The interesting thing I have seen with SEM is that competition is not always obvious. Impressions and clicks at the campaign or even keyword level may not reveal geographical or specific query based competition that may be targeting productive traffic.

Measure, test, measure is pretty much vital. With a lot of SEM, the decisions people make are often based on a model and a collection of assumptions. It is as important to assess these as it is the specific tactics that are used. This is especially true in a market like Adwords, where the tools available to the advertisers change frequently.

The defender in SEM should actually have the advantage, assuming that they have been paying attention to the data that they collect. It is the advertiser entering the market that has to accept a higher level of risk as they start to build a picture of the query space.

Again, thanks for the comment, it was pretty interesting.