CiviCRM Community Forums (archive)

100% Agreed.

I'm trying to convince a big user to do it (for their donation pages). I tried for a campaign (not linked with civi) genetify, works nicely and definitely something anyone that does serious campaigning should do

If we can get some traction, doing a/b on mailing would be great too

The ExtendedMailingStats module extension doesn't do anything explicitly related to A/B testing, though it might perhaps be used to compare results.

I'm certainly interested in mechanisms to further optimise fundraising appeals. My recent interest is more along the lines of using donation history to do things like adjusting the suggested donation amounts in line with what the person is likely to give.  Similarly I've been thinking about mechanisms to send more or less email to people depending on their history of actually opening it.  Given that individualized targeting capacity, we'd still need ways to collect info on user responses in order to optimise the settings for that.  Early days, and I'm just thinking it over so far.

Regarding A/B testing, I like the approach of eg Optimizely, where the testing is externalised about as far as possible from the CMS and CRM, and put squarely in the hands of non-technical people to run tests.  That's not a free service, but it should pay for itself, and be cheaper than development time in most cases.  There are things you can't do that way without some back-end support though.

Happy to discuss further, either here or privately, and yes, I do have available time coming up soon.

mc0e: I like the idea of incorporating some predictive analytics capabilities into our offering.

I am hoping to get do some work using R spatio-temporal analysis on voter data combineing census info, polling info, canvass results, historical info where available on the voting intentions in prior campaigns, etc. In a former career I worked on GIS environmental modelling systems.

I am currently pitching for a different client an integration with social media intelligence on contacts using fliptop, fullcontact or similar products. Depending on budget, time, etc. it might also be possible to append consumer info. The more data points, the better the likelihood of getting good predictions out of data mining techniques. I piloted a bit of this in a by-election a few years ago.

Lots of fun possibilities...

I've used this for a/b testing mailings

http://svn.civicrm.org/civicrm/trunk/CRM/Contact/Form/Search/Custom/RandomSegment.php

Split the group then send 2 mailings.

Bit clunky to split the mailing though.  Would be great to have some a/b testing built into CiviMail. as Joe mentioned sending the winning variation to the remainder is the gold standard for email marketing a/b testing as you get the 'benefit' of the winning variation immediately. Back when I was an email marketer I'd send to 10% vs 10% of list and then send to the remaining 80%.

For donation pages/forms etc.. we use http://visualwebsiteoptimizer.com/ not sure how you could integrate it though, we use it in combination with google anayltics ecommerce tracking for contribution led testing.

Would love to get some predictive analytics integrated but we're a little way off that as an organisation.