Glossary¶

Definitions of the terms you'll see across Squoosh. Each entry links to the page that covers it in full.

Audience The synthetic shoppers that run a test. For an A/B experiment you choose a Rigor level and Squoosh sizes the panel; each shopper tries both versions. A Conversion Report uses a named panel of 10 shoppers spread across the archetype styles you pick. See Run an experiment and Conversion Reports.

Auto-stop Ending a test early when interim results cross a threshold. A/B experiments do not currently use confidence-triggered auto-stop; they finalize after their shopper sessions have ended. See Experiment lifecycle and stopping.

Calibration Shaping your synthetic shopper pool to match your real traffic — device mix, traffic source, and location — from your connected Shopify or Google Analytics data. You calibrate the pool once; every experiment reuses it. See Create synthetic shoppers.

Confidence How sure Squoosh is that the result is real and not random noise. The standard verdict uses an 80% confidence bar. A measured result (High rigor and above) uses sampling intervals instead of a confidence meter. See Reading lift and confidence.

Control page (A) and Variant page (B) The two versions you test: the current page (Control page (A), tagged "A · Control") and the new version (Variant page (B), tagged "B · Variant"). You supply both — Squoosh does not write the variant for you. The default way to provide version B is Two URLs: paste a second, separately hosted page. See Variant ingestion.

Conversion Report A single-page funnel-leak diagnostic. You pick a named panel of 10 synthetic shoppers across the archetype styles you choose and point them at one page; Squoosh crawls the page to discover the funnel, walks the panel through it, and shows you where and why shoppers drop off. There is no second version, and no winner, lift, or confidence. See Conversion Reports.

Conversion Report panel The 10 named synthetic shoppers who walk the funnel in a Conversion Report. You assemble the panel by choosing which archetype styles (Budget hunter, Deep researcher, Impulse buyer, Brand loyalist, Skeptic, Gift buyer) take part; Squoosh spreads the 10 across them. Each shopper has a name and can be watched individually. See Conversion Reports.

Credit The unit you spend on experiments, Conversion Reports and Site Crawler captures. Each workspace has one shared credit balance. The run wizard shows a test's exact credit cost before you launch. See Credits and tests.

Estimated lift The lift reported by an A/B test at Standard or Medium rigor. Squoosh's AI reviewer compares the two pages and grounds its read in what those shoppers did. It is a directional read, not a statistically conclusive result; the wizard labels it Estimated · AI review. See How Squoosh decides a winner.

Experiment Insights A plain-language summary of what drove the result, shown directly under the verdict after a run finishes. It explains how the two versions differed and why shoppers decided the way they did. See Reading the verdict.

Lift How much the variant changed your conversion rate versus the control, as a relative percent. A positive lift means the variant converted more; a negative lift means it converted less. It is the measured difference between the two shopper groups, shown on the same scale as the per-arm rates — not a discounted estimate. See Reading lift and confidence.

Measured lift The lift reported by an A/B test at High rigor or above in Two URLs mode. Squoosh counts verified conversions on each page across the paired synthetic shoppers and reports the difference with a 95% sampling interval, shown as Village conversion lift; there is no confidence meter or reviewer override. The wizard labels it Measured · conversions. See Conversion-evidence results.

Method How an A/B test arrives at its lift: Estimated · AI review at Standard or Medium rigor, or Measured · conversions at High and above with Two URLs. The wizard shows it under the Rigor slider and in the Review & run summary, and the result page restates it. See Run an experiment.

95% CI A range describing sampling uncertainty under the calculation's assumptions, for example +6% → +37%. A narrow range means a more precise estimate; a wide range means less certainty about the size of the effect. It is not a measure of agreement with human-customer results. See Reading lift and confidence.

Primary goal The conversion you optimize in an A/B test: Add to Cart or Checkout. See Run an experiment.

Shopper match A score showing how closely your synthetic shopper pool reproduces your recorded traffic, written as "N% across K calibrated attributes" — the denominator names how many attributes were actually matched to real data, so the number never overclaims. It appears once you connect Shopify or Google Analytics. See Synthetic shoppers.

Shopper task A shopping objective the synthetic shoppers attempt during a run — for example, browse a collection and decide whether to add an item to the cart. Each test runs several tasks, and many shoppers run each one. See Run an experiment.

Site Crawler A property-scoped capture of desktop and mobile pages, component descriptions and recorded interactions. Replay follows saved states; it does not use the live site. Captures do not yet replace synthetic-shopper experiment sessions. See Site Crawler.

Rigor How rigorous an A/B test is, chosen on a slider in the wizard: Standard, Medium, High, Extra High, Max or Ultra. Higher rigor runs a larger experiment, takes longer and uses more credits; from High up the lift is measured and runs on Site Crawler captures (Settings → Features → Site Crawler). See Run an experiment.

Synthetic shopper An AI shopper that browses your pages the way a real customer would — reading, clicking, and deciding whether to buy. Squoosh sends a panel of them to each version of your page and records what they do. See Synthetic shoppers.

Test One experiment run: an A/B test or a Conversion Report. A/B pricing depends on the selected paired sample. The sidebar estimate uses the default A/B price and remaining monthly allowance. A Site Crawler capture is not a test for the monthly limit. See Credits and tests.

Traffic profile How the synthetic shoppers arrive at your page: Organic / Direct (browsing) or Paid Ad (PDP) (landing on a product page from an ad). It sets the source and intent of the shoppers, not a split between the two versions. See Run an experiment.

Verdict The decision Squoosh files when an experiment finishes: Variant B wins, Keep the original, or Too close to call. "Keep the original" is a real outcome, not a failure, and there is no ship button — rolling out a variant is always your call. See Reading the verdict.