We do 'conversion improvement' exercises for our clients - at least those that 'get' it enough to be receptive.

One of the tools in our toolbox is 'profiling' and developing target 'personae' which we can then target with their own specific sales message.

Of course, the key to getting this to work is being able to get the right message to the right people.

So far we:

• Segregate them by organic search phrase (ie. making the pages fit the persona of the most profitable incoming keyphrase, and building new pages around the keyphrase/persona combination).


• PPC along similar lines.


• Geo-target the message.


• Use 'lightbox' pop-ups that ask the important question, then take them to the page aimed at them (testing at the moment, but woking surprisingly well so far)

Questions:

• Is anyone doing anything smart with cloaking - if so, how do you see the risks these days? (especially as, with a little care, you can disguise cloaking as split run testing)
• Any clever tricks for interweaving two different messages on the same page? (ermm...I don't know..  question buttons that open divs perhaps? simlar ideas?)
• Anyone got any other useful tricks/techniques for matching visitors to the best message for them?

..... and for JasonD in particular, any plans for taking Pzyche into more mainstream markets?

In an ideal world, you'd be able to poll your analytics package for historical information on the user - i.e. what pages they've visited, what search terms they've used etc., and tailor any page using this information. Currently I'm using Yahoo Analytics, where you can't do this. However, higher end systems such as Unica NetInsight can provide visitor-level data. Even with Yahoo, using the API, you can get data on aggregated most-viewed pages though. Our database architecture has a white-list of pages, so we can use the Yahoo data to apply a generic 'page-weighting' for all users. Doing this with user-level data would be very powerful. Yes, you can roll your own, but there's always going to be an element of crossover with data that's hidden away in your analytics.

What sort of signals do you use.

Stuff like?

If google toolbar installed liklihoos of male x 1.1

If 11pm Friday night and referee on of 20 sites we know of then likely to be male 20-35 with some disposable income.

Quote
if you are interested in SEO you have a huge negative correlation to gambling, compared to Mr or Mrs average?

Certainly applies to me - if I can't calculate that the odds are in my favour, then why do it? 

Jason I know I've been begging for retail, but any chance you'll have some other financial stuff coming up soonish? It would be quite useful in the pensions and life insurance side of things and there should be a big affilaite side to that one too?