Google Web Optimizer - Things I've Learned
In October 2006 Google launched a new tool for AdWords advertisers called Web Optimizer. Web Optimizer allows site owners to experiment with different versions of a web page to determine which combination produces the most conversions from site visitors.
I have been using Web Optimizer for the past few months and have become quite familiar with the implementation process and have learned some tips and tricks along the way.
Process
First, I have found it easiest to plan your experiment through before you even jump into the Web Optimizer interface; it will make the whole process easier later. Some questions you need to consider are:
- What page(s) do you want to test?
- What sections of this page do you want to change?
- What different variations do you want to test for each section?
- What resulting user action do you desire?
- What page(s) will users end on if they perform the action you desire?
For example, let's say I have an online store that sells widgets. This is how I might answer some of the above questions.
- I want to test the product page www.abcwidgets.com/bigwidget.html
- I want to change the product image and description
- I have 2 different images (one close-up and one full-product) and 3 different text descriptions
- I want the user to add the product to their shopping cart
- Users will end up on the shopping cart page www.abcwidgets.com/shoppingcart.html
With this information in hand, it is fairly easy to follow the instructions provided as you work your way through the Web Optimizer set-up process. At certain points you are required to modify the HTML code for your pages, so if you are not comfortable with this you will need the assistance of a web programmer.
Challenges
There are some scenarios that aren't as straightforward as the example provided above. For instance, what if your test page and conversion page are on different domains? Maybe they're on different sub-domains? What if you are running Google Analytics as well? What if you have several experiments with the same conversion page? Thankfully, Google addresses all these issues and more in their installation guide.
Tips
Here are some tips I've learned while planning and implementing experiments that may make things easier for you.
- Check and double-check your experiment implementation (especially your variations) using the preview button, because once the experiment is live, changes can not be made.
- If your experiment is risky (i.e. may turn away visitors) you may want to limit the percentage of site visitors that participate in your experiment.
- It is possible to have multiple test pages and/or multiple conversion pages in the same experiment; you just need to put the code on each page you want to test/track.
Google Web Optimizer is a powerful tool and I have just mentioned a few of the things I have learned while using it. The best way to discover how Web Optimizer can change your site's performance is to begin creating experiments. Start small until you get the hang of things, but then with some creativity you'll be able to test all sorts of combinations on your pages and many different user actions.
If you like this post then please consider subscribing to our RSS feed.
I have seen that you have been experiementing with Google WO.
Do you know of any shopping carts which support W.O.
Mainly PHP. I have been searching for a lot of shopping carts but none supports it and all of the developers of those carts have never heard of Multivariate testing.
It is possible that some shopping carts that provide access only to admin backend support Web Optimizer. The easiest way to find out is testing. You can use cart's page editor to put JavaScript code on pages where you would like to run a multivariate test.
Thank you very much Alex, for replying.
Alright, if its only javascript then it can be done and its easy.
But I would like to know, since php shopping carts are dynamic and they products image and their details are retrieved from the database, how can we have multiple images to get displayed without changing the URL of the product.
Where do we have to provide the variations, is it in the shopping cart or google.
I am moving away from Oscommerce/Zencart.
I am looking into Viart and jShop shopping carts.
Both of them say that its easy to implement the javascript code.
From your comments, it seems that you are wanting to run an experiment that manipulates the products (which are displayed dynamically from the database). Manipulating dynamic elements on your page is sometimes possible, but it really depends on what you are trying to test. If you could provide some more details on what exactly you are trying to test I might be able to provide some more help.
As to your question about the variations, they are created during the WO experiment set-up through a form provided by Google. Prior to this, you do have to add some code to your test page to specify what areas of the page you would like to change (these are your sections where the different variations would show up). If you need more information on the coding involved in setting up a WO experiment you can read more here: http://adwords.google.com/support/bin/answer.py?an...
Thank you very much for adding to the conversation.
I am mainly looking for implementing Multivariate for testing the products page layout, homepage layout etc.
I would like to see which page performs the best and is most useable for visitors.
For what i understand, we have to insert code on the page which needs to be tested and if one needs to test out Headlines, then one has to provide another page which a different headline. This is easy to incorporate in simple html sites.
But when php comes into play and its dynamism, this creates some sort of problem.
Earlier, i was looking into another Multivariate software which uses Taguchi method unlike Google WO. Kaizentrack and Splittestanalyzer maybe you might have heard of them.
I contacted their developers and both of them told me that its not easy to incorporate Multivariate testing into Dynamic Sites especially php since they generate images, headlines from the database.
Im in a dilemma, how to use Multivariate testing with a php shopping cart ? If its only about inserting a few lines of JS, then most developers of carts are charging excessive amounts of money.
Maybe an example would make things easier to understand. Let's say you want to test 3 different layouts of your product page (regardless of which product is being loaded). This can be accomplished by creating 3 different CSS stylesheets, one for each layout you want to test. If you need help on how to do this, it's best to talk to a designer, as this is not my area of expertise. Once you have your 3 stylesheets created, on your product page (let's call it products.php) you would put a link to one of those stylesheets (ex. <link href="layout1.css" rel="stylesheet" type="text/css">). It really doesn't matter which stylesheet you choose, but this will be know by Google Web Optimizer as the "original" content.
Then while you're setting up your WO experiment, when defining the section code you want to test, you would amend your products.php code to look something like this:
<script>utmx_section("StyleSheet")</script>
<link href="layout1.css" rel="stylesheet" type="text/css">
</noscript>
In otherwords, you are creating a section around the stylesheet reference.
Then in the next step of the WO experiment set-up, you are required to specify the variations you want to test. These are NOT different pages to load, but simply the text you want to swap inside the section you created in the previous step. So, continuing with the example, your 2 variations (remember we want to test 3 layouts - an original + 2 variations) might look like this:
Variation 1: <link href="layout2.css" rel="stylesheet" type="text/css">
Variation 2: <link href="layout3.css" rel="stylesheet" type="text/css">
You are basically limited to HTML when entering text for your variations, since WO swaps this after all server-side code has run. You can manipulate lots of things related to layout/styling depending on where you specify your sections; it's really just a matter of experimentation.
So, in my example, when a user came to products.php page on your site (let's say they're viewing product XYZ), the product would be dynamically loaded and one of the 3 layouts would be used. The nice thing about WO is that a cookie is set on the user's machine so that they will continue to see the same layout (whatever variation they were first served) no matter what product is loaded on the products.php. So, if a minute later they go to view product ABC, they will see the same layout as when they viewed the previous product, but obviously the content will be different (dynamically loaded from the database). This makes for a consistent user experience (provided they don't clear their cookies).
Hopefully this makes things a little more clear; I'm not sure if this is what you were meaning when you said "testing the products page layout". Obviously if you are using a 3rd party shopping cart where you do not have access to the code, you will be limited in the types of experiments you can run because it will be difficult to put your section code right where you want it.
I would recommend reading the documentation found here: http://adwords.google.com/support/bin/topic.py?top... - it is very helpful in understanding how WO works, the different types of things you could test, and how best to implement the code.
Thanks for the detailed explanation.
I never thought that i could bring css into the equation.
I'll test it out and see what comes up. Ill report back if it works or not
Thanks again : )
P.S. How much would VSI charge a small business for internet marketing