rogerd - 3:14 pm on Jul 20, 2005 (gmt 0)
Quite true, linear. Day to day variations make it tough to compare different approaches. To the extent you can do a split-run test, that's probably the ideal. Of course, since it the test has to do with ad frequency rather than ad design, you'd probably have to make the split cookie-based so that each visitor always stayed in the same group for the duration of the test.