
The Basics of Split Testing
02-Feb-2012
Wayne Sharer
It’s time to start again with the testing and tracking series. It’s been some time, however, if you’re serious about having a winning business, you need to understand theses basics because it is how you profit AND increase your profits.
Let’s get started…
Split testing is the process of testing various different versions of the same website to see which one performs the best.
The Importance of Split Testing
If you’re not split testing, you’re leaving a lot of money on the table.
When it comes to how well your page converts, small improvements make a very big difference. The difference between a 1% conversion rate and a 1.2% conversion rate isn’t just 0.2% – It’s a 20% improvement.
If that’s your first product, then chances are that 0.2% improvement will have a very drastic impact on the rest of your business. It may improve your whole company’s profits by 20% or more!
Furthermore, this is an increase in bottom line that you don’t have to pay for. Instead of spending more money on advertising, you’re spending less per customer because you convert more of your visitors into buyers.
Most business owners simply never split test. They’re leaving tons and tons of money on the table.
Imagine increasing your first product sales page by just 0.2%. Then increasing your second product sales page by 0.2%. Perhaps you can increase your advertising by 0.2% and even your email marketing by 0.2%.
How quickly do you think this will add up?
There’s almost no end to how far you can take split testing. The best marketers in the world have split tested their websites to the point where they can get conversion rates that are as high as 12%!
Meaning out more than 1 out of every 10 visitors will buy something when they visit the website.
Data Contamination
When split testing, it’s crucial that the data you gather comes in an “alternating” fashion.
A lot of business owners ask by they can’t just try one headline for a couple weeks then switch to a different headline and see which one sells more.
The problem with this approach is that the data will often get contaminated. There are many factors that could affect results over the course of a week or even a day.
Perhaps sales are slower on Tuesday. Perhaps a major event lured people to the TV rather than the computer. Maybe it was payday on one of those weeks which caused a surge in sales.
There are so many factors that come into play; it’s very difficult to tell what really affected the sales data.
The best way to track the data then is to alternate the page visitors see after each visitor. In other words, visitor 1 will see site A, visitor 2 will see site B and so on.
There are two different types of split testing: A/B split testing and Multi-Variable Split Testing, which can then be broken down into Full Factorial and Fractional Factorial split testing.
A/B Split Testing
Split testing was first invented before the internet ever came into existence. Magazine and newspaper advertisers wanted to run some tests to see which ads converted better.
They invented what was called “split run” testing. Newspaper printing works based on a “roller” of sorts. They put two different ads on two sides of the roller, so that each ad that was printed would alternate between advertisements.
This is similar to how A/B split testing works.
A/B split testing is the process of splitting your traffic between two web pages. Each visitor alternates between seeing Page 1 or Page 2.
A “cookie” is placed on your user’s computer to track which web page they saw. If they come back to the web page afterwards, they’ll automatically be directed to that web page. In other words, once a visitor has seen either of the two pages, that’s the page they’ll see each time they come back.
This is the most basic model of split testing. Test two different versions of a web page; see what works better and then moving on to the next version.
International speaker, author, and entrepreneur. Retired navy officer, former commanding officer. Over 35 years of leading, coaching, mentoring, and speaking.
Read more