In my previous article, I looked at defining your success goals and what to measure when running multivariate tests. Let's now look at your site factors and learn how to select the right ones to test.

By now, your marketing goals are clearly defined and you're ready to run a multivariate test to optimize your site's marketing effectiveness... but which elements, or factors, should you test?

What Is a Site Factor?

First, let's define what a factor is. A site factor is a distinct, single element that you can control for testing purposes, and each version of a factor is called a variation. Through testing, you'll discover that some factors are influential and have a causal relationship with user behavior—modifying them changes how users behave, ideally motivating them to reach your marketing goalposts (e.g., checkout, registration, or simply spending more time on your site.) Others factors are not influential, and have little to no impact on what users do.

One of the keys to multivariate testing is in selecting factors to test, and defining them so that they may be accurately analyzed.

Many unique factors exist on every site, and they vary in scope from narrow to broad. For example, a factor could be an element on a single page, such as a headline, image, or paragraph of copy. Or, it might span pages or exist across a particular section, such as a navigation component, product category promotion, or a call-to-action. As well, a site factor could span your entire site, such as a text style (CSS), or grid layout spacing (e.g., how much space is allocated for main content vs. advertising vs. widget areas, etc.).

Choosing Site Factors

If multivariate testing gives us the ability to simply throw everything at the wall to see what sticks, what's the big deal with choosing specific factors to focus on? Why not just test everything and anything? While identifying factors isn't that hard, coming up with a broad range of variations takes time and effort. For example, think of the creative resources required for coming up with five variations for each of 20 different site factors, which could include everything from the page title to the copyright in the footer. While it's technically possible to test it all factors and their variations at once, it's much easier to be organized and test fewer factors that have a higher likelihood of actually making a difference.

By thinking like an end user and navigating through your site, you can uncover the main factors that are directly in the paths through which your users make their way toward reaching your marketing goals. What are the words that you have to read, the images that you must view, and the links that you're supposed to click on?

These touch points are your low-hanging fruit and are often the factors that, when optimized, will give you the biggest ROI because they are encountered in every meaningful task or process. A few more specific examples include registration forms, product presentation and layout, checkout funnels, and high visibility slots used for offers, cross-sells, or ad space.

Have a Clear Hypothesis

Referring back to my previous article, whenever you are running a test, make sure you have a clear, defined hypothesis. This is nothing more complicated than filling in the blanks in the following statement: "What would happen to ___ if we did ___?" For example, what would happen to newsletter opt-in rates if we changed the privacy headline? This simple yet important step ensures that you know what you are testing, and how you'll determine whether any changes increased performance.

Properly Isolating Your Site Factors

When defining factors on your site for testing, be careful of accidently creating factors that actually consist of two or more factors. For example, most headlines are at least four factors: words, font, color, and size. This means that when defining a factor and producing variations of it for a test, you should change only one thing per variation.

To elaborate, let's work through an example of what not to do, and see what happens if we were to test these two alternate headlines: "buy one get one free" in red, and "two for the price of one" in blue.