Skip to main content

Testing options

What types of tests can I run and why would I choose each one?

James Harber avatar
Written by James Harber
Updated this week

There are different options for you to choose from when selecting which type of test to run. Each option has its own merits and also limitations so which should you choose and why? This guide should help you pick the correct one for what it is you are trying to achieve.

1. A/B/n

Core idea

Compare one baseline (A) against two or more variations (B, C, D …) simultaneously.

What actually happens

Incoming traffic is randomly split across all variants at once (e.g., 25 % each for A, B, C, D).

Typical use-case

You have several alternate headlines, colour schemes, or page layouts and want to see which performs best.

Key point

Still a “single-factor” test—the page changes as a whole, but only one version of the page is seen at any one time.

You are able to use both the visual and advanced code editors to build Abn tests.

Learn more about building Abn tests in the visual editor here.

Learn more about building Abn tests in the advanced editor here.


2. Split (also known as redirect/bounce tests)

Core idea

Each variant lives on a different URL; users are silently redirected to one of those URLs.

When to choose it

• Very large or back-end changes that are hard to inject with client-side scripts (e.g., different checkout flows).
• You need to test server-side technologies, CMS templates, or entire sub-domains.

How it differs from A/B/n

In A/B/n the testing tool rewrites the page after it loads; here the user lands on a completely separate page, so performance, SEO factors, or back-end logic can be evaluated too.

You are only able to use the advanced code editors to build split tests.

Learn more about building split tests in the advanced editor here.


3. Baseline

Core idea

Measure current performance of a metric/several metrics to understand the 'normal' behaviour.

Why it matters

Without understanding the 'normal' it can be difficult to identify if there is an opportunity to test nor how big the opportunity is.

Typical setup

No visible changes to the end user and a data collection exercise only.

You are only able to use the advanced code editors to build baselines.

Learn more about building baselines in the advanced editor here.


4. Target

Core idea

An experience shown only to a specific audience segment (the “targeted” group).

Two common senses

1. Targeted variation: You show a new hero banner only to visitors from paid-search campaigns.
2. Target metric focus: You declare one metric (e.g., average order value) as your “target”. The variant that maximises that metric wins.

Key distinction

The baseline stays broad and unchanged; the target group/variation is purposefully different and/or narrowly scoped.

You are able to use both the visual and advanced code editors to build targets.

Learn more about building targets in the visual editor here.

Learn more about building targets in the advanced editor here.


5. Multivariate Test (MVT)

Core idea

Tests multiple elements at once and every possible combination of those element states.

Example

You pick 3 headline variants and 2 button colours → 3 × 2 = 6 combinations served in parallel.

What you learn

• The individual effect of each element (main effects).
• Whether certain element pairings interact (interaction effects).

Trade-offs

• Requires far more traffic to reach significance because each combination gets only a slice of visitors.
• Analysis is more complex; gains are clearer when elements are independent.

You are only able to use the advanced code editors to build MVT's.

Learn more about building MVT's in the advanced editor here.


Quick decision cheat-sheet

If you need to…

Choose…

Compare several whole-page concepts quickly

A/B/n

Test an entirely different back-end or template

Split-URL / Redirect

Understand the 'normal' or current behaviour

Baseline

Serve a new experience only to a specific segment

Define a Target variation / audience

Discover which of several on-page elements actually drives lifts

MVT

Did this answer your question?