Multivariate Testing

Conversion & UX

Also: MVT · Multivariate Optimisation

Combinations = variants of element 1 × variants of element 2 × variants of element 3 (and so on)
CombinationsProduct of all variant counts
Traffic demandMuch higher than A/B testing
Use whenYou want to find winning combos
Not forLow-traffic pages

Quick definition

Multivariate testing (MVT) is a method of testing multiple page elements at the same time to find which combination performs best. Unlike an A/B test that compares two whole-page versions, MVT isolates the effect of each element, such as a headline, image, and button, across all possible combinations simultaneously.

Run the numbers
Total combinations to test12.00

Each combination needs sufficient traffic to reach statistical significance. Add more elements and the required traffic scales fast. If the total combinations exceed what your page traffic can support in a reasonable timeframe, run a focused A/B test instead.

How it varies across Australia

MVT requires substantially more traffic than A/B testing to reach statistical significance across all combinations. Most Australian mid-market sites lack the volume to run MVT reliably on anything but their highest-traffic pages. The conversion gain from a well-run A/B test on most pages outweighs the theoretical upside of MVT.

See conversion benchmarks across Australian industries

MVT vs A/B testing

A/B Testing(A/B)

Compares two or more complete page variants against each other. Fast to run, low traffic requirement, answers one question at a time.

Best for: most pages, most budgets
Multivariate Testing(MVT)

Tests multiple elements simultaneously across all their combinations. Slower, traffic-hungry, reveals interaction effects between elements.

Best for: high-traffic pages, element interaction questions
Full Factorial MVT

Every possible combination is tested. Most statistically robust. Requires the most traffic.

Combinations = product of all variant counts
Fractional Factorial MVT

A statistical subset of combinations is tested. Lower traffic requirement. Less complete picture of interaction effects.

Common in tools like Google Optimize (now sunset)

What it actually means

Imagine you want to test three things on a landing page: the headline (two versions), the hero image (two versions), and the call-to-action button colour (two versions). That's two times two times two, which gives eight combinations. MVT runs all eight simultaneously, with real visitors split across all of them, so you can see not just which headline works but which headline works best when paired with which image and which button colour.

That's the promise. The catch is that eight combinations each need enough traffic to reach statistical significance before you can trust the result. If your page gets five hundred visitors a month, you'll be waiting years. If it gets fifty thousand, you might be done in a fortnight.

The other thing MVT reveals that A/B testing can't is interaction effects. Two elements that individually test neutral can produce a significant lift when combined, or two individually strong elements can cancel each other out. MVT is the only method that finds this.

Most conversion teams reach for MVT when they should still be running A/B tests. The traffic maths almost always resolves that question. Run the combinations calculation first.

MVT tells you which combination wins. A/B testing tells you whether your hypothesis was right. Most pages need the second answer more urgently.

How to calculate it

Combinations = variants of element 1 × variants of element 2 × variants of element 3 (continue for each additional element)

Worked example. You want to test a headline with 3 versions, a hero image with 2 versions, and a CTA button with 2 versions. Combinations = 3 × 2 × 2 = 12. Each combination needs enough traffic to reach significance. If you need 500 visitors per combination for 80% power, you need at least 6,000 visitors before you can read the results reliably.

The Australian context

Australian ecommerce and lead-gen sites rarely have the traffic volumes that make full factorial MVT practical outside the top handful of pages. The sites where MVT makes sense tend to be high-volume retail and financial services landing pages, where even a small percentage lift is worth the testing investment. For most mid-market Australian businesses, the better use of conversion budget is disciplined A/B testing on the pages that matter most, before graduating to MVT.

Where people get this wrong

Running MVT on a low-traffic page.The more combinations you test, the more traffic you need per combination to reach statistical significance. Underpowered MVT produces noise dressed up as results.
Testing too many elements at once.Each additional element multiplies the number of combinations and the required traffic. Three elements with two variants each is manageable. Five elements with three variants each is almost never achievable.
Stopping the test when one combination looks good early.Early leaders in MVT frequently flip as traffic accumulates. Stopping early is how you ship a change that looked like a win and was actually noise.

Related terms

Common questions

When should I use MVT instead of an A/B test?

Use MVT when you have a high-traffic page, you want to test multiple elements simultaneously, and you specifically want to understand how those elements interact with each other. If you're testing one clear hypothesis with limited traffic, an A/B test will give you a reliable answer faster.

How much traffic do I need for multivariate testing?

It depends on your number of combinations, your baseline conversion rate, and the lift you want to detect. A rough guide: multiply the number of combinations by the sample size you'd need for a single A/B test. More combinations means substantially more traffic before results are trustworthy.

What tools support multivariate testing in Australia?

VWO, Optimizely, and AB Tasty all support MVT. Google Optimize is no longer available. Most Australian teams run MVT through the same platform they use for A/B testing. The tool matters less than whether your traffic volume can support the test.

Can I run multivariate testing on a mobile app?

Yes. Most major experimentation platforms support in-app MVT. The same traffic constraints apply. Mobile apps with large daily active user bases are often better suited to MVT than equivalent web pages because the session volume is higher and more predictable.

Keep exploring

About New Rebellion

New Rebellion is a marketing intelligence consultancy. We build tools, score Australian businesses on how their marketing actually performs, and publish Debrief every day. This dictionary is part of how we work in the open.

How we think →