Note from the author: This is an example of bad prediction on my part. I thought that the concept behind "fuzzy logic" would quickly invade the fields of advertising and media research. So far, I have been wrong. If you have a prospective viewpoint on this, I'd like to discuss it.

(Click here if you wish to subscribe to our free mailers.)

by Jacques R. Chevron

(As published in the Viewpoint: Forum section in Advertising Age, Nov. 25, 1985)

Our every-day language is notably imprecise. We avoid absolutes and communicate by giving our audience a direction, a sense of what we mean. But it works! When we say: "This bald man is a frequent smoker," every one understands: "Bald" refers to a quantity of hair somewhere between "balding" and "completely hairless"; "frequent smoker" is less than "chain smoker" and more than "occasional smoker." To better communicate our thoughts, we often confuse things a bit further: Saying a man is "bald" or a "frequent smoker" seems too absolute a characterization. He will, most likely, be "quite bald" and a "rather frequent smoker." The vagueness of qualifiers "quite" and "rather "doesn't hinder communication.

Imprecise language can best reflect human reality.

Yet, when marketing began using computers, we were forced to precisely define the terms we use in order to make them acceptable to the computer. Thus a "frequent cigaret smoker" is defined, let's say, as one who smokes one pack or more per day, and a "bald man," as one having 5,000 hairs or less. In this precise world, the poor man who loses just one of his 5,001 hairs becomes instantly bald. The one who regularly smokes 19 cigarets every day is not a frequent smoker.

A new mathematical tool is now available, which could help change all that: Introducing FUZZY LOGIC. The brainchild of Mr. Lofti Zadeh (the former chairman of the electrical engineering and computer sciences department at the University of California at Berkeley).

To start with, a Fuzzy Logic definition of a frequent smoker would be "smokes about one pack or more a day." Then, rather than categorizing a given smoker as "frequent," the Fuzzy Logician would allocate it a coefficient (between 0 and 1) to indicate how well the smoker can be identified with the "frequent smokers" set. With 19 cigarets per day, the coefficient could be .75; with 25 cigarets, .95, and so on.

In other words, Fuzzy Logic reflects the way most of us speak, allowing for all the qualifiers one may use when talking about someone as a frequent smoker. When saying, "he's a rather frequent smoker," we do not give "rather" a meaning related to a certain number of cigarets. We give our impression of how well the subject belongs to the group of "frequent smokers." We use "rather" as a fuzzy coefficient.

This theory has wonderful applications to marketing and advertising. It can enable the analysis of consumer answers, as imprecise as they naturally come. It should put an end to the "women 18 to 49" target audience definition, which has no reason for being in fuzzy language (they'd probably become "young and middle-aged women"). It should improve our analysis of consumer emotions, where the language is often purposefully rendered vague with a plethora of qualifiers: In Fuzzy Logic, "extremely" can multiply "very" and one can take the square root of "few" or solve a problem like "if most student are rather frequent smokers, how many smoke frequently?"

The answer is "most" multiplied by "rather", of course.

No, I don't smoke -- not even funny cigarets.#