Split testing is a method of finding out which of two items is more popular, converts better, etc. It’s based on the profound wisdom that “you don’t know what your customer likes best… they do.” For the purpose of this article, I’ll talk about a simple text advertisement on a webpage. But know that you can use this same concept for graphics, products, colors… in fact, for just about anything that contains two choices that you want to determine which is the most popular or effective.
So suppose you’re creating a short webpage, and you want to improve your conversion click-thru’s… the number of people who see your ad and are interested enough to find out what’s on the other side by clicking on it. One thing you’d want to try is two different headlines, to try to get more people clicking than before. What you’d do is write your two headlines, and create two copies of the webpage you’re using it on. Version 1 has Headline 1 on it; Version 2 has Headline 2 on it. And here’s the important point:
NOTHING ELSE IS CHANGED on the two pages. They’re identical except for the different headline. Now, using whatever traffic method you choose, send at least 100 people to your site. But do it like this…** send 50 people to Version 1 page, and the other 50 to Version 2. **
After you see how many people clicked through to your next desired action on each separate page, you’ll know which one was more effectiveattractive to the viewers. So if 10 out of 50 people clicked thru on Version 1, and 3 out of 50 clicked thru on Version 2, then Version 1 was more than 3 times as effective in converting viewers. And since the only difference between the pages was the headline, you know that Headline 1 was the winner. Now you might decide to try testing something else, maybe page color for example. So do the same thing… put up two different pages with identical text, graphics, headlines, etc. and change ONLY THE BACKGROUND COLOUR between the pages.
Split your traffic evenly and randomly between them, and very soon you’ll know which the most popular color was. And on and on! Split-testing gives you a scientifically accurate way of determining what works best. You have to provide a statistically valid sample before making a decision (that’s why I suggest at least 100 viewers).
Note: It’s best to run the test elements at the same time, to the same audience, under the same conditions, etc. That way, your test ‘sample’ contains only the single variable you’re trying to test. No point in testing a page in October, then trying a different headline for a test in April… too much has happened between those times, you’d have no idea what mood the viewers were in.