A/B Page Testing

[Mike Sack]

Let's do a quick poll on A/B Page Testing.

I do a lot of A/B testing and find that I can quickly improve conversion rates from just doing these simple tests.

[rustybrick]

Maybe this thread is not being replied to because some people might not understand what "A/B Testing" is.

A/B Testing is used to describe when you test two types of ads (doesn't have to be ads) and see how the two ads differ in terms of your desired response.

For example: I put up two ads at Google AdWords. I run one on a particular day and have the ad go to a particular landing page with a blue background. I notice a clickthrough rate to the page is X and the order conversion rate is Y. Then on a different day, I run the same ad but change the background color of the landing page to red. Does the clicktrhough rate to page the page change? Probably not, its the same ad. Does the conversion rate change? The goal is for the conversion rate in increase.

So you isolate variables and test how a tweak to a variable can affect a metric your measuring.

More info here http://www.jupiterwebevents.com/webinars/0401opt/

[Alavina]

A/B testing should improve your page. Maybe many can't be bothered that much... it's a question of effort you put into your site...

[AussieWebmaster]

I use A/B testing everywhere... the creatives are juggled.. I actually like Google's optimzer for pushing the most traffic to the ads that are working best.
I do it for versions of landing pages and even graphics.

Hell I even do it with dates... what lines work best, what restaurant helps close....

Life is one big A/B test!!!!

[JOhannes]

Hi,

We have out site hosted by an ISP, who uses an Apache Server.
In order to do A/B Testing and track it with our Websidestory Professional software, how do we make sure visitors to our site are served different pages (Visitor 1 page A, visitor 2 page B etc.)?
Our site is implemented in static HTML and CSS.

thanks

JOhannes

Below is an explanation for larger sites, but we don't have that option available resource wise.

"Several approaches to how to split the traffic, essentially one is hardware driven and one is software driven.
Hardware - We have multiple servers due to our volume of traffic and our need for redundancy. Our routers and load balancers split the traffic to the different servers. In this case we put the A code on one (or one set of) server and the B code on the other one or other set of servers. The issues here are to ensure that your load balancer does not introduce bias - for example all visitors from the Northeast go to one test. Also, if a server goes down - this event can skew the test.

Software - We are a very dynamic, database driven site using Coldfusion. Using the software, we can write code that essentially gives visitor 1 a path to test A, visitor 2 gets the path to test B, visitor 3 and so on. While the initial programming is harder - it is much easier to move tests through this process
Having talked with other companies, most A/B testing is based on a version of the above. Depending on a site's configuration one or other of the approaches might make more sense."

[AussieWebmaster]

The easiest way to do this would be to change the pages for a set amount of time and then compare the activity of both.

[JOhannes]

Thanks AussieWebmaster,

Hmm, that might be a workaround, but not give true randomness because that would be at different time and parts of the week etc. which might attract different visitors to the site.

I did some more searching and found

http://www.offermatica.com/

and

http://www.optimost.com/

I don't need all the bells and whistles with Multi-Variant Testing etc., but some simple A/B testing would be nice.

A webmaster told me:

"For splitting traffic to carry out an A/B test in 50% - 50% or 33%, 33%, 34% you can use PHP. The easiest way would be to define different style sheets. Every visitor would get a PHP session and the server would memorize, what version (A or B) the previous visitor received.

More info you might be able to find at

www.php.net

and complete classes under
www.phpclasses.org

i.e. logging

I don't know PHP.
Does anyone has a ready script for this?

Thanks

JOhannes

[AussieWebmaster]

Thanks AussieWebmaster,

Hmm, that might be a workaround, but not give true randomness because that would be at different time and parts of the week etc. which might attract different visitors to the site.

I did some more searching and found

http://www.offermatica.com/

and

http://www.optimost.com/

I don't need all the bells and whistles with Multi-Variant Testing etc., but some simple A/B testing would be nice.

A webmaster told me:

"For splitting traffic to carry out an A/B test in 50% - 50% or 33%, 33%, 34% you can use PHP. The easiest way would be to define different style sheets. Every visitor would get a PHP session and the server would memorize, what version (A or B) the previous visitor received.

More info you might be able to find at

www.php.net (http://www.php.net)

and complete classes under
www.phpclasses.org (http://www.phpclasses.org)

i.e. logging

I don't know PHP.
Does anyone has a ready script for this?

Thanks

JOhannes
You can do a true A/B test with those methods... mine was a down and dirty quick solution...