What is Multivariate Testing

So you've heard the buzz. Multivariate Testing is the word... But if you're like a lot of people, you're probably wondering what it's all about. Today is your lucky day- because I have a (relatively) complete explanation for you...

Before you will understand Multivariate testing, lets review some facts about statistics and testing. In science class we learned something called the scientific method. This principal states that to test a something, we should only change one thing at a time. By only varying one element and measuring results before and after the change, we could learn the relative impact of that change. When attempting to optimize a web page this way, we would make just one change to the page at a time. After repeating the experiment enough times, we could prove our results with some certainty.

Multivariate testing refutes the above notions. Using mathematical formulas and specialized arrays, it is possible to test many aspects of a system at the same time. Not only are you able to learn what worked and what didn't, but also the optimal combination of page elements to maximize the response rate. This creates the net effect of doing many simple experiments at the same time.

In fact, it becomes possible to run the equivalent of thousands of tests simultaneously.

How will it help me sell more on my site?

Through the magic of Multivariate experiments, the possibilities are virtually limitless. You can rapidly learn the best combination of page elements- things like your headlines, product images, and price points, to maximize sales or lead flow. Stop guessing what the ideal design for your page is because you don't have to. Through Multivariate experiments, you can learn if the picture of the girl with the white background, or the man with the blue background works better. Find out if the long form with detailed instructions works better than the short form with short instructions.

Not only will you learn which works better, but also, which combination of elements works best. In the above example, you might find out that a long form with short instructions actually works best. You never even included this in the original test. Or you might find the best combination in the first example is a man with a white background, another combination that you never considered.