Google Optimizer Code – or How I learned to survive the “test”
I was fortunate enough to have a client in need of comparing two callouts on one page. Like most clients, they use Google Analytics. They wanted to run the Google Optimizer tool on a couple pages. This is also supposed to enhance SEO performance.
Initially they wanted to try two different layouts of one page. So we ran a simple A/B variable test. After running the test live, they found that users interacted with option B almost twice as much as A. These were great results.
Developing and refining the code was not so simple. There are several forums out there that helped me determine the correct way to implement the test. There are at least 4 different ways to integrate the optimizer code into the existing GA tags. Each integration depends upon how the tags are included on the page – SSI’s, hard-coded – and the type of server – apache, jsp, .Net.
Google provides a “testing” page to check your code. It has to pass the test in order to be recognized by Google. I finally finagled the code into the existing GA tags. And it passed the Optimizer testing page.
The client was happy and now wanted more testing. This time they wanted to see compare callouts on one page. In other words, 1/2 the users would see callout A on page 1 the other half would see callout B also on page 1. Now I really had to flex some code muscle.
I built the two versions and ran them through the Optimizer test page. After several modifications, I still couldn’t get a “pass”. So I had to delve into the world of “multivariate” testing.
I’ll save the gory details, and just say that it took a good week of development to successfully build the pages. I think i held a small party when the Optimizer test gave me a “pass”.
The client could see click-thru rates on the same add with different wording. A very granular test with amazing results. Who knew that a couple words could change a buyers’ mind so drastically.
After surviving such a code-intensive workout, I kept all the clips, bookmarks, and trial code for future use.
If another client decides to test, I’ll be ready. Let’s just say, it wasn’t easy, but it was worth it.
I would like to add that other sites use the A/B testing as well. Just saw this note from Nieman Lab about how the Huffington Post uses A/B testing for their headlines. Pretty cool stuff. Then again Nieman Labs is just plain amazing!