Multivariate Testing is a method of experimenting with variations of particular elements in a feature implementation, such as the headline, images, copy, etc in a landing page or application launch screen, or other critical moments of truth in a customer journey, in order to determine which variations of said elements are best suited to improve conversions.
The method is similar to A/B testing, however, multivariate testing experiments are run with a higher number of variables and generally provide deeper insight on how to optimize your page. In multivariate testing, your feature implementation becomes a combination of elements which can be decomposed and tested simultaneously.
To help break down this process, let’s assume you’re working with the following elements: header, page images, and copy.
If you were doing a multivariate test of these elements, you’d create variations of them:
| Header Variations | Image Variations | Copy Variations |
|---|---|---|
| Header 1 | Image 1 | Copy 1 |
| Header 2 | Image 2 | Copy 2 |
| Header 3 | Image 3 | Copy 3 |
The purpose of a multivariate test is to try out different versions of these variations, as illustrated below:
| Test 1 | Test 2 | Test 3 | Test 4 | Test 5 |
|---|---|---|---|---|
| Header 1 | Header 1 | Header 1 | Header 2 | Header 2 |
| Image 1 | Image 2 | Image 3 | Image 1 | Image 2 |
| Copy 1 | Copy 1 | Copy 1 | Copy 1 | Copy 1 |
As you can see, the number of possible variations can stack up quickly, and this is using only three elements. The complexity of a multivariate test can grow exponentially, making it difficult for your team to manage. Software solutions allow you to run multivariate tests more efficiently as they can experiment with a multitude of possible combinations.
As the multivariate test gathers data over time, you’ll be able to separate the wheat from the chaff and discern which combination of the variations performed the best. For instance, maybe Header 2 + Image 1 + Copy 3 gets the most conversions, making it the winning combination that you decide to run with from then on.
How To Create a Multivariate Test
When creating a multivariate test, it is best not to include too many elements since every element you include more or less doubles the number of combinations you’ll have to experiment with.
Not to mention that all elements aren’t created equal. For instance, if your test includes headers, call to action buttons, and footers, you may discover that footer variations make little impact on conversions.
Some good steps to follow when creating a multivariate test (using a landing page example):
- Use your analytics data to do an evaluation of the page and identify what is and isn’t working with it.
- Once you know which elements are hampering performance, order them based on the amount of damage they’re dealing to the page’s quality.
- Formulate a hypothesis regarding the elements you want to test. Ask questions like: If I fix these issues, what impact will it have on the page’s conversion rate? What about the page’s overall performance?
- Launch the test, and as it is going, document it. Doing this formalizes the process and makes it easier for others to provide feedback on it later.
- Once the test is complete, analyze the results. Pay attention to what did or didn’t work and conclude whether your hypothesis was correct. You can use the data generated by the test to make appropriate changes to your web-page/app, or you can use it to create follow up tests.
