General Question

gambitking's avatar

How do i perform "A/B Split" or Multivariate Website Testing?

Asked by gambitking (4206points) February 26th, 2010

Okay techie people, I am trying to test two variations of a new page on my website. I want half the visitors to see page “A” and the other half to see page “B”, all at random of course. What’s the best/easiest way to do this?

Observing members: 0 Composing members: 0

6 Answers

jaytkay's avatar

Whoever does the coding can do it with PHP or ASP.Net or some other dynamic language. The choice of language depends on web server support and the coder’s skills. The ASP.Net code could look something like this simplified pseudocode.

<% If randomnumber is even then %>

<html>
<body>
This is the even page
</body>
</html>

<% else %>

<html>
<body>
This is the odd page
</body>
</html>

<%end if %>

drClaw's avatar

Google Website Optimizer. Learn it. Love it. Live it…

http://www.google.com/websiteoptimizer/b/index.html

It will allow you to do A/B split tests, but more importantly it will allow high level multivariate testing. You should also be using Google analytics or some other analytics that collects as much (if not more) data.

ETpro's avatar

The trick isn’t so much how to provide a pair of pages and chart which is delivered and what the user’s response is. That’s relatively straightforward. The trick is developing software to make sense of the information gained without having to wait years to gather enough data. Most sites have relatively low traffic to a given page, and it takes a lot of data-points to build a reliable statistical model.

gambitking's avatar

@drClaw – yep, exactly what i’ve been lookin’ for. Thanks!

gambitking's avatar

@ETpro -That’s what Google Analytics is for.

ETpro's avatar

@gambitking No, GA won’t drive traffic to a page you want to test. You need enough hits to get a meaningful answer, or testing may be telling you something that doesn’t match the true answer at all. Too small a sample can leave you thinking that what A is better when a statistically accurate sample would have shown B to be the winner.

Answer this question

Login

or

Join

to answer.

This question is in the General Section. Responses must be helpful and on-topic.

Your answer will be saved while you login or join.

Have a question? Ask Fluther!

What do you know more about?
or
Knowledge Networking @ Fluther