Split URL testing: what it is and how to run one

What split URL testing is

Split URL testing is a type of split test where each version of a page lives at its own URL. Visitors who land on the original page are split at random: some stay, and the rest are redirected to the new page. With a 50/50 split, half of the visitors to /pricing see /pricing-new, and the test compares which page leads to more sign-ups or purchases.

People often use "split testing" to mean any A/B test. A split URL test is the specific kind where the versions are separate pages, which is why it’s also called a redirect test.

Split URL test, A/B test or multivariate test?

The three test types answer the same question, which version converts better, but they deliver the versions in different ways. Pick the one that matches how big the change is.

• A/B test: the testing tool changes elements on the same page, such as a headline, an image or a button label. It suits most changes and is the quickest to set up. See what A/B testing is.
• Split URL test: the new version is a whole page at another address, built in your CMS or by your developers. It suits redesigns, new templates and pages that would take dozens of edits to rebuild in a visual editor.
• Multivariate test: several sections of one page change at once and every combination is tested. It needs much more traffic. See how multivariate testing works.

The statistics are the same for all three. A split URL test needs as many visitors per version as an A/B test of the same size of change.

When to use a split URL test

Use a split URL test when the new version is easier to build as its own page than as a set of edits to the old one. That usually means a big change, such as:

• A redesigned landing page or homepage with a new layout.
• A new product page template that changes the order of the photo, price, reviews and buy button.
• A shorter checkout or sign-up flow on a different set of pages.
• A page built on a new platform or page builder before you switch the whole site over.

For a single headline or button, an A/B test is simpler, because there is no second page to build, track and remove afterwards.

How a split URL test works, step by step

A split URL test runs on the original page’s traffic and sends a share of it on to the new page. Setting one up takes four steps:

• Build the new page and publish it at its own URL, with the same tracking as the original.
• In your testing tool, enter the original URL as the control and the new URL as the version. Some tools also accept a pattern, so /product/shoes can map to /product-new/shoes for every product.
• Pick the goal that decides the test, such as a purchase or a submitted form, and set how traffic is split.
• Run the test until it has enough visitors to call a result, then publish the winner at the original address and remove the test page.

Each visitor keeps the version they were given, so someone who saw the new page sees it again when they come back.

Common split URL testing mistakes

Most split URL tests that go wrong break at the redirect. Check these before you start:

• Losing campaign tags. If the redirect drops the query string, UTM parameters and ad click IDs disappear, and paid traffic on the new page looks like direct traffic.
• A flash of the old page. If the original page appears for a moment before the redirect, visitors in the new version see both pages, which skews the result.
• Missing tracking on the new page. If the goal or the analytics code isn’t on the new URL, that version reports almost no conversions.
• Uneven traffic. Redirects that fail for some browsers, or visitors who leave before the redirect finishes, can make the split differ from what you set. That’s a sample ratio mismatch, and results can’t be trusted until it’s fixed.
• Redirect loops. If the new URL also matches the test’s targeting, visitors can be sent around in a circle.

Does split URL testing hurt SEO?

Split URL testing doesn’t hurt SEO when it’s set up the way Google’s guidance on website testing describes. At the time of writing, that guidance says:

• Add a rel="canonical" tag on the new page that points to the original URL, rather than a noindex tag.
• Use temporary redirects (302) or JavaScript redirects, never permanent (301) ones, so the original stays indexed.
• Show search engines the same thing you show visitors, with no special version for crawlers.
• End the test once you have a result, and remove the test URL.

Split URL testing in Cuttl

In Cuttl’s Web Experiments, a split URL test is one of the test types, next to A/B, multi-page and multivariate tests. You enter the original URL as the control and the new page’s URL as the version, then pick the goal.

• Each version can send visitors to an exact URL, to a pattern on the current URL (so a whole group of pages maps to new ones), or to a URL returned by your own code. A test can have up to 8 versions, including the original.
• The redirect keeps the visitor’s query string, so UTM tags carry over to the new page.
• The original page is held back until the redirect happens, so visitors in the new version don’t see the old one first. The redirect replaces the original in the browser history, so the Back button doesn’t bounce visitors back into the test.
• Cuttl counts the visitor in the test before the redirect, and the report warns you if visitors aren’t split as set up.
• Both pages get their own heatmaps and recordings, so you can see where visitors on each version clicked and stopped.

Split URL tests are available on every Cuttl plan. When the new page wins, you publish it at the original address and finish the test.

Common questions