Google Web Site Optimizer Tool: I Told You So

Notice: This is a discussion thread for comments about the SitePoint article, Google Web Site Optimizer Tool: I Told You So.
__________

The TOS section of their form to download the software doesn't work at the moment. It doesn't move the user on to the next point.

[posaune02]

I'm assuming this is just a typo in the article and not also your attempted 'fix' for HTML validation but in the article it says <script type="javascript/text"> where it should be <script type="text/javascript"> although that mimetype is now obsolete. It should be <script type="application/javascript">

By the way, you ought to add OpenID support to the forums. It shouldn't take 15 minutes to post a stupid comment. I hate having to register, wait for email, confirm, just so I can tell you about your typo.

[rayver]

You can run A/B scripts on a site. Thats awesome!

I was just talking to one of my buddies the other day from Marketing Experiments and this was one of the things he mentioned they do a lot to test different things and see what works best.

Thanks for the article.

[atetlaw]

Originally Posted by posaune02

I'm assuming this is just a typo in the article and not also your attempted 'fix' for HTML validation but in the article it says <script type="javascript/text"> where it should be <script type="text/javascript"> although that mimetype is now obsolete. It should be <script type="application/javascript">.

while 'text/javascript' is obsolete I believe IE will not run a script with a type attribute of 'application/javascript' (http://annevankesteren.nl/2006/05/javascript-mime-type)

So we're in the strange position of having to specify a type attribute to ensure validation but only able to use an obsolete one....

[ole90]

It was interesting enough; though i doubt i'll be using it anytime soon, due to my internet traffic being non-existent. By the time i have suffiecent results from something, i'll probably have moved on by then and the testing project won't matter.

Good Article

[peach]

I was always thinking I'd make a split/multivariate testing script myself but maybe this will evolve into something I can use.

[DaveWoods]

It's hard to prove either way, but there may be some effect on the PageRank of your original page if you left the experiment running for a long time.

I've not tried using the optimizer yet (it does look extremely useful though) but I can't see that being a problem from a PR or SEO perspective. Presumably the JavaScript does some form of redirect to pass some new users to the new page and some to the original page and therefore because it's all being handled by JavaScript, the backlinks will always point to the original page and therefore PR will be uneffected.

If I was in any doubt, I'd also suggest disallowing the new page from robots.txt file to ensure that it didn't get listed.

I'd go for "text/javascript" myself...

[posaune02]

I understand using text/javascript in the type attribute on a script, to cover for IT. Though I believe IE7 now properly runs scripts that are returned from the server with an application/javascript Content-Type header.

Though this is really irrelevant to the matter at hand, which is of course that the article specifies javascript/text as the type attribute. (and has yet to be corrected.)

[mkoenig]

Waat?

[panther786]

I was just talking to one of my buddies the other day from Marketing Experiments and this was one of the things he mentioned

[deaconmpg]

Yup, I currently am running Optimizer Experiments on several sites and it is most revealing. Also you can get the Test to run across your whole site if you use the multi-variant testing and include the JavaScript snippits in everypage. I am testing the effect of display security badges to get users to fill out our contact forms. So I have the security badges on/off through out the whole site template using the Multi-Variant testing.

[prutwo]

In the report image it's showing the original with 2544 visitors and the test version with 1461 visitors.

This sounds suspicios as each should get the same amount of traffic?

Any comments?

[mkoenig]

Do you have 2 pages or just 2 sets of code that google juggles back and forth?

Decision in 12 hours!

If I were you, I would run that test at least a week. How can you be sure that it wasn't a specific user segment that preferred one version over another? A restaurant by day may have a much different clientele at night when it's a 21 and up bar. It appears that traffic is not your problem. Why not set GWO to test with a subset of traffic say 10% over a longer time span. That would definitely increase the reliability of your results.