Test a different page with a split URL test

A split URL test (also called a redirect test) is useful when the new version is a whole different page, for example a new landing page your developer built, or the same page with a different layout under another address. You don't change anything with the visual editor or AI: you compare finished pages.

Create the test

1. Open Optimize → Experiments and click New test.
2. Under What do you want to test?, choose Send visitors to another page (split URL).
3. Fill in Name of the test.
4. Under Original page, paste the address of the page visitors arrive on today, for example https://yourshop.com/landing. This page stays the original.
5. Under Variant pages, paste the address of the new page next to B. Click + Add variant page to test up to three pages (B, C and D).
6. Click Create split URL test.

Variant pages must be on your own site's domain (a subdomain is fine) and start with https://. A query string such as ?layout=new is fine. Varly refuses other websites, so a test can never send your visitors to someone else's site.

Varly then opens every page, takes screenshots on desktop, tablet and mobile, and checks whether the Varly snippet is on each variant page. This usually takes a minute or two. Then the status changes to Ready to review.

Review the pages

The experiment page shows:

• Original and variant page: screenshots of each page. Switch between Desktop, Tablet and Mobile.
• Pages in this test: every page in the test, with its full address. Under each variant page, Varly says whether it measured visits to that page in the last 30 days. No visits is normal for a new page, but then check that the snippet is on it. Varly counts visits per page, not per query string: a variant on the same page as the original with only a different query string (for example /?layout=new) is marked "Same page as the original"; if the snippet is on the original, it is on that one too.
• Check before launch: warnings, for example a page that answered with an error, a page that forwards to another address, or a page where Varly couldn't find its snippet.
• Preview on your site: a link per page. See Preview the pages below.

Preview the pages

Under Preview on your site, click Open the original or Open Variant B. The page opens on your real site with the Varly preview bar at the bottom. Click another version in the bar to go to that page. Only you see the bar, nobody is sent anywhere, and nothing is measured. Use Copy link to share the page you're looking at. See Preview and check a test.

Launch

Launching works like any other test: pick the goals, who takes part and the plan, then choose Dry run first (recommended) or Live. See Launch, pause, resume and end a test.

Some goals don't work for a split URL test:

• Clicked an element isn't available: each version is a different page, so a click on one page can't be compared with the other.
• The test page and the variant pages can't be the goal, because every visitor in that group lands on them. Pick the page visitors should reach next, such as a thank-you page, or Added to cart or Placed an order.

In a dry run, visitors are split and counted, but nobody is sent to another page. When the split looks fine, click Go live: send visitors to the variant pages.

What visitors see

• Visitors in the original group see your page as always.
• Visitors in a variant group briefly load the original page. Varly counts them, then sends them on to their variant page right away. The address keeps the query string and the part after #, so UTM tags and ad click IDs still work.
• Each visitor keeps the same version on every visit in the same browser.
• Only visitors who accepted cookies take part. A visitor who accepts cookies only after about five seconds, or whose page loads very slowly, sees the original that time and isn't counted, so no one is sent away while they're already reading.
• A visitor who arrives straight on a variant page, for example from an old link, is not counted in the test.

Varly counts every visitor on the original page, in the version they were given. So the results, goals and Results by segment work the same as for any A/B test. See Read your test results.

Common problems

• "The variant page must be on your own site": the address is on another domain. Use a page on your own domain, or add the domain under Site settings first.
• "Variant B is the test page itself": you pasted the original page as a variant. Paste the address of the new page.
• "Varly couldn't find its snippet on …": add the Varly snippet to that page, the same way as on the rest of your site. Without it, Varly can't measure what visitors do on that page, such as adding to cart. If you install Varly with a tag manager or after cookie consent, Varly may not see it in the check; open the preview link to make sure.
• "The pages could not be checked": one of the pages didn't load, for example because it doesn't exist yet or isn't public. Check the address in your browser, then click Try again.

For developers

The snippet reads the test from its built-in configuration, so there is no extra request. Before it sends a visitor on, it sends the exposure with keepalive, then calls location.replace, so the original page doesn't stay in the browser history. Query parameters of the current address are added unless the variant address already has a parameter with that name. The snippet never sends a visitor on when they're already on that variant address, and the page directly after a redirect never counts or redirects again for the same test.