then calibrate a pool of synthetic shoppers whose device, source, and behavior mix matches your real visitors. See Set up Squoosh and Calibrate synthetic shoppers.
  2. Run an experiment. Enter your Control page (A) and Variant page (B), pick a conversion goal, and launch. See Run an experiment.
  3. Read the verdict. When the run finishes, Squoosh tells you to ship the variant, keep the original, or re-test. See How Squoosh decides a winner.

Documentation

Section What's in it
Getting started What Squoosh is, signing in, adding a site, calibrating shoppers, and running your first experiment.
Core concepts Synthetic shoppers, how a winner is decided, reading lift and confidence, and credits and tests.
Experiments What you can test, shipping variants, watching shoppers, the run lifecycle, and reading the verdict.
Conversion Reports Diagnose a single page's path to conversion without a second version.
Integrations Connect Shopify or Google Analytics, and keep synthetic shoppers out of your analytics.
Account and billing Credits and billing, workspaces and properties, and settings.
Help FAQ, glossary, and troubleshooting.