When to use it
- Compare existing Recharge, Stay.AI, or SKIO plans for one or more products.
- Create a Recharge plan with a different recurring discount and measure whether it attracts more subscribers.
- Compare first subscription orders by the test group that acquired the customer, and later renewals with Recharge or SKIO.
- Compare selling coffee beans as a one-time purchase at USD 39 against a monthly subscription at USD 31.20, to learn whether the subscription earns more per visitor over 90 days.
When not to use it
- To change checkout content, delivery, or payment options, use a checkout test.
- To change a product’s one-time price, use a price test. A one-time purchase test group in a subscription test keeps the product’s current one-time price.
Before you start
Your subscription app owns the plans and the widget shoppers use to choose them. ABConvert selects plans in that widget, so set up both in the app first:How it works
Choose the products first, then choose one purchase option per test group. A subscription test group applies one shared offer across all selected products. Each product keeps its own price. A one-time test group sells the same products at their current one-time prices without a plan. Recharge, Stay.AI, or SKIO owns the subscription plans and bills their renewals. ABConvert applies the assigned offer to new purchases; it does not change existing subscriber contracts. Visitors outside your target audience keep the normal storefront purchase choices.Set up the test
Choose Subscription in Create new test. The setup has four steps: Base info, Targeting, Subscription offer, and Review.1. Define the comparison in Base info
Name the test after the question you want to answer, such as “Care kit: one-time vs monthly.” Write a hypothesis that says what you expect to improve and why. Choose your primary metric before starting. Revenue per visitor helps compare the value of the first purchases. Conversion rate helps compare how often visitors place an order. Use later subscription results to assess renewals; the first order alone does not tell you the full value of a subscriber. Set up Control and the other test groups. For a simple comparison, split traffic equally between Control and Variant A.
Choose a primary metric on the Base info step.
2. Choose who enters the test in Targeting
Use audience targeting to choose which visitors can enter the test. For example, include all visitors or limit the test to the country where you offer subscriptions. Set the share of eligible visitors that enters the test. The test group percentages divide that share between Control and the other test groups. Visitors outside the test keep the store’s normal purchase options. Use targeting rules when you need a particular visitor to see a specific test group, such as when checking the setup yourself.
Traffic allocation controls how many eligible visitors enter. Audience filters decide who is eligible.
3. Choose the provider and products in Subscription offer
Select Recharge, Stay.AI, or SKIO. If prompted, connect it in the dialog. Follow the Recharge, Stay.AI, or SKIO setup guide if you have not connected the provider yet. Select every product you want to include, then select Continue to plans. All test groups use this product selection. Use Edit products to change it later. If a product is missing, check its plans in the selected provider and refresh the plan list.
Select all products for the test before choosing test group offers.
4. Set Purchase options for each test group
In Purchase Options, set Purchase type for each test group. Choose Subscription and a Shared plan, or choose One-time purchase, which needs no plan. At least one test group must use Subscription. A shared plan applies the same offer across every selected product, such as monthly delivery with 20% off. The comparison table shows the resulting price for each product and test group. Prices are per unit and can vary by product variant.
One-time versus subscription across two products in one test.
5. Keep the purchase options separate
Each visitor sees one-time purchase and, in a subscription test group, the plan chosen for that test group. They never see another test group’s plan or a plan the test created for another test group:
Purchase-option visibility stays on while a test owns Recharge plans. The summary shows the offer assigned to each test group.
6. Preview and launch from Review
Continue to Review. Check the visitor split and the product and purchase option assigned to each test group. Use Preview to inspect every test group on your storefront before launch. For each selected product, check the following:- The purchase option matches the test group.
- The price, product variant, quantity, and delivery interval are correct.
- The cart keeps the intended one-time purchase or subscription plan.
- Checkout shows the expected recurring terms for a subscription purchase.

Use Preview from Review before sending real traffic to the test.
Compare one-time purchases and subscription offers
Choose the setup that answers your question:- One-time test group
- Subscription test group

Control shows one-time purchase without a subscription choice.
write_plans permission. Creating a plan saves the draft first. Set the discount and delivery frequency, then select the new plan. SKIO and Stay.AI plans must already exist in those apps.
A Stay.AI plan, a SKIO plan, or a Recharge plan you create in Recharge stays on the product for every shopper unless its name starts with *, as in * Deliver every month. Shoppers assigned to that plan see it. The product page and other plan lists in your theme drop the * and the space after it. Remove the * when you want every shopper to keep seeing the plan. Plans you add with Create plan are hidden without that prefix, including from the Recharge customer portal.

Create a Recharge plan with its own name, discount, and delivery interval. Selecting Create plan saves the plan in Recharge.
Include multiple products in one test
Select all products together, then choose a Shared plan for each subscription test group. ABConvert lists offers available across every selected product. For Recharge, matching includes the discount, billing and delivery schedule, quantity, and other plan terms. Matching plan names alone are not enough. For example, the screenshot uses two products with the same monthly 20% discount. Control offers either product as a one-time purchase. Variant A offers either product on that monthly subscription. A shopper receives the same test group across both products; the shopper does not need to buy both products. If No Shared Subscription Plan appears, add matching plans in your provider or remove a product with different terms. Then select Refresh plans. Create plan inside ABConvert is available for a single selected product; prepare matching multi-product plans in Recharge instead.Use the APIs and MCP
You can also prepare a draft with a connected assistant or your own integration. The details in this section are optional for store setup.Developer setup with the REST and JavaScript APIs
Developer setup with the REST and JavaScript APIs
type to subscription, choose the provider in shared.subscription, and give every test group one subscription change for every tested product. Every test group must cover the same products.A subscription row names a Shopify Product GID and an existing provider plan ID. SKIO and Stay.AI also need the Shopify selling plan ID. For a one-time test group, set purchase_type to one-time and omit both plan IDs. The API saves a draft; preview and launch use the same provider checks as the admin.The JavaScript API exposes getSubscriptionOffers() for every product in the visitor’s assigned test group. One-time rows have no plan ID. It also reports type: 'subscription' in getAssignments() for a standalone test. An assignment describes the selected change; the browser script confirms application separately on a matching product page.Create a draft with a connected MCP assistant
Create a draft with a connected MCP assistant
list_shop_products to find every product and list_subscription_plans to find plans owned by the connected provider. Pass one offers row per product in each test group to create_subscription_test. Set a test group’s purchaseType to one-time to leave its rows plan-free. MCP can create the draft and run get_launch_checks, but it cannot launch a standalone subscription test.Try this prompt with your connected MCP assistant:Use subscription with a combined test
In Create new test → Combined, use Add change → Subscription. Choose Recharge or SKIO (Stay.AI isn’t available in combined tests), select the products, and configure Purchase type and Shared plan for each test group. The subscription setup follows the same product and shared-offer flow as a standalone test. Add other supported changes, then preview the complete combined test for every test group. All changes share the test group’s assignment. Results measure the complete package of changes, not the separate effect of the subscription offer. SKIO plans must already exist in SKIO; Recharge plans can also be created in ABConvert. Price changes, offer discounts and URL redirects cannot be combined with a subscription component. ABConvert checks the provider connection, product and plan before preview or launch. Incompatible choices are unavailable in setup.Use subscription with personalization
In Personalizations → Create personalization, add a Subscription change. Choose Recharge, Stay.AI, or SKIO and select the products. Choose Subscription with an existing Shared plan, or One-time purchase for all selected products. Set the audience, preview, and roll out the personalization. Visitors who do not match keep the normal storefront, including their original plan choices. An assigned subscription test takes priority over an overlapping subscription personalization; eligible personalizations use the most recently activated experience. Recharge plans can be created in Recharge or explicitly kept after an ABConvert test. A Recharge plan still owned by a test cannot be selected for personalization because that test can remove it during cleanup. SKIO and Stay.AI plans must already exist in those apps. In-app plan creation is unavailable for personalization. Personalization applies a plan to a new purchase. It does not change existing subscriber contracts. Its lifecycle results describe the targeted cohort, not causal lift against the normal storefront: this setup does not create a randomized holdout.Read the results
Subscription Metrics uses Shopify first orders with a subscription line. It shows subscription orders, subscription revenue, units, and rates by test group. Renewals are separate: Subscription Lifecycle uses verified Recharge, Stay.AI, or SKIO events for subscribers, retention, churn, and renewal revenue. See Subscription metrics for the metric definitions.
First-order subscription metrics from the demo store. These sample figures are not evidence that one offer is better.

Lifecycle results follow subscriptions over time. The demo subscriptions are too recent for the retention windows.
End a test and keep a plan
While a test runs, ABConvert hides the plans a shopper shouldn’t see. When it ends, shoppers see every subscription plan on those products again, so decide what comes next before you end it:- Keep one test group’s offer. Roll it out as a personalization. ABConvert ends the test and shows that plan or one-time purchase to the audience you choose, hiding your other plans. The audience starts as all visitors, because shoppers outside it see every plan.
- Show all plans again. End the test.
Move existing subscribers to the winning plan
You don’t have to move anyone. Subscribers on a losing plan keep working as they are. If you want everyone on one delivery frequency, your subscription app does the move, not ABConvert:Wait for the results you need
Tell your subscribers first
Move one subscriber and check the price
Update the rest in bulk
- Recharge: Bulk update subscriptions with the Import tool
- Stay.AI: bulk updates from the Subscriptions page, described in its release notes
- SKIO: Bulk operations guide
If preview is blocked
Check the message in the test review step. An active test on the same product can block Preview or launch. End or pause the conflicting test only if you intend to change its traffic and results, then retry Preview. Confirm the provider connection and each test group’s plan if the message points to either of those instead.Where another plan can still show
While a test is running or in preview, a shopper sees only their test group’s plan on the product page, in the Recharge cart drawer for a product already in the cart, and in another plan picker on the page when that picker belongs to one product. Search suggestions, recommendation cards, and on-site upsells are included when they offer a plan choice. The product image and title stay. Checkout for a product already in the cart uses the plan chosen on the product page. It does not list that product’s other plans. A shopper can still meet another plan in these cases:- Checkout and post-purchase upsells. These add a product and choose a plan on their own. Leave the product in the upsell. Do not offer a plan from a test group that shopper is not in. A plan name that starts with
*is one of those plans. Add the product with no plan, or with the plan you sell outside the test. - The first moment of the page. The extra plan can flash while the page loads.
- The app embed is off. Turn on the ABConvert app embed. With it off, shoppers see every plan.
- A plan picker ABConvert does not recognize. A custom widget can still show the plan. If a plan list is not tied to one product, a name that starts with
*is hidden only when that shopper is not in a running or preview test. - The name saved on the plan. Checkout, order email, and your subscription app’s portal keep a leading
*. The product page and other plan lists in your theme drop the*and the space after it. - Before the test is running or in preview. Search suggestions, recommendation cards, and on-site upsells can still list every plan. The product page still hides a plan whose name starts with
*. - A subscription already in the cart. That line keeps its plan while the test is a draft, paused, or ended. ABConvert changes the line only for a shopper who is in a running or preview test.
Common mistakes
- Offering a test plan in a checkout or post-purchase upsell. Those offers choose a product and a plan on their own. Leave the product in the upsell, and point it at no plan or at the plan you sell outside the test.
- Launching a Stay.AI or SKIO plan without a leading
*. Every shopper can see that plan. Start the plan name with*before launch, and remove the*only when you want to keep the plan. - Expecting SKIO or Stay.AI plans to be created in ABConvert. Create them in the app before you choose the product and plan in the test.
- Testing on a template without the widget. Shoppers on a product template without your app’s widget see no plans. Check every template the test can show, including any used by a template test.
- Running two subscription apps on one product. Shoppers see both widgets, and a test only controls its own app’s plans on the product page. ABConvert warns you in the test setup.
- Adding products whose offers only share a name. Match the actual plan terms across products before choosing a shared offer.
- Judging renewal value before the billing date. Allow subscribers to reach the renewal periods you want to compare. First-order revenue alone cannot show later subscription value.
- Treating Preview as a renewal check. Preview shows the storefront offer. Verify provider events separately in Recharge, Stay.AI, or SKIO.
FAQ
Why can't I turn off Show only the assigned subscription plan?
Why can't I turn off Show only the assigned subscription plan?
Why can a checkout upsell still sell another plan?
Why can a checkout upsell still sell another plan?
Why can visitors outside my test see other purchase options?
Why can visitors outside my test see other purchase options?
