A/B vs. multivariate testing: What’s best for your website?

May 18, 2022 | 7 minute read
Christopher Santini
Senior Consultant, Oracle Maxymiser Professional Services
Text Size 100%:

So you’ve decided to test the effectiveness of your website but don’t know where to start. Maybe you’re thinking, what’s website testing, and what’s the best strategy for me?

Should I use A/B or multivariate testing, and what’s the difference between the two?

A/B and multivariate testing strategies possess distinct sets of strengths and weaknesses when it comes to website optimization. Let’s compare the two and determine which is the best option for your website.

A/B testing

A/B testing is a method of website optimization that compares how two versions of an element on a webpage perform.

In this method of testing, you alter a single element on the test page, such as the placement of a product side rail (see the diagram below). Each of these page versions is commonly referred to as a variant.

You would split traffic, most commonly 50/50, to the two-page variants and place visitors at random into either:

  • The control experience (variant A)
  • Or the updated experience (variant B)

From there, you measure which version of the page drives better key performance indicators (KPIs).

The goal is to determine which variant is more successful at getting visitors to complete a desired action. By measuring conversion-focused KPIs, you can analytically determine which version of the page is more effective at converting your visitors.

What KPIs can you test? You could try:

  • Call-to-action (CTA) clicks
  • Subscriptions
  • Purchases
  • Page views
  • Bounce rate

A/B testing visual exampleHear from Oracle Marketing Consulting’s email guru Chad White about the best ways to save time and money with your A/B testing. Check out his blog “A/B Testing pitfalls: How marketers can avoid costly mistakes.”

Multivariate testing (MVT)

Multivariate testing (MVT) is very similar to A/B testing, but you use it to compare multiple elements of a page at once, rather than just a single element.

In our previous example, we used A/B testing to test the placement of a side rail, a single page element. However, with MVT, you can test multiple site elements simultaneously, such as but not limited to:

Site visitors would be randomly shown a combination of these elements to determine how each new element on the site interacts with each other to form the most effective page for conversion. Rather than splitting the website traffic 50/50 between the default and the variant, like in an A/B test, traffic in a multivariate test is most likely split between a much greater number of variants to ensure you are accurately measuring every possible combination of page elements.

Apply the following formula to find the total number of variants in a multivariate test:

(# of Element A Variations) x (# of Element B Variations) x … = (Total # of Variants)

For example, in the diagram below, we have three elements—content, image, and pop-up. Assuming we have an existing and a new variation of each element (A and B), we would have two total variations for each of the three elements.

Therefore, this formula would give us eight (2 x 2 x 2) different possible combinations consisting of the control and seven variants. The variant structure would be as follows:

Element

Control

Variant 1

Variant 2

Variant 3

Variant 4

Variant 5

Variant 6

Variant 7

Content

A

A

A

A

B

B

B

B

Image

A

A

B

B

A

A

B

B

Pop Up

A

B

A

B

A

B

A

B

Multivariate testing visual example

Find out how Brazilian retailer Armazém Paraíba used testing and website optimization to help better manage their sales channels and understand consumer behavior.

Which is right for you?

When determining which method is best for your website, you should consider many factors, such as your goals, web traffic, time frame, and expertise.

When first beginning website optimization, it’s probably best to start with A/B testing, as it is a less complex testing method with a simpler test structure.

A/B testing is also a better option if the amount of traffic you receive on your web page is low. By having fewer variants, tests can reach statistical significance and conclude quicker than MVTs, which require greater web traffic to provide reliable data.

Should you want to test a single page element but have more than two versions of that element, an extension of A/B testing—known as A/B/n testing—is the answer. A/B/n testing allows an extended number of variants to exist, from the third to the “nth” version, to determine which performs the best.

A/B tests are the most common form of testing as they can be implemented easily and provide accurate data quickly to drive website optimization.

On the other hand, multivariate testing is much better suited to large-scale tests, such as home page redesigns where many elements of the page, such as imaging, copy, or even specific audience segments, are to be tested at once.

Suppose you have the luxury of a longer time frame and a high level of website traffic. In that case, running an MVT is a very powerful way to test the effectiveness of multiple page elements simultaneously. It will save you the hassle of creating numerous tests.

While MVTs may sound intimidating, they can be an invaluable source of information for uncovering the best combinations of page elements. However, when testing on dynamically changing sites, as a general rule, avoid MVT testing, as winning experiences derived from MVT tests rely on the interplay between the different elements of that experience. Therefore, when conducting large-scale tests on dynamically changing pages, A/B/n testing allows for faster iteration of new test ideas.

It’s important to analyze your internal objectives and resources before building out your testing strategy. A/B testing and multivariate testing are two different paths to the same goal: providing consistently optimized experiences for your consumers.

You can use these methods exclusively or interchangeably, but it’s important to view website optimization as an ongoing process regardless of which method you use. As consumers, markets, and commercial climates change, so should your testing strategy. Doing so ensures your website is consistently optimized to drive the highest conversion rates and the most optimal customer experiences.


Find out more about how testing optimizes your website, landing pages, digital marketing, email campaigns, and more. Check out these resources about A/B and multivariate testing.

Christopher Santini

Senior Consultant, Oracle Maxymiser Professional Services

Christopher Santini is a Senior Consultant for the Oracle Maxymiser Professional Services team with experience working closely with large global brands in the Financial Services and Insurance industries. Christopher spearheads optimization strategy which drives digital transformation by implementing industry best practices and leveraging Oracle Maxymiser’s best-in-class platform and professional services.


Previous Post

What Walt Disney knew about failure

Analisa Flores | 3 min read

Next Post


Marketing automation today: Myths, truths, and trends to drive results

Michael McNichols | 10 min read