A test without a goal is just a traffic split. When you create an A/B test, you pick one primary goal so Instant knows what "winning" means, plus a market, language, and traffic split so the right shoppers enter the experiment.
Configure these in the A/B testing form when you create a test. You can also change the goal later from the test's results page.
Goals
Pick one goal per test. Instant uses it to rank variants and declare a winner once confidence is high enough — above 90% for your selected goal.
Conversion rate
Track which version drives more completed purchases.
This is the default goal, and the most direct link to purchase outcomes. Use it when your hypothesis is that the change increases the share of visitors who place an order.
Revenue per visitor (RPV)
See which version brings in more revenue per unique visitor.
Use RPV when both spend and conversion might move — for example upsell modules, bundles, or merchandising that can raise revenue even if conversion rate stays flat.
Average order value (AOV)
Find out which version increases the average spend per order.
Choose AOV when you expect bigger baskets rather than more orders — gift wrapping, recommended products, or bundle layouts. AOV is calculated from orders that convert, so low conversion volume slows learning.
Click-through rate (CTR)
Measure which version gets more engagement.
Use CTR for upper-funnel changes where the main outcome is a click (hero CTAs, navigation, promotional blocks) rather than an immediate purchase. CTR can reach a result faster on high-traffic pages, but a CTR win does not guarantee more revenue.
Goal selection guide
Hypothesis | Suggested goal |
Clearer product page increases purchases | Conversion rate |
Upsell or bundle grows revenue without necessarily more orders | Revenue per visitor (RPV) |
Merchandising increases spend per order | Average order value (AOV) |
Hero or CTA drives more clicks | Click-through rate (CTR) |
Market and language
Under Localization, choose:
Market — which Shopify market the test runs in
Language — which locale within that market, shown when the market has more than one
Instant resolves each variant's URL against the selected market and language and displays the resulting Market URL. Use these so localized copy or currency context doesn't mix traffic from other storefronts into the same experiment.
Loading markets needs the Shopify read_markets permission; if it's missing, Instant prompts you to re-authorize Shopify. Full Shopify Markets workflows across all layout types require the Business plan. See Shopify Markets and localization.
Traffic split
Under Traffic split, set the percentage of enrolled traffic each variant receives. Each value must be between 1 and 99.
50/50 for two variants is the usual default. With two variants, the split auto-balances so the two percentages total 100%.
Uneven splits (for example 80/20) keep most shoppers on the control while you still learn from a challenger.
With more than two variants, set each percentage yourself. The first variant is treated as the Control when you have three or more.
How shoppers enter the test
Tests use an A/B test link (shown in the form and on the test's page). Direct traffic to that link so Instant can assign a variant and count sessions toward your goal. Visitors who never hit the test link are not enrolled.
Each enrolled unique visitor consumes one A/B test session from your monthly plan pool. High-traffic tests use the pool quickly — watch your remaining sessions in the A/B testing panel. The pool resets monthly; upgrade if you hit the cap (Upgrade your plan).
Common mistakes
Wrong goal for the hypothesis — a CTR-optimized hero may win clicks but lose on conversion or RPV.
Mixing markets or languages — untranslated or wrong-currency variants skew results.
Multiple big changes in one variant — you won't know what moved the needle.
Changing a variant mid-test — this mixes data; duplicate or start a new test instead.
Next steps
Set up variants and launch in Create an A/B test.
After the test runs, interpret the outcome in Read A/B test results.