How To Test Websites, Part 1 - Split Testing

In a previous post, I showed why testing your website is important.

Now, let's look into how you can test your website.

There are basically two types of testing you can do on your website: split testing or multi-variate. In this part, we will cover split testing. In part 2, we will look at multi-variate testing.

Most internet marketers probably use split testing.

With split testing, you change one part of your web page (for example, the headline) while keeping the rest of the page constant. You then send equal traffic randomly to each alternative page, and see which headline results in the most conversions.

The advantage of split testing is easy to implement. Just create N number of alternative web pages that are identical except for the part to be tested. Then, on your actual page, use javascript and the rand() function to randomly forward 1/N users to each alternative page.

For example, if you want to test 4 headlines at a time for www.mydomain.com/salespage.html, create pages salespage1.html, salespage2.html, etc. Then have salespage.html send 1/4 of your traffic to each alternative page. I have an example of using JavaScript to randomly forward a web page on my Unix blog.

The disadvantage to split testing is that you can only test one variable at a time. It might take many months of testing / tweaking parts of your sales page before your page starts to reach its optimal conversion rate. This will result in a lot of missed sales.

In part 2, we will explore an alternative testing approach that optimizes sales pages must faster.

In part 3, we will examine MuVar, one of the most powerful website testing software packages on the market.