How to Split-Test with Google Website Optimizer

March 31st, 2009 | 11 comments

Split Testing How many times have you heard “The only way to know for sure is to test… test, test, test” right? Probably so many times you hate hearing it anymore. Unfortunately though it’s very true! Not only that, but testing can increase conversions… in some cases making the difference between $50/day and $100/day or more! You may be surprised how much of a difference one or two small “tweaks” will do to conversions. The only way to know anything for sure is to test it!

BUT HOW?

When it comes to sales copy, design, layout… basically anything that appears on a Webpage, the answer is to split test. But once again, the question is HOW do you do it properly?

In the past there have been specialty scripts and software that did it fairly accurately, but not long ago Google created something that has made split testing very simple and very accurate. It’s called “Website Optimizer.”

Split Testing with Google’s Website Optimizer

1. If you haven’t already, go through the Google Website Optimizer sign up process.

2. Click “Create an Experiment”

3. You now have 2 options. You can do a basic “A/B” test or you can do a Multivariate test. With the first option you’re just testing changes you’ve made on one page versus the existing page. For this I recommend only making one change, as you won’t know how each effected your stats with multiple changes. The “A/B” test is great for a simple split test of your headline etc. With the second option you can test multiple changes on a single page. I like to make things simple and test one thing at a time, so for this example we’ll be doing a simple “A/B” test (option 1.)

4. After clicking “A/B Experiment – The simplest way to start testing fast” I’m presented with a checklist. I need the original page I want to test, in this case I’ll be testing the headline of my Article Marketing Guide. I then create a new page by copying all of the html from the index and pasting into a new page and only changing my headline, since that’s what will be tested. It then asks for a conversion page, which in this case will be the thank you page. I have all of that ready, so I click “Create.”

5. Now you have 3 more steps. First name your experiment. Just name it whatever you want. I’ll name mine “AMD Headline Test.” Second insert the URL of the original in the “Original page URL” (in this case it is http://www.articlemarketingdomination.com/) Now enter the new page you created (in this case it is http://www.articlemarketingdomination.com/index2.php) Third enter the URL of the conversion page (in this case it’s my thank you page) click continue.

6. Select “You will install and validate the JavaScript tags”

7. Follow the instructions… very “cut and dry” You’re just copying and pasting code into your 3 pages. Click “validate pages” and if they’re good click “Continue.”

8. Ensure everything is correct and click “Start Experiment”

All Done! Wait a few hours and you’ll start to see stats

img credit: mil8

Related Posts

11 comments

John (3 comments.)
31st March, 2009 at 4:33 pm

Josh, thanks for that tip. I have heard that with some CPA marketing it is better to have a very simple page with no graphics. This subconsciously indicates to the viewer that you are not selling anything. Now, I have a way to easily test that for myself. No-graphic version coming right up !!
Josh Spaulding
31st March, 2009 at 4:39 pm

Absolutely… there are many possibilities with it. I’d be interested in your findings as well. I’m a graphics freak, so you know what I think will convert better.
Courtney (19 comments.)
31st March, 2009 at 4:42 pm

Terrific! this was a question I was going to ask in your forum very soon to split test the sales page of my ebook. It’s great to know that Google offers a free and easy to use version. Thanks for the instructions.
Steve (4 comments.)
31st March, 2009 at 6:12 pm

Thanks Josh for another great (and Free) tool. I’ve been wanting to do some testing on a new product page, but have not got around to evaluate software, etc. You saved me a bunch of time! Thanks again.
Milan Kosanovic (2 comments.)
1st April, 2009 at 9:51 am

Thanks Josh. Testing is a subject that’s definitely worth investing time in.

Here is one tip:

Multivariate test -
Cons: you need much more traffic than with A/B test to get some statistically valid data with Multivariate test.
Pros: Testing different more than one change is useful, but it is not the true power of Multivariate test. The true power of it is testing different factors combined. For example, you might discover that the certain headline with the certain opening paragraph are the best converting combination, while other headlines and paragraphs when tested separately give the same or better results than the one that wins (when combined).
Milan Kosanovic (2 comments.)
1st April, 2009 at 10:01 am

Josh, may we suggest headlines for you to split test?

I’ve seen your: “Learn from one of the most Elite Article Marketers and Start Dominating your Niches with Effective Article Marketing!”

and thought to make my version:
“Discover the Technique an Elite Article Marketer uses. Dominating Niches with Effective Article Marketing Guaranteed!”

“learn’ might suggest school/boring etc. while “discover” has only great connotations…
WordsmithBob (1 comments.)
4th April, 2009 at 3:56 pm

I’d really like to see an example of someone using split-testing. My tendency is to be impatient and have two totally different pages but this really seems to go against the idea of split testing.

You might want to make sure that people understand that split testing should be for small or inceremntal changes.

For instance, two identical pages with different headlines and everything else the same. Or if you’re going to use graphics, thry not to alter the text at all so you know that it really was the graphics that made the difference.

While this kind of testing can seem really boring and tedious, it’s how science is done and that’s a good precedent to follow.

Also, make sure you run your test over a statistically significan’t period. When we do organic search engine optimization for a client, they start emailing us within a couple days wanting to know why they aren’t on the first page of Google.

They impluse is to start altering thing immediately to try to improve results. Yet, we tell the client that we need to give the SEO at least a month before we do any tinkering to get significan’t data to analyze.

You need to do the same thing with your split testing. This must be a long-term process.
Lee (1 comments.)
6th April, 2009 at 7:01 pm

Wow, I have never heard of this tool until now. I wish I had come acroos this a while back. I have been trying to do this whole process manually up until now.

Thanks again!
Automatic Web Site - Jon Leger Case Study | MasonWorld Internet Marketing Strategy | Internet Based Marketing Blog
8th April, 2009 at 2:34 pm

[...] test different elements on the page.Â I had that post about half written, and then I saw that Spaulding had written a post on Google Website Optimizer.Â Â So, since that was exactly what I was going to do, I wrote Josh a nasty email and then decided [...]
John Taylor (1 comments.)
23rd April, 2009 at 11:42 pm

Interesting blog post. What would you say was the most important marketing factor?
Kamlesh Desai (1 comments.)
22nd May, 2009 at 1:00 am

Very interesting and very well explained. Google has all the tools –thabks for letting us know.

This is a very useful time, effort and money saver tip.