Research Guru by Alice Irvan, PRC, APR

Gone are the days…of “perfect” random representative sample

September 21, 2009 at 11:26 am | Posted in marketing research | Leave a comment
Tags: marketing, marketing research, online surveys, sampling

If the days of a perfect random sample ever existed, they are gone like the snows of yesteryear.

Those of us in marketing research understand this and are working with the sampling available to provide the most appropriate sampling frame possible for each project. We know that a RDD (Random Digit Dialing) sample is not perfect due to the increase of cell only households, especially for urban dwellers, minorities and younger adult populations. Add to that the issues surrounding contacting cell phones without prior authorization, and the issue becomes more complex.

The purpose of a marketing research project is to obtain information for decision-making on specific issues. Explaining the sampling method used and listing any limitations that exist are good research practices when reporting findings. This information allows research users to consider the “perfection” of the data.

Sampling people’s opinions is not the same as randomly sampling objects as they come off an assembly line. Individuals may not be easily reached during the period of a research study; individuals may decline to participate; individuals may have different methods of preferred contact such as by email rather than by telephone.

An article in the September issue of MRA’s Alert! “The Virtues of Consistent Bias: Online Research Must Move On” by Steve Gittelman and Elaine Trimarchi, PRC, provides a thoughtful look at online sampling, specifically. They recommend that we consider the element of consistency when using online sampling, and they suggest using multiple sources, if possible.

The key concept of this column is that for many populations in marketing reserch, identification of all members is frequently difficult at best, and impossible at worst. Additionally, it may not be possible to reach a completely random representative sampling by any one method. The “old” concept of always using probability sampling no longer works in many, if not most, instances. Instead we must define our sampling practices and make decisions using our best judgement. We must move forward into the age of non-probability sampling fully informed.