By Ken Erickson PhD - Paceth CEO
Theoretical sampling often seeks maximum variation rather than a "representative" slice of reality (Miles and Huberman 1994).
In other words, anthropologists (or any ethnographers, really) are interested in the systematic study of the contexts surrounding a particular consumer product or business practice. They want to flesh out the real-life meanings behind product choice, purchase, and use.
What is this meaning business, anyway?
Its more than words. Just ask a surrealist. (Or Foucault.)
Linguists, who are supposed to know something about meaning, are often asked to explain how one knows what a word means. Usually, their answer is that to understand what a word means, one should see how the word is used in ordinary speech (Ogden and Richards 1952). Understanding the context in which a word is used (and the contexts in which it may not be used) is the key to understanding its meaning(s). The same is true of IT products and services like mobile phones in India. Mobile phones are part of multiple contexts--home, work, family, street, train, and so on. And as they move from pre-sale to sale to delivery to use in a variety of contexts, and, finally, to disposal (or resale), what they mean and how they are used changes quite a bit.
To understand the range of meanings mobile phones may have for Indian people, or for any kind of people, you have to see them used in context.
If anthropologists find meaning in the contexts that surround what people do then why would the individual person be the unit of measurement around which to build a sampling design? Clients may ask, "How many people will you observe? What kinds of sampling frame will you design, and what kinds of people will fit into that frame?" Our answer is often "We don't know." That is hardly a satisfactory answer when one is trying to win a research contract.
So there is trouble in the sort of means-based statistical clustering used for determining market segments and, likewise, trouble in many client's expectations around sampling. The first kind of trouble lies in the selection of questions or question categories for the initial segmentation questionnaire. How can you know that the questions the client used were the right questions to ask? How did the client determine which dimensions of taste or practice to include or exclude, and what did they overlook completely?
The next trouble comes in selecting the factors for clustering. Which were the most significant? Usually they use those that are most significant in statistical teams but is a statistical norm--taken as a moment in time--the most strategic element to select from a moving target like the evolving use patterns surrounding mobile phones in India? Does the norm include the tail ends of the curve--the outliers, the users on the edges of the normal pattern? If the pattern is put into motion through time, the users at the edges--early adopters and adaptors--then the users--and contexts that surround--them who are at the edges of the normal curve most certainly should be included in design research.
Next: Toward Ethnographic (contextual) segmentation.
Works Cited Part II
Matthew B. Miles and A. Michael Huberman.
1994 Qualitative Data Analysis: An expanded Sourcebook. 2nd Edition. Thousand Oaks: Sage.