Split testing vs A/B testing

Definition: Split testing is another name for A/B testing: splitting visitors at random between two versions and measuring which converts better. In practice "split test" often means a URL split test, where each version is a separate page and visitors are redirected, while an A/B test changes an element within one page. The statistics are identical.

Same method, two names

Both terms describe random assignment of visitors to versions and a comparison of conversion rates. Marketers and tool vendors use them interchangeably; "split testing" is older and slightly more common in email marketing, "A/B testing" in web optimisation. Nothing about the design, the sample size or the significance test changes with the name.

URL split tests

The versions are two separate pages, often two designs of a landing page. The testing tool sends a share of visitors to the second URL and keeps them there on later visits. This suits redesigns and layouts too different to swap element by element. It needs care with search engines: the variant should carry a canonical link to the original so the two pages are not indexed as duplicates.

Element tests

Both versions live in one page; the tool shows one visitor the original headline and another the alternative. There is no redirect and nothing new to index. This suits testing one change at a time, which is also the design that tells you which change mattered.

Which to choose

Element test when you want a clean answer about one change. URL split test when the versions differ in structure, or when a page is built in a tool that makes duplicating it easier than editing it in place.

A worked example

A store wants to test a new checkout layout. As a URL split test, the new layout is published at a second URL and half of the checkout visitors are sent there; the canonical on that page points back to the original. As an element test, the store would instead test one change inside the existing checkout, such as moving the trust badges above the pay button. The first answers "is the new design better?"; the second answers "do the badges help?".

Common mistakes

• Running a URL split test without a canonical, so search engines index both versions.
• Leaving the redirect in place for months after the test ends.
• Sending a visitor to a different version on each visit; assignment must stick.
• Testing a whole redesign when one element is the real question.

In AB Test WP

AB Test WP runs both kinds: element tests matched point-and-click or by a CSS class, and URL split tests whose redirect runs in the page head before anything paints, so cached pages work. The variant's canonical points at the original by default, with noindex and off as the other options, chosen once in Settings and applied to every URL test on the site, and a URL test can send visitors to a page on another subdomain or site, where the plugin emits no canonical or noindex of its own but still credits the conversion back through a signed token the entry site issues.

Related: Element tests vs URL split tests, in detail · How to split test landing pages in WordPress · all glossary terms.