The Basics
A good split test is made up of several elements:
1. One element to change, such as the headline on your sales page.
2. Two different versions of the page, A and B.
3. Traffic to these pages; drive traffic through social media, or paid ads for an even faster set of results.
4. Track the results of A and B until you have a clear "winner", that is, a page on which more people click, subscribe or buy.
5. When deciding on a winner, make sure the number of visitors to each page is on a par to determine if the results are statistically significant. For example, if A has 100 visitors and 4 sales, and B has 200 visitors and 8 sales, there is no winner. They are tied. In this case, either continue to run the test, or pick your favorite, such as B, and run a new variation against it, which we can call C.
If on the other hand A and B both have 4 sales, but B has twice the traffic, A is the winner because it has a 4% conversion rate versus a 2% conversion rate. Use A as your "control" and then try to beat it with version C. If A is still the best, it remains the control. If C emerges as the winner, C will become the control. Then you can test it against D.
When you are happy that your headline is performing well, test other aspects of the sales letter until you've got the best conversions you can manage.
6. Run the test for as long as it takes until you get a clear winner. Use that as your "control" page.
7. Go back to Step 1 and run another test using a headline (technically headline C).
Note: If neither A nor B are a clear winner, choose the version you like best, and run a second test against that version until you get a winner.
Setting up the tests is pretty easy. You have to stay organized. Tracking the data can be a bit tougher, because you will have to use one or more sets of tools and/or analytics to get a full picture of what has happened on the test.
For example, you can use Google Analytics and Experiments to create the split test, but you will have to also use your sales data or email marketing subscriber information to determine how many people completed the call to action you were requesting.
Things to Watch Out For
Low Traffic Sites
If your site does not get a lot of traffic, it will take longer to run the test and get statistically significant numbers. Or, you could run an ad to drive traffic to A and B to speed up your test and get a winner faster.
Test for 7 Days Minimum
Test for at least a full week in order to account for traffic and habit variations from one day to the next. Chances are the results you get on Sunday might be very different from Monday or Thursday.
Be Clear about Your Goal
Set up your test to have one variable and one goal. For example, test the headline only, and see what impact it has on sales. Test the subject line of the email only, and see what impact it has on open rates.
Don't Be Dazzled by Other Numbers
It is easy to get distracted when you are using a hugely robust program like Google Analytics. Focus on your goal only in order to get the data you need to take action; that is, decide whether A or B is the control and then test against C, find the new control, if any, test against D, and so on. You will soon get a winner, and more conversions.
