Split Testing

Definition

Split testing is an experiment that divides visitors into groups and shows each group one version of a page, element, or flow, so you can see which version performs better. In common use it means the same thing as A/B testing . If you are looking for software, go to split testing software .

What Split Testing means in A/B testing

The "split" is the audience, not the page. Everyone in the test is randomly assigned, each person sees one experience, and you compare those groups on a metric chosen in advance. People also use the phrase for a broader redesign, and they use split URL testing when each version lives on its own address. Those are scopes of the same idea. The split testing vs A/B testing post is where that wording is sorted out.

Why Split Testing matters

Without a split, a before-and-after change mixes the new page with a new week of traffic. The split holds the time period constant and lets the version be the thing that differs. Calling every comparison a split test, including a sequential launch with no randomization, throws away that guarantee.

Example of Split Testing

A landing page sends half of new visitors to the current headline and half to a shorter one, on the same URL, over the same two weeks. Signup rate is the metric. That is a split test. Sending the same two headlines to two different URLs and splitting traffic between those URLs is still a split test, and it is specifically a split URL test. Looking up a tool to run either one is a different query, answered on the software page.

How to use Split Testing

Name one change, one audience, and one metric. Randomize by visitor, not by day. Read the result after the sample you planned. Use the comparison post if you are unsure whether you mean an on-page change or a split URL. Use split testing software if you are choosing a tool.

Common mistake

A common mistake is treating "split testing software" and "what split testing means" as the same search. They are not. Another is running a split URL test and then forgetting a canonical, which is an SEO problem the definition here does not solve. The split-URL glossary entry and the software page cover the implementation.

Related A/B testing terms

• A/B Testing
• Client-Side Testing
• Split URL Testing
• Server-Side Testing

Related guide

Split testing software: The product page for running split tests. This glossary entry only defines the term.

FAQ

Is split testing a different method from A/B testing?

Usually no. Both names mean you split the audience and compare versions. Some teams reserve 'split testing' for a full-page or split-URL comparison and 'A/B testing' for a smaller on-page change. The comparison post walks through that usage. The statistic is the same either way: one randomized split, one primary metric.

Where do I go if I want split testing software?

Use the split testing software page. This glossary entry does not compare tools or describe how to install one. It only says what the method is.

Does every split test need two URLs?

No. Two URLs is split URL testing, which is one way to run a split. A headline change on a single URL is still a split test if visitors are divided between the original and the edited headline.