Boost your upsell conversion by continuously testing your offers. With A/B testing, one funnel shows two versions of its offer to your customers and measures which one converts better.
💡 In short: variant A is your current offer, variant B is a second, fully independent offer. You choose the traffic split, the app assigns each customer once, and shows you when a winner is statistically clear.
How to run a test
Open the funnel and switch on the A/B test in the bar at the top of the offer editor.
Use the same bar to toggle between variant A and variant B while editing. Variant B is a complete offer of its own: product, discount, texts and layout can all differ.
Set the traffic share for variant B, anywhere from 1 % to 99 % (default 50 %).
Publish. From now on each customer is assigned to one variant.
How customers are assigned
The variant is chosen once per checkout, on our servers, the moment the offer qualifies. Everything that follows for that customer, the impression, the price they pay, and the conversion, is attributed to the variant they actually saw. The analytics can never mix the two up.
To find the orders of each variant in Shopify, switch on Tag orders in the Optimize tab and enter a tag for variant A and one for variant B. The order gets the tag of its variant when it is updated with the upsell. You can change the tags while the test is running.
Reading the results
The funnel list shows an A/B badge and a results panel with the raw numbers per variant (impressions, conversions, revenue) plus a verdict. The verdict is based on a proper statistical test and suggests a winner only at 95 % confidence or higher.
⚠️ Important: give the test time. With only a handful of impressions per variant, differences are noise. Wait for the verdict rather than judging after a day.
Picking a winner
Click Pick winner once the verdict is in. The winning offer becomes the funnel's main offer, the test is marked completed, and every customer sees the winner from now on. Variant B and the per-variant history stay visible in the panel, so you can always look back at the outcome.
Tips for meaningful tests
Test one change at a time. If variant B has a different product AND a different discount, you will not know what made the difference.
Discounts are the strongest lever, so they are usually the best first test. See Offer discounts.
Edits apply per variant. Changing the product in variant A leaves variant B untouched.
Related articles