has any one done split testing in joomla?

I would like to do split testing based on the refer that the user has when they enter the site. So based off what was the keywords they searched for, or the referring site. I want to give different content. Measure the response and see how I should build the site?

Hmm.. Thats sounds interesting. ( how like you!) How would you do that? Could you do it with conditional statements in the index.php?

I will probably end up writing something.

My thoughts would be to make a special template, and a component for static content.

When a customer first comes to the site, I would need to look at the referrer and drop a cookie to have that information. At that stage through them in to a (cookie based) group, based on that group change the module names, css files, or content as needed. Then I need to make sure I can track those users so I can see how users relate to that content instead of the other one.

Sounds very kewl!

Qns:

1. do you think it would slow site much

2. could it be done without the component (for simplicity's sake) I'm always trying to simplify things down to a level that I can understand as I'm still fairly basic in my php and java skills!

I'm loving this idea not for testing but for content customising a website........

it should not really slow down the site, a little added to the processing, but not much.




the order of the site being processed (at least in 1.0) is:
1. Component
2. Template
3. Modules

Plugins/mambots happen in different parts of the chain.

So if I want to effect say the module displayed/order then it can be done in template or component.

If I want to change what the text is shown, then I have to do in a component.

Hi gang,

I've been considering this too, but in J1.5. I've been asked to show a different home page each visit, from a selection (say 3) to see which one performs best.

I thought of using an existing feature e.g. banners module or newsflash, which at least show 'random' content and can be measured. Then uou just put your whole homepage content in there. But they'd need to include an RSForm and we'd want to know which copy produced the best RSForm submission results. So yeah, we use affiliate cookies anyway, which helps - can capture clickthru responses and know number of page displays of each page and which copy produced results.

Bit concerned about the randonmess of the random function though. I'm not sure it will produce a consistent even split over the course of time.

Interested to hear how other people do this?

Another (much easier) solution would be to put a Javascript snippet in at the top of the home page which redirects to one of 3 SEF URLs, based on either random or counted page displays. Then the page hits could be traced directly in the article manager and the clickthru on the RSForm will still record source. Could also replicate your RSForm 3 times and include one copy of the RSForm in each page - thus no cookies involved.

if your happy to use javascript then you can look at google website optimizer
www.google.com/websiteoptimizer