Matthias Felleisen wrote at 05/06/2011 10:41 AM:
For what time period should we leave the description constant to test this conjecture?
Someone mathematically-inclined did something similar-sounding a decade(?) ago, for US national political campaign fund-raising. From what I could gather, the campaign Web sites would automatically experiment with varying the approach, *on a per-visitor basis*, to find optimal ways of presenting itself. I think they were searching the combinatorics of different text, positioning, dollar amount alternatives, etc., and also correlating with messages in the news. In one respect, they had it easy: dollars donated immediately through the site is an exceptionally good success metric. I suspect that the party didn't publish on their setup (just guessing; I don't follow such things), and the person I'm thinking of has moved on to doing different kinds of work.
-- http://www.neilvandyke.org/ _________________________________________________ For list-related administrative tasks: http://lists.racket-lang.org/listinfo/dev
