Beyond A/B Testing: Multivariate Testing with Marketing Automation Platform “Marketo”

Best practices are courses of actions specific to your industry, your company, your branding, your specific landing page, your content, and your request. Multivariate testing is how you determine your specific best practices for you alone.

Many professionals these days seek out “best practices” in every possible situation. “What are the best practices for creating landing pages? What are the best practices for increasing conversions? What are the best practices for a call-to-action button color?”

Universal “best practices” really mean universal “good practices.”

Why should you do multivariate testing for landing pages?

    • The results of multivariate regressions are a teaching tool for future design decisions. What call to action button performs the best across all campaigns, landing pages, and traffic sources? Instead of having a sample size of 1 (the problem with A/B testing on individual pages, since products and copy vary so dramatically), we can test the effect of changing a submit button’s color across all pages. Do our customers for our specific brand in our specific industry convert better with red buttons or green buttons, all else equal?
    • A/B testing has a small sample size. The results are more qualitative than quantitative. You can generally see that one version performs better than another, but you don’t know if that same change will have an effect with a campaign promoting a different resource and with different landing page copy.

  • You can determine “best practices” are for your specific company, which are different from many factors. Universal best practices are a myth. Your customers might respond better to some things than others.
  • Multivariate regression results show which investments are performing better and worse. With this data, you can make controlled experiments. Maybe more MQLs (marketing qualified leads) come from one PPC campaign than another. You can use this information to see which campaigns need to improve, or which campaigns need to be cut.

 

How do I get the data for that kind of multivariate testing within Marketo?

Wow, I’m glad you asked.

Using Marketo text tokens to call different CSS styles and stylesheets is how we gather the data for multivariate landing page testing. We code some extra tokens into our responsive landing page template. If we’re looking to see the effect of a different call-to-action button color on our responsive Marketo landing pages, through many separate campaigns, we can create a top-level text token with the “B” version CSS. Whether this is one line:

<style>
#mktFrmSubmit{color:red!important}
</style>

or an entire external stylesheet (<link href=”http://etumos.com/mkto/universalTestB.css”>), we can automate testing throughout all campaigns.
Marketo Multivariate Testing Example
Because Marketo tokens work in landing page templates, we planned ahead and placed extra tokens in the HTML template. These are the following:

{{my.universalTestAStyling}}

{{my.universalTestBStyling}}

{{my.universalTestCStyling}}

When these tokens are undefined, no HTML is rendered. That means clean HTML code with no unnecessary parts. When we want to do testing, we can define top-level Marketo text tokens to start testing.

With this data, we can run multivariate regressions. Throw in some control variables for campaign source (e.g., PPC campaign), resource promoted (e.g., Buyer Personas), and landing page content, and we have some controlled marketing automation tests.

I post interesting little Marketo hacks and workarounds on Twitter all the time. Let’s network.



Pin It