I talk a whole lot about split testing in every walks of optimization whether that’s on your internet site, off your internet site, in your design, when email marketing, etc. I figured so it was time I finally addressed what’s split testing and moreover just how to split test.
First thing’s first, what’s split testing? Split testing on its most basic level describes taking multiple versions of something and alternating the two to see which performs better.
Webmasters use split testing a whole lot because they always want the highest possible conversion on their goals whether that goal is to market a product, get someone to sign up because of their email list thus generating a lead, as well as just keeping their traffic on their site for longer.
Common subjects of split testing are the copy or design/layout that you use on your own site.
You may get as macro in split testing as changing the entire layout of your internet site or as micro as changing one word in your call to action.
Given that we’ve covered the what, let’s cover the how in just how to split test.
Split testing is really as simple as taking two or more versions of whatever it is that you wish to test and interchanging them with one another with the idea being of tracking analytics while doing so.
Like, when you have a sales page for your product you may test everything from the header graphic, including and excluding testimonials, the placement of the testimonials, your “buy now” button (call to action, color, size, shape, placement, etc.)
With regards to tracking, typically you’ll wish to see which version of what it is you’re testing converts better towards your preconceived goals.
If it’s a sales page, likely every change which you’re making splittesting.com on that page is to encourage visitors to click right through to the purchase page. In this case you are able to track your results simply using Google Analytics and tracking how many views you get to each page.
Any changes that you make while split testing come in an effort to have the two numbers as close together as you are able to as this suggests that everyone who visits the sales page ultimately clicks right through to your purchase page.
There typically is never an “end” as it pertains to testing; you must continue to accomplish it as you always desire to be improving your conversion rate. You can also go ahead and test the copy on your own purchase page when you have control over that page, as well.
As you’re probably gathering, as it pertains to this type of testing, being anal could be the name of the game.
With email marketing, split testing is really a major part of the process and many of the better email marketing companies make split testing as simple as possible. I use AWeber, for instance, and they’ve an alternative to check everything you are able to imagine.
Your web form, for instance, or the form on your internet site which people use to sign up for your email list obviously plays a position in how many visitors to your internet site go ahead and sign up for your list. You can cause as many versions of your online form as you want, varying it in terms of text and shape, color, etc., then choose how often you would like each of the web forms to look on your own site.
In this manner you’ve multiple versions of the exact same form appearing randomly and interchangeably on your internet site and never having to swap them out yourself, and AWeber tracks the sign up rates for every one.
Then, following a period of time, you are able to check back in to see which performed the very best, then take that version and produce a few copies (also quite simple to accomplish in AWeber) of this web form which you can tweak to split test against your original one, starting the process anew.
With regards to precisely how long to offer before choosing the winner during each split test session, I don’t recommend a particular span of time such as a week or perhaps a month so much as I suggest that you allow the full time so a significant level of traffic can visit your site.
This makes it so that you may get an even more realistic idea which version performed the very best so that you can discount randomness or anomalies which are more prevalent with smaller levels of traffic.