What is multivariate testing (MVT)?

Multivariate testing (MVT) is a form of experimentation that allows you to test multiple variables from a SaaS product at the same time. The goal of multivariate testing is to determine the combination of variations with the maximum positive impact.

Multivariate testing vs. A/B testing method

Both multivariate testing and A/B testing are methods of experimentation used to improve the performance of an app or website. But there are key differences between the two.

Also known as split testing, A/B testing compares the performance of variations of a product, in-app experience, or webpage to determine which performs better against a control group. For example, a project management platform may test whether a drag-and-drop interface will improve product adoption.

On the other hand, multivariate testing enables you to test multiple elements of your app at once.

For example, a social media management platform trying to launch a new post-scheduling feature may release different versions to test the placement of the feature, the button text, and the feature’s tooltip.

The differences between multivariate and A/B testing can be summarized as follows:

• The number of variables: A/B testing compares one variable against a control group, while multivariate testing compares two or more variables or versions against each other.
• Number of versions: There are two versions in an A/B test – a control version and a variation version. Conversely, multivariate tests compare multiple variations (depending on the number of variables tested).
• Audience: You can only conduct an A/B test on one audience segment at a time, while a multivariate test can be conducted on the same or multiple audience segments at the same time.

Pros of multivariate testing

Using multivariate testing is ideal when you have to test different possible combinations of elements or variables. Some of its benefits include:

• It saves you time by enabling you to test all the possible combinations of variables at once. This differs from an A/B test, which would require multiple tests with only two variations per test.
• It helps you see how different elements interact with each other, making it easier for you to redesign a feature to have the most impact.
• Multivariate testing requires a larger audience for results to have statistical significance. This makes its test results more significant than A/B testing.
• Thanks to the multiple variables you are testing and the significant data points they’ll produce, you can extrapolate results to other unknown situations.

Cons of multivariate testing

There are also some drawbacks to conducting a multivariate test in SaaS, including:

• Running multivariate tests is more complex and time-consuming than running a traditional A/B test. Because it is a form of full-factorial testing, the number of variations needed for a test can add up quickly.
• Testing multiple variations at once reduces the allocated traffic for each variation. As a result, multivariate testing has higher traffic requirements than A/B testing to attain statistically significant results.
• Because you’re testing multiple variables at once, it may be impossible to tell if one of the variables tested has any measurable effect on the conversion goal.

How to run multivariate tests

Running a multivariate test can feel confusing and complex, but it doesn’t have to be. Here’s a step-by-step guide to help you obtain meaningful results from your test.

Define what you need to test

When running a multivariate test, the first step is to define what you need to test. Start by setting a testing goal. Is the test geared toward conversion rate optimization? Are you trying to reduce churn or improve the user experience?

Once you know your goals, you can identify the specific elements of your product that you need to test. You can also determine how many variables or variations you need for your test.

Some elements you can test in a multivariate test include:

• Call-to-action buttons
• Product Features
• Onboarding flow and steps
• Marketing campaign elements
• In-app messages and notifications

Formulate a hypothesis

Next, you need to formulate a hypothesis. A hypothesis is a prediction about how a change to one or more elements of your product will affect user behavior.

For example, you may have a hypothesis that adding a new product feature will boost customer retention. Or, your hypothesis may be that changing the position of an in-app notification will improve conversion rates.

When formulating a hypothesis, ensure it is specific. The more specific the hypothesis, the easier it will be to test and measure your results.

Create multiple variations

Now that you’ve laid out the foundation, it’s time to start building out your test. First, you’ll need to create variations of the element(s) you’re testing. These variations can be very different or only slightly different versions.

For example, if you’re testing whether the positioning of an in-app notification can improve the conversion rate, you can create different variations of the page with the notification in varying positions

Each of your variations must be different enough to help you identify which one is performing the best.

Segment your audience

It is also important to segment your audience for a multivariate test. This means showing different versions of the product or webpage to different user groups.

For the example from the previous step, for instance, a segment of users may find the notification modal at the top, another at the bottom, and another in a corner panel or sidebar.

Segmenting your audience like this can help you to get more accurate results from your test.

Define trigger settings

Trigger settings determine when and where a user will be presented with a variation of your product. It can be a specific period of the day, after a specific interaction(s), or due to other factors like location, demographics, etc.

For example, to improve your user interface, you may create different versions and send one version to users within a geographical location and another to users in a different area.

Whatever your trigger settings are, ensure they are contextually relevant to each user.

Decide when to end the experiment and run the test

Once you’ve created your variations and defined your trigger settings, you now need to decide how long the experiment will last and when it will end.

The length of your test will depend on a variety of factors, from the number of users you need for your test to achieve statistical significance to your daily website traffic (or daily average users).

Once that’s done, it’s time to run your test.

Gain insights from test results

Finally, the testing period is over. It’s now time to analyze your results and identify the best-performing variation. There are a variety of statistical tests you can use at this stage, such as the chi-squared test or t-test.

Note that with multivariate testing, you can get three (3) types of results, including:

• Clear winner: A clear winning variation with statistically significant results. Once you’ve identified this variation, you can end your test and implement it in your product for all users.

• Not enough data: At the end of the pre-defined test period, the test results are inconclusive because there isn’t enough data to pick a winner.

• No significant results: At the end of the test period, neither segment has had a significant impact on the results.

Common use cases for multivariate testing

There are multiple use cases for multivariate testing in SaaS products. However, the three most common use cases are:

Test the experience with new feature announcements in-app

Run a multivariate test to test different variations of an in-app new feature announcement. This is very important as your announcement is critical to ensuring users adopt the feature.

Your test should include variations with different CTAs, media forms (GIFs, videos, or images), and copies to see which combination produces the best result.

Optimize email campaign performance by testing different variables

Use MVT testing to power a data-driven email marketing campaign. You can create email variations with varied:

• Subject lines
• Body Content
• Messaging
• Opening line
• CTAs, etc.

Your test should help you identify the best combination that improves the email open rate and conversion rate.

Test multiple elements for landing page conversion rate optimization

With multivariate testing, you can create different landing page variations by combining variations of every single component, including the following:

• Body copy
• CTA Button
• Headers
• Image placement
• Color Palette
• Scrolling depth

At the end of your testing period, your goal will be to confirm the combination of components that produces the highest conversion rate.

Conclusion

Multivariate testing focuses on testing multiple elements of your product or landing pages to identify the best-performing combination. Although it can be time-consuming, the results are very rewarding.

Thankfully, Userpilot’s latest iteration makes it easy to conduct multivariate tests in a few clicks – without any technical knowledge. Book a demo today to learn how you can conduct an MVT test with Userpilot.