I think there's a better way if you're doing webapps. Since you can change the price for each customer in the backend code you can actually start doing statistics and find out at which pricepoint you maximise your profit.
It might be a better way in theory, but do you know of anyone who's actually employed the technique without severely alienating customers? You mention Amazon as a general A/B tester, but Amazon's price testing is the most prominent example I can think of and it ended with a public apology. Price discrimination (even if it's only a testing phase) only seems to fly if it's not obvious, and it'd be obvious for most webapps.
True. I didn't mean that some markets don't support price discrimination, just that customers' perception of it -- if they know it's going on -- is terrible. With airline tickets, people see prices fluctuating over time: lots of bumps but also some apparent connection between the time/effort involved with buying early and a generally better price. Customers seem to be more infuriated if they can easily tell that someone else is getting a better price on the same exact seat at the same exact time. Given the fluctuations and the expectation of fluctuations in airline pricing, there's not much direct comparison going on among airline ticket consumers.
More info here: http://www.maximise.dk/blog/2009/01/getting-product-pricing-...
(Full disclosure: it's my own rarely updated blog)