In search, it is a misnomer that a user will run a query and then go through the results in chronological order from first to tenth.  More often than not, a user will run a query, skim the top-line results and see if anything strikes his or her fancy.  If not, then the user will tweak and re-run the query to see if the new set of results contain more relevant results.  The basic act is actually quite an accepted research methodology for search-and-retrieval.  You run a query, assess the relevance of a sample of results, and then use the feedback to refine your query.  I actually think the basic method at work here moves into a number of other spheres.

At the SES conference, one underlying theme was the recommendation to simply put an ad up and see what type of results you get.  Based on those results, you tweak the ad accordingly and put another one up there.

I think the method represents a profound shift in an approach to advertising because suddenly, advertising becomes a dynamic ongoing act instead of a static one.

Traditionally, advertisers and marketers are dealing with one-shot mediums: magazines, billboards, broadcast commercials.  Consequently, in the traditional model a tremendous amount of research, planning, and design went into the production of an advertisement.  The refinement of an ad was normally based on industry expertise or personal preference.  In the interactive model, an ad’s refinement should be based directly on audience feedback because with Web ads, you can literally tweak the ad in real-time in response to the feedback you are getting. 

Basically, my recommendation is that when planning a Web marketing campaign, the best strategy should simply be to craft an ad based on the research and expert input, but go live as quickly and efficiently as possible.  Once you’re live, study the results and tweak the ad based on best practices as frequently as you can.