When we began working with American Motors in the early 1970s, one of our first
contributions, and one which initially seemed counterintuitive to our client, was to
narrow its previously used research sample of all potential new car buyers to only those
why would consider an AMC car. Our rationale was that the data from nonconsiderers of
AMC had been distorting the research because these responses tended not only to be
more negative, but also were more heterogeneous and generated more random error.