Posted by samhassell on October 10, 2007 at 6:03am
Hi Guys,
I've just finished a basic AB Split Testing module for Drupal 5 that allows you to select 2 themes and offer them to alternating visitors. It tracks whether they reach a supplied action page (such as a thankyou page after a sale) and then lets you know the conversion rate for each theme.
You can download it from:
http://webzen.com.au/drupal-split-testing
I've applied for CVS access to get this into the module directory. Any suggestions, bugs etc are welcome.
If anyone knows pgsql can they please check / edit the install file and send a patch - at the moment the sql for pgsql is the same as the one for mysql (which i know).
- Sam.
Comments
Sweet
Thanks alot for sharing! I will definetly test this module asap cause what you have wrote so far sounds quite promising. The only thing which doesn't seem to be perfect yet ist that I'll have to disable caching. Is there any change that you'll update the module adding this feature or is it technically impossible?
Sebastian
ps: Tryed to comment on your blog, ither it doesn't work or the comments go to moderation quere without a notice.
I have wanted this for sooo long in Drupal
Thank you!!! I assume one would duplicate Theme A, call it Theme B and then make some minor tweaks to it rather than an entirely different theme. My request would be integration with Google Analytics so that I can see how well each them progresses down a conversion path. In other words, maybe Theme A is better for step one and theme B is better for step 3. Is there anything in an analytics package that we could filter on to show which theme was in use?
Thanks,
Ben Finklea, CEO
SpryDev Search Engine Marketing home of the Drupal SEO Podcast
We guarantee web profits.
512-989-2945 x204
mobile: 512-632-4222
f: 512-857-0212
ben@sprydev.com
Ben Finklea
Chair, Drupal Branding and Marketing* Committee
*BAM!
I got this module in the
I got this module in the repository now so there is a project page for it at http://drupal.org/project/split_test
Sebastian: I havent actually tested it with caching on, but from my basic knowledge of the Drupal bootstrap process, the caching would occur before the code that assigns the theme is executed. So therefore all visitors would see the same theme. As I said though, it may work and I may be wrong.
Ben: Yep just a minor tweak, test, minor tweak, test, rinse, lather, repeat. I havent really used Analytics beyond the basics so not too sure how you want to integrate that. I am looking at creating another module to track sales funnels so you can see where in the sequence you are losing people (like for a shopping cart, or for upsell, downsell style sequences). Just a matter of finding the time to do it I guess.
Thanks for the input!
Why Themes and not Nodes?
Hi Sam:
I was considering developing my own A-B testing module that tests two different nodes and I found your post. Do you mind if I ask why are you testing different themes vs. different nodes? Is it to specifically test the quality of the theme, or are you using it to split-test content (too)?
-Mike Schinkel
http://mikeschinkel.com/
-Mike Schinkel
http://www.mikeschinkel.com/blogs/
Amazing!!!
i was thinking for a while how to get a/b testing into drupal. for me the key in a/b testing are many small steps. means just minor changes to the layout, colors and buttons, then compare and move on. if you do too many changes at once you will not be able to identify the factor that brings improvement. That's possible with your template switching of course.
But it would be great to have the variants as variables on template level. Means you could include a placeholder in the template and drupal would randomly switch them and record.
Another suggestion: Multivariant testing would be great. so if i care about the background color of a landing page for example, i could set up 50 variant and then see pretty fast which one kicks in and which one not.
anyway, great module. i'm writing this before i even downloaded it, cause i've been waiting for it soo long.
oh, this is old, why i have
oh, this is old, why i have not seen it before.