There are two ideas in "Advertising Effectiveness" by Jerry W. Thomas that seem especially important.
The first is that you should use the same pre-testing methods consistently. This makes sense because if you are pre-testing advertisement A and advertisement B, in order to compare their effectiveness, you need to use the same tests, or your results may just reflect a difference in testing method.
The second really logical idea is that you need to use a mathematical model to measure brand name recognition as a result of the advertisement. If people don't remember the brand and buy it as a result of the ad, the ad is a failure.