Google Website Optimizer: How to Split-Test Your PPC Landing Pages for Maximum Conversion

Google Website Optimizer can be accessed in your Adwords Interface. Just click the Tools Menu and choose Website Optimizer from the list of tools!

Google Website Optimizer can be accessed in your Adwords Interface. Just click the Tools Menu and choose Website Optimizer from the list of tools!

Google provides a tool in their Adwords interface called the Google Website Optimizer. And it has a number of useful features that you won’t find in any other tracking software.

Split-testing pages works by taking two versions of your landing page that are exactly identical except for the one variable being tested. Be it a single word different in the headline, colour, or position of your optin box.

They then automatically forward clicks on your ad to one of the two versions of your page, alternating between them to fairly and accurately determine which results in the desired action of a sale or optin.

Whenever the clicker views a key action page, such as your “thank you” page after opting in, the download page for a piece of software you give out as a demo, or a thank you page after making a purchase, Google counts that as a successful conversion. 

A running record is kept of how many views each version of your test page received relative to the number of times your key “successful conversion” page was subsequently viewed. You can then look at two ratios you compare to decide which page results in the best rate of conversion.

 

Accurate Conversion Tracking

Tracking is rarely ever exactly accurate. The reason is that much of it is done with cookies to find out who the unique visitors or action takers are. Cookie tracking has its problems–for example, if a prospect uses a text browser or has their cookie security turned on high, you will not get any stats tracked for them.

All the same, most browsers store cookies and allow the tracking software to do their work.

Other third party tracking software does a good enough job. But they track all clicks–those you pay for and those you don’t. Google has procedures in place to cancel out forged clicks and other useless actions. Which makes the only clicks worth tracking those that Google produces. Since these are the only ones you actually pay for. That’s why I prefer to use Google’s website optimization tool.

 

Geek Talk: How to Define a Clear Winner

 

Google Website Optimizer Telling Me There is No Clear Winner Yet
Google Website Optimizer Telling Me There is No Clear Winner Yet

Statistics tells us that if you toss a good coin 100 times, it will land 50 times on heads and 50 times on tails. This is only theoretical, though, provided the coin is fairly weighted on both sides. In reality, you may observe 40 heads and 60 tails instead. But if you keep tossing the same coin, the you will get closer to the theoretical outcome of 50/50 heads/tails.

Few marketers split-test correctly because they act on too little data or otherwise they wait too long and waste opportunities to profit with their better landing page.

In the same way, if you toss the coin only 30 times, you are EXPECTED in theory to find it lands 15 times on heads and 15 times on tails. But in reality, the outcome could be extremely one-sided. And if you were to act on this data saying that the coin weighs more on one side than the other, you would probably be wrong.

Similarly, your two test-landing-pages may well show bias in favour of one versus the other even if they are the exact same page. Especially if measuring conversions with only data about a few clicks.

Running tests that update and track every click and conversion along with confidence statistics will help you determine if there is a large enough margin between the two test pages, relative to how many results were observed. Confidence statistics work by figuring out what the theoretical difference between your two landing page versions would be if it were due to the “randomness” phenomenom I described with the fair coin. If there is enough difference seen between the theoretical outcomes and the actual outcomes, you can safely conclude that the change being tested has an effect on how the page performs.

Rather than make these difficult calculations yourself the Google Website Optimizer will do it all for you and interpret the results on your behalf.

When Google says you have a winner, you can trust to kill the loser and keep the winner and repeat with another test variable.

 

I have used the Google Website Optimizer tool a lot in my business. I test the hell out of every landing page until I get a version that converts at the highest rate possible. This technique has been responsible for me seeing thousands upon thousands of Home Business leads at $2.35 each.

Start using the Google Website Optimizer today. It’s simple. Just create two versions of your best landing page and let the website optimizer tell you the winner!


To learn more of Jim’s PPC advertising techniques, check out PPC Domination.

Author: Jim Yaghi

Jim Yaghi is an advertising consultant and traffic expert, with a background in Artificial Intelligence.

4 thoughts on “Google Website Optimizer: How to Split-Test Your PPC Landing Pages for Maximum Conversion”

  1. Done that already Jim. I had issues with split testing via Google Website Optimizer. The conversion page I was using was an affiliate link which Google wouldn't let me include in the test as the conversion page had to be on my own domain. Is there a way around this where I can use an affiliate product as my conversion page or do I have to mock up a 'thank you' page on my own domain and potentially miss out on affiliate purchases during the 'testing' phase??

    Vince.

    1. Pretty simple man! Two ways to do this…

      I use a different method, but here's the quickest least technical method. Create a new thankyou.html page on your domain. Make that your "Thank You" URL with your autoresponder. Then with that page, create two horizontal frames. One should be 50 pixels high and the other should take up the rest of the page.

      Make the source of the top frame a page called "status.html" and in that place your action code from Google and write a message like "Thank You, etc, etc". Then in the lower frame, make its source the affiliate product's sales page.

      To see an example (and copy code), just fill out and submit the optin form below then view the source of the resulting page. That's the method I use in this blog :-)

      Jim

  2. Done that already Jim. I had issues with split testing via Google Website Optimizer. The conversion page I was using was an affiliate link which Google wouldn't let me include in the test as the conversion page had to be on my own domain. Is there a way around this where I can use an affiliate product as my conversion page or do I have to mock up a 'thank you' page on my own domain and potentially miss out on affiliate purchases during the 'testing' phase??

    Vince.

    1. Pretty simple man! Two ways to do this…

      I use a different method, but here's the quickest least technical method. Create a new thankyou.html page on your domain. Make that your “Thank You” URL with your autoresponder. Then with that page, create two horizontal frames. One should be 50 pixels high and the other should take up the rest of the page.

      Make the source of the top frame a page called “status.html” and in that place your action code from Google and write a message like “Thank You, etc, etc”. Then in the lower frame, make its source the affiliate product's sales page.

      To see an example (and copy code), just fill out and submit the optin form below then view the source of the resulting page. That's the method I use in this blog :-)

      Jim

  3. That's pretty darn smart, Jim. This way you can actually get stats off of any affiliate campaign.

    And that would work for tracking conversions too, right? genious!

  4. That's pretty darn smart, Jim. This way you can actually get stats off of any affiliate campaign.

    And that would work for tracking conversions too, right? genious!

Leave a Reply

Your email address will not be published. Required fields are marked *