By normalization, I mean first bidding a flat rate on a large number of keywords and then after a few months, tweaked it so that the CPA for each keyword was the same.
Eg:
Initial Keyword Bids
$0.50 Widget Blue
$0.50 Great Widgets
$0.50 Awesome Widgets
$0.50 Widgets for hire
After two months, the CPAs turns out to be
$5.00 Widget Blue (10 clicks / aquisition)
$2.50 Great Widgets (5)
$10.00 Awesome Widgets (20)
$1.00 Widgets for hire (2)
Let's say you want to normalize around $5 CPA. So you normalize the keyword bids on Target CPA / Clicks Per Aquisition ..
$0.50 Widget Blue
$1.00 Great Widgets
$0.25 Awesome Widgets
$2.50 Widgets for hire
Has anyone tried that? What impact have you seen on your over all revenue and profit? Did you continue to tweak? Do you see a convergence of some sort to meeting your target CPA? Have you found a more affective formula for converging on an ideal bid rate per keyword?
It would be even better if Google could do it on a (keyword phrase, location, time) tuple.
Eg: if I broad matched on "blue widgets" and it found that "great blue widgets" converted at a CPA of $1 at aol in canada at 3am then at 3am on aol in canada it would increase the bid rate by five times (assuming I'm targetting a 5$ CPA)
Obviously, PPC engines will never be able to compete on price when it comes to PPC advertising.
What they will be able to to is compete on reach and how well they convert, and so therefore I believe that functionality like this is only real future for an increase in added revenue for PPC search engines.