Split URL tests

Compare store pages or a checkout on another domain: prepare destinations, check reachability, and understand which visits and orders count.

A split URL test compares pages instead of changes: the control is the origin page visitors already reach, and each variant is a different page of the store or an HTTPS checkout page on another domain. A visitor assigned to a variant is navigated to that variant's destination URL, once. Store destinations are measured after arrival; checkout destinations on another domain count the visit when it leaves the store. Campaign parameters on the original address (utm, gclid and the like) survive the navigation, so ad reporting keeps working. The control never navigates — it stays on the origin page.

Traffic in a split URL test is always split by Proa itself: the assignment and the navigation have to be one decision, so external split tools cannot drive it. See Traffic split and targeting.

1. Build the destination page

A destination on the store needs a real, published page carrying the same Proa snippet as the rest of the site. Build it however you want — it is a full page, not an overlay. Then prepare it for testing:

  • Point the destination's canonical at the origin, or mark it noindex. Otherwise search engines index two competing pages. Proa does not write to your theme — this step is yours.
  • Keep the destination out of the navigation, the sitemap, and your ads. Traffic that lands on it directly is outside the test and is not measured; the test only reads visitors who entered through the origin page.

Checkout pages on another domain

Turn on Checkout pages on another domain in Site settings to use a Checkout Champ funnel. Use its first page as the destination. Install its Shopify plugin and enable Send Initial Orders for that campaign.

Proa counts the visit before sending it to the checkout. Orders count when the checkout exports them to Shopify. Visitors arriving directly at the checkout, zero-value orders and funnels without that export stay outside the test. Cart and checkout steps on that page are Not measured.

After creating the draft, use Copy test link to place a real order and check its attributes in Shopify under Additional details. Test cards do not reach Shopify. Orders through test or preview links do not count in the experiment.

With analytics permission, the checkout URL also carries a visit identifier and, when visitor data is enabled, a visitor identifier. Your checkout and its scripts can read and retain those values. See the Privacy Policy and DPA for the limits of consent and deletion controls.

2. Point each variant at its destination

In the wizard, set the Origin URL — the page visitors already reach — and a Destination URL per variant. The control shows "Stays on the origin URL"; it needs nothing.

The honest cost: one extra navigation for variant visitors, on their first arrival at the origin page. The wizard states it up front. If the origin page is highly latency-sensitive, weigh it.

3. Check the destinations before the test starts

The wizard checks each destination on your store and names the outcome per variant: the page exists and carries the snippet, or what is wrong — page not found, snippet missing, page unreachable, or a storefront password in the way. A destination without the snippet would receive visitors and measure nothing, so the check exists to catch exactly that before any visitor is split.

The test page keeps the same checks after creation, and QA tools adds a checker that shows where each group lands — paste the origin URL as a visitor would reach it and see which variant goes where. Each variant also gets a preview that puts your own browser session in that variant, without affecting any visitor or any number.

External destinations show Reachable when the page responds. This check does not confirm the Shopify export or order attribution.

What is measured and what is not

  • A visitor entering through the origin is counted for the assigned variant. Store destinations measure the funnel there; external checkouts count orders once they reach Shopify.
  • A visitor who reaches a destination page directly — from a shared link, from search, from an old ad — is not part of the test. That traffic is excluded rather than silently counted, which is why the destination stays out of the navigation.
  • Screen size targeting applies to split URL tests; page lists do not, because the test is anchored to one origin page by definition.

What is fixed and what is editable

Destinations are editable while the experiment is a draft; once it has gone live they are part of the record. Everything else follows the normal experiment lifecycle.