Does Anyone Use Google Website Optimizer?

[RDude]

I was just checking out some stats in Google Analytics and was wondering about Google Website Optimizer. Does anyone here use it and has it helped your conversions at all?

Thanks:cool:

[fantastvik]

this is really looking like a fantastic tool to me, first of all i will try my adult jobs (http://www.pimpwork.com) site on it!

[gaydemon]

Looks interesting, but not tried it yet. Might do actually, just looks a tiny bit complicated.

[random]

I have been using various split testing & multivariate testing tools for years, w/GWSO as the most recent.

I'm only on my 2nd test using GWSO specifically.

The first was in mainstream (so I won't link from here), and it did help confirm our gut feeling on which one of 4 'hero shots' was most effective.

The second is currently running on a clients alt/goth/emo porn (http://alternadudes.com/red) site where I am helping them to test 8 different headers in one test, that is being split-tested against a different set of tour vids.

(Either use a different browser or clear the cookies to see a different one).

GSWO has some advantages in that it's free, hosted, and documented.

One thing to note is that it tests EVERY combination. While this is the most accurate way to test, it can require a huge amount of traffic to get any actionable data depending on how many variations/segments you have, and your conversion ratio.


If you are interested in the math you can check out this spreadsheet here (http://3ob.com/split-testing/split-test-sample-size-calculator.html), which is from an old workshop I gave at Internext.

OR


If you just want to see how much traffic it would take you can use this online split test/ MV Test calculator (http://jeffrandom.com/calc/#last_segment)I built, but never got around to documenting.


Just add a 'segment' for each element that you want to test, and adjust your conversion ratio. If you like you can vary how confident you want to be in the results, and see how that affects the amount of traffic required.

Oftentimes, it makes sense to take action w/o waiting for statistical significance, or to instead constrain your tests when you want to be sure.

Basically the more things you test the more traffic you have to send for it to mean anything.

GWSO doesn't let you target by behavior (AFAIK only 2 tests in), but for basic testing it's plenty and even in many ways is up there.

No matter what you use, if you have any significant volume to work with you can often get more from your existing traffic by continually doing smart testing.

So whether you use GWSO, or any other tool I do encourage you to test, if/where you have enough traffic passing through, and enough expected gain to make it worthwhile.

This post is getting kind of long, so if you want to know more about this kind of stuff let me know. Otherwise I hope that helped.

[RDude]

Some great information Random. Thanks so much! Sounds a bit complex, but once you wrap your head around it it's not too bad. LOL.

Regards...