The ability to split test is a powerful tool and thanks to Google itâ€™s completely free. Using Google Website optimizer may sound complicated but in truth it couldnâ€™t be easier to set up your first split test.
There are 2 flavors of split testing available through Google Website optimizer. A run of the mill A/B test and the more exciting multivariate test. An A/B test is when you want to test one page against another and a multivariate test is when you want to test different page elements against each other.
If you like this article, you might want to check some of our previous articles on The Benefits of QR Codes, A/B Testing, Tricks to Increase the Speed of your Blog, HTML5 Applications, and Creating Depth in Web Design.
Before you take your first steps when testing here is a check list of what you need.
A Conversion Page:
If you know anything about Analytics you will know what a conversion page. Itâ€™s the page a visitor will get to once they have completed a goal, if someone sends an enquirer through your site they will most likely be directed to your thank you page, this page should then be your conversion or goal completed page. When testing you will need to designate these pages and add the conversion code to them.
When we test for the first time most of us make the mistake of not setting a clear set of goals. Its important that before you start you must ask yourself what do you want to achieve? Most likely the answer will be to increase conversions, but you may also want to lower the bounce rate of your site, or you may want your visitors to stay on a page long enough to fully read that page.
A conversion is the most tested element on any site, but there are a host of other elements that could be tested, that when combined can help to increase conversions. I guess what I am trying to say is that looking at the bigger picture may not be the best solution. If you break the process of a conversion down you may be provided with more useful elements to test.
Now that you have a clear set of goals in mind the next question you need to ask is how you will achieve them. One of the biggest mistakes I have made in the past is that I did not have a check list of what page elements I wanted to test and in which order. I tested everything, but the way I tested was incoherent and erratic, which resulted in minimal improvements, if any.
Since those first failed tests this is what I have learned, start with the page you want to test, brake that down into a list of the different elements on that page, from my experience with ecommerce sites there are usually around 6 different page elements
Pricing Call to Action Header
Related Products Images Copy
Now that you have a list of the different elements to test, build a list of what you plan to test with, starting with the elements that will have the best chance of improving your conversions. I always test in this order, Price, Call to Action, Header, images and related products.
As An Example Of How I Would Test Price,
- Offer free shipping
- Lower the prices but charge shipping
- Show the prices less Vat
- Show the prices including Vat
- Offer a free Gift
- Offer a free gift over a certain threshold
- Change the color of the pricing
- Change the font
- Change the position of the prices
- Remove the pricing
- Show the RRP, your Price and the saving
- Show the RRP and “call for a discount”
- Show the savings if they buy more than one
There is 13 different ways of testing the price, there are probably another 20 variations you could test if you wanted to. This is just one element of 6 that we need to test, if we were to test 13 different variations for the 6 different elements thatâ€™s 78 variations we need to test. Because there are so many different variations, always start with the ones that will have the best chance of improving conversions and remember there are no wrong answers, if a test fails chalk it off and try another.
So now that we have 78 different elements to test, we also need to test them in conjunction with each other, this is where the power of the multivariate test comes into play as it will allow you to test different combinations together. Before I start to test a combination of elements I will test each page element on its own, once I have a winner, I will add the winner to my site and test the next page element on the list. The golden rule here is to never test to many elements together, doing so will mean you could wait months before Google decides on a winner, find a winner in each element before you move to the next.
This is the best advice I can give, if your first test does not work, try another and keep testing for at least 6 months, small gradual improvements over a long period is the real secret to split testing your site successfully.