A/B Testing with Google Website Optimizer


Here’s my question:

Which of these  images will decrease my bounce rate?

Layout A with ‘Angry Bear’ image:


Layout B with ‘Happy Bear’ image:

Google Website Optimizer seems like the tool to use.  And, since I’ve never done an A/B test like this before, I really have no idea what to expect.

Jump over to Google website optimizer, and sign in.  There’s no problem here because I already have a Google  account.  I don’t have to pay, because it’s free. Easy so far.

RTFM? Nah. You write a better FM

Today, I avoid the help section.  This might throw up a red flag and it’s certainly not my typical strategy, but hear me out.  I don’t want to ‘learn’ about A/B testing.  I want to A/B test.  I tried that ‘book learning’ last week and I got stuck in the weeds.  I don’t want to be trained.  I want to run some tests.  I want to understand A/B testing and I want to be ‘leet, but I want that understanding to come from “doing it”, not “reading about it”.

I click on start experiment.  In order to test, I need to complete three tasks. [Did a magically creature give you the three tasks to complete? -ed]

1. Choose the page I want to test.  Got it, http://www.alaskagold.com/index.html.

2. Create alternative versions of the page. Got ’em, http://www.alaskagold.com/index-a.html and http://www.alaskagold.com/index-b.html.

First problem: “These URLs could be bookmarked by your users, so after your experiment finishes, you may want to keep these URLs valid.”  Sort of sucks, but okay.  Will this mess with my Analytics?

3. Identify the conversion page.  “This is an existing page on your website that users reach after they’ve completed a successful conversion” … Hmmm.

Second Problem: The goal is to decrease bounce rate or drop-offs of this front page.  So, I don’t have a single conversion goal. All and any page of the site is a conversion page, in this instance.  Perhaps this tool isn’t the best way to test this.

All right. Making an executive decision, I decide the pick the most linked to page from the test page: http://www.alaskagold.com/valdez/index.html.  It gets 13% of the clicks off the test page. I have no idea why I chose it this way.

All three tasks completed, it’s time to click the create button!

Name the test.  Paste the URLs (which I’ve already collected because of the last step – way to go Google!). Name the variations something meaningful, I suspect this will be useful later.  I name them ‘angry bear’ and ‘happy bear’.

Next, the process asks “Who will install and validate the JavaScript tags?”. I will.

Install the JavaScript

One script for the test page.  One script for the variation pages. One script for the conversion page.  There is a validation feature which, in my case, caught a problem with the conversion page script.  Got it fixed. Good to go.

I select to start the experiment and I get this message:  “You’ve successfully launched this experiment. Reports will be available in 2-3 hours.”

Here’s where is gets squirrely

I okay – to check it I open up another browser (Chrome) and look at the test page… hmmm, It’s showing the original and not one of the variations.  This shouldn’t happen, because I chose to send %100 of the traffic to one of the variation pages.  I go back and look for the preview and it’s showing the original page for all the variations. What about the conversion page? That’s not shown in the preview section. There is a problem here.

huh, I wonder if it’s a problem with the script not being in the exact first thing after the opening <head> tag.

The code isn’t in the place the Google Website Optimizer wants it, directly after <head> (red arrow)

Note:  I think my problem is that I use Dreamweaver templates and the template won’t allow me to put the code correctly.  The help says: Place the combined block of code immediately after the <head> tag.  But this is strange, because all of the other scripts really don’t care about placement – only that they are INSIDE the <head> tag.

I test this hunch by removing the template from the pages – so the code can be right next to the <head> tag –  and creating a new test.  And, it works!

Here’s my dilemma: I could remove those pages from the template or add a new template ‘editable region’.   Removing the template tags isn’t the best option because then I would have to fix it after the test.  Putting in a template tag directly next to the <head> might work, but it’s a pain to try it out.

But is the placement of the code the REAL problem. I’ll take a break to think about it.

I just drank 3 carrots!

[Smoothie Break … 1 hour later]

WTF, it’s working now?  When I go to http://www.alaskagold.com/index.html, I get forwarded over to index-a.html.  But there is no test running on index.html. Okay, I left the code on the server, remote file… but which code was it? You can’t go back and see the code snippet once you pass that screen in the setup wizard.  Was it the code from the first test or the second?  Are they different? Which one has gone rouge… I feel the rage monster about to bust out.

Okay, count to 10. Time to reset, take a deep breath, and go back to square one. Something is obviously EFFED here.

I’ll start a new test. I want to use Google Optimizer to see which of these pages will engage visitors to click further into my site – reduce my bounce rate.

Step by step and very carefully, I re-create the test.  Collect the URLs, paste in the scripts, and upload everything. And, without too much fuss, everything is smooth and working.


I wonder.  This is a real mystery – I have no idea why it wasn’t working and then started again.  Did it fix itself?    The scripts are working despite not being exactly after the <head> tag.  So, I wonder why they say that.  Is it a big deal?

We’ll see what the reports say. It might be a day or so, according to the site.

Frankly, I’m just glad to get it to work.  I’m not sure what happened but I would not recommend this to a noobie.  Not to say today wasn’t time well spent.  I’m convinced of the data-driven, evidence based thing is not a fad.  And, using this testing techniques will improve websites.  It just times time and energy to learn.  So, recanting my previous statement, I would recommend it to a noobie, with the caveat to be prepare to spend some time.


The next challenge will be to set up an experiment in WordPress.  There are plugins for A/B testing with Google Optimizer and methods for doing it ‘by hand’. I may tackle this next week.

3 comments on “A/B Testing with Google Website Optimizer

  1. Re-reading this post, It feels like I’m watching a 20 something version of myself hacking away at a problem.

    I do feel that reading manuals and being trained is important, but jumping in and testing something is equally important.

    I’ve concluded the test and I’ll share the results soon. Which do you think was the better homepage image?

    • Hey Bear!
      Thank you. Google Optimizer is a great way to split up the traffic and, if you have a unique conversion URL, it should work perfectly to give you actionable information. Also, the more traffic you have, the better the results (and quicker)

      Good luck and let us know how it works!

Leave a Reply