Multivariate Experiments with Google Website Optimizer

Uploaded by Google on 16.08.2008

>> Multivariate testing is a way to test multiple variables on your web page simultaneously.
For example, using multivariate testing, you could identify the headline, image and promo
text on a page as distinct page sections, and then create three different variations
for each one. When the experiment was running on your page,
a user might see headline A, image B and promo text C all together, or headline B, image
C and promo text A. Multivariate testing works well if you have
specific sections on your page that you'd like to optimize.
However, if you have low page traffic, you may prefer to use Website Optimizer's AB testing
feature in order to obtain more reliable results. The first thing you'll need to do is name
your experiment. Pick a descriptive name that will help you
differentiate this experiment from others. For example, "My Landing Page Multivariate
Test." Your visitors won't see this name.
Next, type in the URL of your test page. This is the page being optimized.
By showing several variations of your test page, Website Optimizer will determine which
version is most effective at convincing visitors to take the desired action.
Now, type in the URL of your conversion page, such as your "Thank You" page or "Confirmation"
page. When a viewer visits this page after visiting
the test page, Website Optimizer will register a conversion is taking place.
Remember, your conversion page doesn't have to be an actual page.
It can instead be an action, such as a the click of the button or link or submission
of a form. For more information on setting your conversion
goal to be an action, click the "Advanced testing strategies" link from the set of links
on this page. Click "Continue" to move on to the next step.
You'll need to add Website Optimizer JavaScript tags to your page's source code.
When you add these tags to your site, no visible changes will be made until your experiment
begins running. If you don't have access to your source code,
or would rather your I.T. team perform this step, select the first radial button.
We'll provide you with a link to installation instructions that you can send to your webmaster
or your I.T. team. If you'd like to add the Website Optimizer
tags yourself, then choose the second radial button.
On the next page, we'll tell you exactly where you'll need to copy and paste our tags in
your source code. Don't worry if you aren't a genius at HTML.
Remember, this is largely a copy and paste activity.
So as long as you're comfortable with basic HTML editing and have access to your web pages
source code, you'll be fine. The instructions on this page tell you exactly
where each Website Optimizer tag needs to be placed in your source code.
Copy and paste the control script provided into the beginning of your test page's source
code, immediately after the opening HTML tag. Next, copy and paste the tracking script directly
before your test page's closing "body" tag. Make sure to use only the scripts provided
in your account, since they contain a specific account number that's unique to your experiment.
For each element of your page that you'll be testing, such as your image or headline,
you'll need to surround that particular element's source code with our section scripts.
Almost anything can be included in a section, including blocks of text, buttons, images,
script, etc. However, each section must span a contiguous
block of content. There are a couple of limitations on the placement
of section scripts that you should keep in mind.
Section scripts cannot be placed inside an HTML tag.
Nor can they be placed inside another block of script.
Section scripts can, however, surround HTML tags or other script blocks, provided the
script block does not contain any "noscript" or "/noscript" tags.
For additional questions on section tagging, check out our multivariate installation guide
from the set of links on this page. When adding section scripts to your page,
you'll need to think of a name for each section your testing, such as "Headline" or "Product
image," that will appear in the parentheses of this code.
Make sure to have a different section name for each section.
This name will not be shown to users, but will be reflected in your reports.
You may also find it useful to view a sample source code by clicking the link near the
top of the page. This will open up a simple color-coded example
of where each script should be inserted on a simple example test page.
After you've completed your section script tagging, copy and paste the conversion script
directly before the closing "/body" tag of your conversion page.
Once all of your pages have been tagged and uploaded to your server, click "Validate pages."
If your pages are externally visible, Website Optimizer will directly check them on your
live site. If your pages aren't accessible, you can use
the offline validation option to upload the HTML source file of your pages for us to check.
Remember, tagging causes no visible changes to your site until your experiment begins
running. Now that you've tagged your pages, we'll provide
you with your original content as a base for you to add your variations.
To create a variation, choose the appropriate page section and click the "Add new variation"
link. Name your variation, and then make whatever
changes you wish to make to that particular section.
In this simple example, we'll add a few text variations of our headline.
Save your new variation and then be sure to preview it to see what it will look like once
your experiment begins. We recommend starting with a small number
of page sections, such as three to four, and also a small number of variations, up to three.
The more page sections and variations you have, the more traffic and time you'll need
to run successful tests. All that's left for you to do is launch your
experiment. You can once again preview each of your page
variations to make sure everything is in order before you launch your test.
Just click the "Preview" link. You can also select what percentage of your
test page's traffic will participate in the experiment.
The larger the percentage, the quicker your experiment will generate significant results.
If everything is satisfactory, click "Launch experiment."
We'll start the multivariate test immediately, and data should be available in your reports
within 24 hours.