News
Easy A-B Testing for Success
February 11, 2010
When building a website, you inevitably spend a large chunk of time deciding which layout is best and which content delivers results. But when it finally reaches the web, how do you know you made the right decisions? Could that sales bump have been 8% instead of 4 if the subscribe button were bigger? How do you know traffic wouldn't have doubled if you had done something slightly different on your home page?
You can find out by testing, and Google's Website Optimizer makes it simple. Go to google.com/websiteoptimizer to get started.
If you have an account with any other Google tool such as Analytics or Gmail you can use the same login otherwise create an account.
Once inside Google does a great job of taking you step by step through the process. Click "Create a new experiment" to get started then choose which type of test (experiment) you want to run: either a simple A/B 
test where you test two or more versions of a page or a Multivariate test which allows you to test content variations from multiple sections on the same page. If you don't know which one to choose, here are a couple of examples.
If you want to compare the conversions for two e-commerce page layouts where one has the "Buy now" button next to the product image and the other has the same button below the product description, you would choose the Google test called "A/B Experiment".
However if you have a subscription landing page with several subscribe links spread across the page and you want to know how each one performs, you would choose "Multivariate Experiment". There is one caveat. If the page you are testing receives fewer than 1,000 views a week, it would take so long to gather enough data on each link to produce a quality test that Google suggests you run the test as an "A/B Experiment" instead.
To continue you will need to decide which page or pages you are testing and identify the conversion page, the page where the site visitor ends up after taking the next step. It might be a shipping address page or a thank you page. If you need to test using multiple conversion pages at the same time because different test links on the pages point to different locations, don't worry. Google can handle that but will only report one conversion per visitor even if the same visitor clicks through to multiple conversion pages.
Once you have given Google the page details, you need to paste some code. Google will produce two code blocks: one block of code is for the experiment page and the other is for your conversion page(s). Think of them as start and finish lines. Google gives step by step instructions for adding the code blocks to your pages. It's as easy as using copy and paste.
Once your code is in place and your pages are published to the web you will need to give Google time to collect data before it can begin displaying test results. How long? That depends on how many people go to those pages and click through. Popular pages may produce good results within a couple of days. Others may take weeks before a clear winner can be declared. In the end, if it's too close to call you can make changes on your pages and try again.
Don't make the mistake of just testing one page. Regularly testing all of your site's most important pages keeps your website and business evolving and competitive. You can be sure your competitors are doing it.