Discission: Validity Procedures in Consumer Behavior Research

Albert C. Rohloff, Market Science Associates
[ to cite ]:
Albert C. Rohloff (1975) ,"Discission: Validity Procedures in Consumer Behavior Research", in NA - Advances in Consumer Research Volume 02, eds. Mary Jane Schlinger, Ann Abor, MI : Association for Consumer Research, Pages: 755-756.

Advances in Consumer Research Volume 2, 1975      Pages 755-756

DISCISSION: VALIDITY PROCEDURES IN CONSUMER BEHAVIOR RESEARCH

Albert C. Rohloff, Market Science Associates

[Albert C. Rohloff is Senior Vice-President, Market Science Associates, New York.]

The validity we look for is the validity we find. The four papers presented here examined many kinds of validity related to data, measurements, models, goodness of fit, constructs. We can infer from these papers, and our own experience, that a researcher will emphasize certain kinds of validation over others.

A specific example illustrates this: Suppose data is available that shows how soon consumers make their first purchase of a new product. This data is presented as a cumulative trial curve. The experienced eye of the analyst indicates that an exponential growth curve would fit this data. Goodness of fit criteria demonstrate that this model fits quite well, not only historical data of the company but data for introductions of other new brands of the company as they occur. The model has predictive validity. The model has been validated in the market place.

A marketing manager may take a different view: It is too good to be true (valid). The market I know is more complex than this. What is wrong here? How can the marketing plan for the next new brand introduction be changed to make this model fit poorly? Notice what the manager is suggesting is an invalidation procedure.

Experience has shown that when a simple exponential growth curve fits trial curves well, it is usually symptomatic of an inefficient (even ineffective) marketing plan. The better the model fits the poorer the plan tends to be!

This example illustrates my opening proposition: The validity we look for is the validity we find.

Lehmann appears to be placing "the predictive model" in a central position. This approach can be limiting if the predictive model becomes an end in itself. Pessemier states: "The commercial researcher's most important problem is not simply to employ valid data and models...it is more difficult...He must find valid relationships that are managerially important."

How can this "managerial" validity be increased? I suggest that if managerial validity is what we are looking for, it is the managerial validity we will find.

An excellent starting point for achieving managerial validity for consumer behavior research is to understand what are the basic beliefs and theories the manager has about consumer behavior. Some of his beliefs can be inferred from the way he markets a brand. When research is keyed to managerial validity, it is natural to identify which beliefs about consumer behavior are most critical and then proceed to test them against alternative viewpoints that may be more valid.

But consumer behavior is, in part, determined by the manager's beliefs. Note that as long as the company introduced new brands in the same way, the feedback on trial purchases was consistent. When an innovative plan was used, consumer behavior changed.

In the appliance example Pessemier describes the creative interpretation of relationships, noting that other research and consumer behavior is also brought in. Often the research is seen as providing the facts and the manager the creative interpretation. As the researcher is more closely integrated in the marketing planning team, he has more opportunity to contribute towards "managerial validity".

Pessemier sets up a dichotomy of academic research and commercial research, seeming to say that the validation criteria for academic research are less demanding. There is a greater demand on academic research for valid consumer behavior theory. The example given demonstrates a serious limitation when the predictive model is held up as the criterion of ultimate validity for consumer behavior research. A model is not a comprehensive theory.

For both the commercial and academic researcher the validity he looks for is the validity he finds. Both can profit by taking a second look.

----------------------------------------