A/B Testing for Upsell Offers
A/B Testing for Upsell Offers
A/B testing lets you show two or more versions of an upsell offer to your customers and see which one sells better. You don't have to guess. For example, if a customer just bought skis, you could show one group of customers ski goggles and another group ski gloves — then let the results tell you which offer works best.
You can run an A/B test on any offer block on these pages:
- Checkout
- Post-purchase
- Thank You
- Order Status
Each version of the test tracks its own Impressions, Clicks, Conversion rate, and Revenue, so you can see exactly how it performs.
Starting an A/B Test
- Open your funnel and go to the offer block you want to test (for example, Upsell Offer 1 on the Checkout step). The block needs to already have a product, collection, automatic, or add-on offer configured.
- Click the ⋯ (three-dot) menu on that offer block.
- Hover over Create A/B test, then pick how the second version should be created:
- Duplicate – copies the current offer as-is
- Product – pick a different product
- Collection – pick a collection
- Auto – let Shopify recommend a product automatically
- Add-on – add an add-on offer

This creates an A/B test panel with Version A and Version B cards side by side. Version A is your original offer and starts as Active; Version B starts as Inactive.

Adding More Versions
Need to test more than two options?
Click + Add Version above the version cards to add Version C, D, and so on, using the same offer-type choices (Duplicate, Product, Collection, Auto, Add-on).
Each version card has its own ⋯ (three-dot) menu where you can change what it offers (Change product, Collection offer, Automatic offer, Add-on offer), Copy offer, or Delete it — as long as the test isn't currently running.
Setting the Traffic Split
Click the ⚙️ gear icon next to Add Version to open the Impression percentage window. Drag the slider to decide what share of visitors sees each version (for example, 50% / 50%, or 33.3% each for three versions), then click Save.
If you pause or delete a version later, the remaining versions automatically split its share of traffic between them.
Running the Test
Click Start to launch the test.
- Your funnel must be published first. If it isn't, a Publish required window appears telling you the funnel needs to be saved and published before the test can start — click Save & Publish to continue.
- Once the test starts, the Start button becomes Stop, and every version's status changes to Testing.
- While a test is running, editing is locked: Add Version and the percentage settings are disabled, and each version's ⋯ (three-dot) menu is grayed out. The only action available is Stop.

All impressions, clicks, conversion rate, and revenue numbers update in real time and also feed into your overall funnel analytics.
Stopping the Test and Picking a Winner
Click Stop to end the test. A Select winner window opens showing every version that took part, each with its offer preview and live metrics (Impressions, Clicks, Conversion rate, Revenue).
The app automatically highlights one version with a Recommended winner badge, based on its performance (for example, highest revenue or conversion rate). It's pre-selected, but you can click any other version to pick it instead — only one version can be chosen.

Click Save to confirm. The version you picked becomes Active and stays live on the page; every other version switches to Inactive. The test panel resets and the Stop button turns back into Start, so editing (Add Version, percentage split, per-version offer changes) is unlocked again — meaning you can adjust versions and launch another round whenever you like.
✔️ Next step
Check our articles to learn more about the app's functionality:
Updated on: 31/08/2026
Thank you!
