To get the most out of this article, you need to be familiar with the following terms: A/B Test and Multivariate Test.
An A/B Test (also called a split test) is a test in which two versions of the page are tested against one another. The test involves the original page and a variation of the page. The variation can contain ANY changes. A/B tests are perfect for beginners, for quick tests and for websites with low traffic. You can get data very quickly in comparison to a multivariate test.
A Multivariate Test is a test in which page variables such as headlines, calls to action, paragraphs, buttons, etc are tested against each other. Google allows up to 1000 multivariable tests at once. For example, if you have six headlines and want to test three variations for each, you can set up: 6 x 6 x 6 = 216 tests. Multivariate tests require a long time to produce data, but are much more effective than A/B.
The Biggest Question of All – Does Google Website Optimizer Affect Search Engine Rankings?
If you do search engine optimization, then obviously this is the biggest question. Why? Because Google Website Optimizer (let’s call it GWO for short) serves different pages to different users (up to 1000 variations). How does the Google bot behave towards it? What if it follows a link and gets a different page all the time? Will I lose rankings? Those are the questions you should be asking.
Here’s a clear answer by Tom Leung, Business Product Manager of Google Website Optimizer team in an interview with conversion guru Brian Eisenberg:
Bryan Eisenberg: "Any time you do, any kind of multivariate, A/B testing, what will the Google search engine think of the page?"
Tom Leung: "Using the website optimizer in and of itself will not affect your Google search engine rankings, because the original content will still be there and be indexed as if you weren’t testing. However, if you use the test and if you find ways to give a better customer experience and then you implement those changes, then you should have every reason to believe that it will be reflected in a positive way in your rankings. Using the tool in and of itself won’t change your rankings one way or another."
The answer is loud and clear. I have not tried testing the index page, so if you did, please let us know your experience (even with that guarantee I am VERY wary of testing the index, since it’s obviously the most important page of the website).
Also, here is a quote from ThreadWatch, quoting Google:
"Website Optimizer is designed to keep your original content visible in the HTML source code of your page at all times. As a result, your original content is visible to crawlers, which means there should be no major impact on search engine ranking."
Apparently GWO and Google Search are integrated, and Google knows when you’re testing and won’t penalize you (an extremely WILD guess). On the other hand, once tests are successful and you implement changes to the original file, don’t be surprised if rankings change, either positively or negatively.
The set up is easy. First choose the type of test you want to run – A/B Split or Multivariate.
Setting up an A/B Test
First choose the page you want to test. Then, create an alternative version of that page. Name your experiment and identify the pages you want to test. Google will ask for the URL of the original and the URL of the test page (Google calls it Variation #). Copy and paste URLs into appropriate boxes.
Next, specify the conversion page. The conversion page is the page that users see when they complete your desired action (sale, sign up, download). The conversion page cannot be accessible from a link, search result or any other way, but by completing your action (otherwise it will mess up data).
At this point, Google will give you four script codes to install. These include one “control script” which will go on top of the source code (before the <HTML> tag) and one tracking script, which will be positioned on the bottom of the source code (after the </html> tag).
Another script, for the variation page, will go on the bottom of your test page after the </html> tag. And the fourth script must be placed at the very bottom of the conversion page. Add all the code and validate it.
Click continue and you’re done.
From now on, GWO will serve different versions of the page to different users. If you’re testing two pages (original + variation) then Google will split traffic 50/50. In the GWO control panel you can view detailed reports, pause and modify the test.
Setting up a multivariate test is trickier, and you will need basic HTML knowledge to pull it off. First, identify the page you want to test. It can be the index, product page, contact us or any other page. Then select the page sections you would like to test, such as headlines, buttons, calls to action, etc.
Name your experiment and specify two pages for GWO – the page you want to test and the conversion page.
When it comes to installing scripts, multivariate testing is more complicated. As with an A/B test, you will need to paste two scripts into the original page, one on the very top and another on the very bottom. You will also need to plug in a script at the very bottom of the conversion page.
Next, wrap each page variable you would like to test in GWO script. If you want to test a headline, then the code would look like this:
<script>utmx_section("specify ANY name here")</script><h2>Qualifying for Home Equity Loan or Home Equity Line Of Credit</h2></noscript>
If you want to test an image, or a button, then the code will look like this:
<script>utmx_section("specify ANY name here ")<img src="images/home-equity2.jpg"></script>
Wrap ALL the elements you want to test with this script. When done, press continue. That’s where the fun begins.
In my test I specified five headlines I want to test. Here’s how this looks inside Google Website Optimizer:
To add variations, simply click on Add New Variation. You can create as many variations as you desire. Keep in mind that all variations are tested against one another, so if you add three variations for five headlines (including originals), the total comes to 243 page variations. It will take some time to test all of them, and you will need a lot of traffic (at least a few thousand visitors per day).
Once you’re finished with the set up, you can check all the variations at once using the Google Website Optimizer test screen.
Next, specify the amount of traffic you would like to test > 100%, 75%, 50%, 25% or 10%.
Press “Launch” and you’re done. Now it’s time to sit and wait for results.
Both an A/B split test and a multivariate test require time to determine the most effective page/combination. Metrics include estimated conversion rate, chance of beating original, observed improvement, and conversion rate/visitors.
Immediately after Google Website Optimizer was released, SEOs started to ask a question: does Google warrant cloaking? Since it provides software that helps serve different versions of the page, GWO can be considered a cloaker from Google. Search Engine Land and other leading search publications had interesting posts with insightful discussions by industry gurus. Here’s a quote from Danny Sullivan (apparently he was quoting someone else, by the way):
"Test various pages, and keep showing spiders the original one during testing. Anyone not seeing what the spider sees is technically getting a cloaked page. Yes, as he notes, that’s what Google Website Optimizer does – so Google itself seems to be giving tacit approval to cloaking."
So does Google warrant cloaking? The answer is still NO, as Google invented the tool to help small advertisers get more out of their AdWords campaigns.
More Articles and Resources Concerning Google Website Optimizer:
I recommend the tool to anyone serious about increasing their conversion rate, since conversion optimization and testing go hand in hand, and Google is giving you an enterprise level tool for free.