Like any research, market research has always recognized that to be certain results of research can be projected to an entire population; you need to eliminate any bias. We worried about things like:
- Representativeness Effects – Needed to not only make sure we selected a random representative sample, but then do everything possible to maximize the percentage of people who completed the survey.
- Interviewer Effects – Surveys needed to be done identically. If one was done by mail, all should be with identical forms. If done by phone interviewers needed to be careful not to lead respondents and to keep pacing at consistent rate.
- Framing Effects– If responses from one question are going to potentially bias a future response then the order should be changed to reflect it. In cases where changing the order merely changes which question biases which, use rotation or split samples so that bias effects can be measured and softened.
I know this is a simplified view of things, but the above three do get at the major forms of bias that we seek to eliminate in market research. In this blog, I'll focus on representativeness and at some point in the future I'll cover the other two.