Psychophysics: the Key to Real Product Differences Through the Consumer

Citation:

Russell A. Bell and John Rossiter (1971) ,"Psychophysics: the Key to Real Product Differences Through the Consumer", in SV - Proceedings of the Second Annual Conference of the Association for Consumer Research, eds. David M. Gardner, College Park, MD : Association for Consumer Research, Pages: 271-277.

Proceedings of the Second Annual Conference of the Association for Consumer Research, 1971     Pages 271-277

PSYCHOPHYSICS: THE KEY TO REAL PRODUCT DIFFERENCES THROUGH THE CONSUMER

Russell A. Bell, E. I. du Pont de Nemours & Company

John Rossiter, Associates for Research in Behavior

Multidimensional scaling techniques are enjoying increased attention in today's marketing research literature. But the fact is that many decisions in applied marketing -- a majority perhaps -- involve a single product attribute. More and more individual consumers whose spending dollars are being sought are becoming increasingly critical of not getting what they are paying for. And very often this criticism centers on unitary product-attribute. The use of unidimensional scaling techniques can protide information for decisions that will provide responsible products in the consumer marketplace. Accompanying the concern with product performance and safety standards has been a developing skepticism toward claims advanced in the advertising of many products. Manufacturers must guard the validity and believability of their advertising and ensure that their products meet the expectations of the consumer.

Along these lines, four general classes of questions typically confront production and marketing executives:

(a) How can I convincingly demonstrate that my product, with respect to characteristic X, is better than my competitor's product? This question is particularly pertinent to advertising which must meet FTC requirements.

(b) Given that I can offer a better product with respect to characteristic X, how far do I need to go, or indeed can I go, before diminishing economic and consumer satisfaction returns set in? This question is important from a cost standpoint and may also involve safety considerations.

(c) At the opposite end of the spectrum, how little of characteristic X do I need to include to be able to provide a low cost product which is still positioned within the boundaries of consumer acceptance? The consumer would thus receive a cheaper product without sacrifice of performance standards.

(d) Finally, how widely can my product fluctuate with respect to characteristic X and still be located within legally established limits and, more importantly, within the boundaries of performance expectations set by the consumer? This, of course, is the problem of quality control.

Judging from clients with whom the authors have collaborated, many executives are reliant on inefficient trial-and-error procedures for answering these types of questions. Yet the four cases outlined above are easily translatable into classic psychophysical terms. Given a product characteristic or attribute X measurable in physical units, what is the nature of the corresponding perceptual or psychological scale which will enable a determination of: (a) the magnitude of a real difference, observable by the consumer, with respect to various levels of the attribute; (b) the shape of the upper limits of the scale which shows how much of an increase in characteristic X can be tolerated and at what physical and thus monetary cost; (c) the shape of the lower limits Of the scale which shows the region of low cost acceptability; and (d) the range of physical variation in characteristic X which is within the range of perceptual or psychological nondetectability. Ignorance or incomplete knowledge of these parameters can result in many missed opportunities in the marketplace, chances to break through with different versions of an established product or to accurately position a new entry.

The purpose of this paper, then, will be to demonstrate how the executive can usefully apply psychophysics to answer questions of the foregoing types.

Assuming that the product attribute of interest is measurable in physical scale units, the first step is to select a method of deriving a correlated perceptual or psychological scale. Several established procedures are available, each being appropriate in certain circumstances. Applications of the three best known procedures are briefly summarized below:

TABLE

Many situations fall into the middle category: the range of product performance on characteristic X can be adequately represented by a dozen or fewer items with different values along the physical scale, the characteristic itself is fairly specific and easily understood by consumers, and the consumers (the ultimate Judges) do not require special training in order to evaluate the characteristic. The present illustration will thus utilize the paired comparison procedure. The technically oriented reader concerned about largely hypothetical objections to the method (Blankenship, 1966; Greenberg and Collins, 1966) is referred to an excellent discussion by Day (1966) advocating the advantages of the systematic probabilistic approach to paired comparison testing followed here. The data reported in this paper are actual; however a different (and fictitious) product is discussed for proprietary reasons. The applied aspects of the analysis could refer to any number of similar products or product attributes.

METHODOLOGICAL EXAMPLE

As an example of how psychophysical data can be used in marketing decisions, consider a decision making situation for the product manager of a spot remover. Let us assume for purposes of discussion that this is a well established market and one that is dominated by three or four major brands that contain only slightly more or less of a given active ingredient along with only slightly different combinations of carrying agents. Thus, the effectiveness of cleaning constitutes the major basis for product differentiation. This is a market environment that exists for a large number of products and decisions concerning an entry Of a new product or even re-positioning of an old product can be made more certain by psychophysical research.

For a spot remover product a typical experimental design would involve the preparation of ten experimental formulas that contained systematic variations in the active ingredient concentration with the current market situation located about midway along this range. The selection of this set of formulas would usually be aided by laboratory analysis and some product performance testing results. Carpet squares that were identical would then be prepared with a stain or soil pattern and then cleaned in a standard fashion. Each square would be cleaned by two of the spot remover formulas, one for each half. Thus, the end result would be 55 carpet squares that were divided in half. Two-hundred consumers would then be asked to Judge each patch as to which side was cleaner. The resulting data would indicate the proportion of times a given formula was Judged to out perform the others. Thus, a family of ten curves would be produced and represent discriminable differences at each solution concentration by employing the standard 75% point as the difference threshold. Figure 1 represents one of these curves for a middle range concentration as the standard for comparison. Table 1 represents the Just noticeable differences (JND) for all ten concentrations. These JND8 are the magnitudes of active ingredient increment that will be reliably reported by the consumer and they are expressed in terms Of ounces per gallon.

The blank spots in Table 1 indicate that a JND does not exist at these points, implying that beyond a certain active ingredient concentration any increases are not detectable by the consumer.

A second data analysis that would provide information for product decisions involves the calculation of constant errors for each solution concentration. This would be done by calculating the value for the solution that was reported as equal in performance to the standard and can be determined from the 50% point on each curve. Either a systematic or random bias on the part of the observer will cause equal products to be identified as different. If the data are systematically biased then the magnitude and direction of this bias can be considered. Table 2 shows the constant error for each of the ten solutions and in this case can be interpreted as random variations of very small magnitude.

FIGURE 1

THE PROPORTION OF TIMES A MIDDLE RANGE FORMULA WAS JUDGED TO OUT PERFORM OTHER MEMBERS OF THE SET

TABLE 1

THE JUST NOTICEABLE DIFFERENCES FOR THE TEN CONCENTRATIONS EXPRESSED IN OUNCES PER GALLON

TABLE 2

THE CONSTANT ERRORS FOR THE TEN CONCENTRATIONS EXPRESSED IN OUNCES PER GALLON

Returning to the original four questions that were posed by the product manager, It can be seen that answers are available from a rather simply executed procedure for each Of them. By locating the curve that most closely represents the concentrations of the current products, It is possible to determine if a claim of better than can be substantiated. From Table 1 it can be seen that a product with concentration of eight or less could have a better than claim leveled against it by a competitor. m at is to say, there is one or more products that could be perceived by the consumer as performing at a higher level on cleaning ability. A product having a concentration of nine or ten has no product that was Judged to be superior. Consequently if the current leading brand fell at level eight or below, a new product claim Of "better than" could be substantiated.

These same data also provide information for the second class of questions. Having located the maximum concentration at level nine and also positioned the current market status, the product planner has identified the flexible range of product modification. If the current market is located at seven or eight then very little discretion is permitted in selecting a new product.

Conversely, a product that could provide economic relief to the consumer along with minor decreases ln performance is a desirable alternative. In making the trade-off decision between economic and performance criteria the product manager would be able to determine the degree Of performance loss that would be associated with a reduction in active ingredient. If the current market leader is located at level six then it is possible to determine if level five is a noticeable drop in performance or not.

The final question of quality control is most easily answered by the data. If a product manager selects a product at any point within the range of the ten concentrations, the JND will indicate the outer limits Of equality. In practice, the quality control tolerances should be set well inside this limit since the arbitrary definition Of the JND is made at the 75% point.

The data from the constant error analysis would also bear on each of these questions; particularly the last. However, in this example there is no systematic CE and the random fluctuations are of such small magnitude that they cannot be interpreted.

DISCUSSION

It should be emphasized that the application of psychological methodology described in this paper is only one of a number that could have been selected. In particular, a variety of procedures exists for deriving a psychological consumer-perception scale to correspond with physical product qualities that can be controlled by the manufacturer. Once the psychological function has been mapped, the determination of product strategy is greatly simplified. Furthermore, the intent has been to show that psychological measurement can be easy to conduct, relatively inexpensive, and capable of producing definitive, readily interpretable information guidelines since the data are obtained directly from the consumer.

The application of such techniques is limited only by practical ingenuity, usually in circumstances where an immediate judgment of product performance is not possible. Nonetheless, tests could be designed to investigate more complex phenomena such as the pain-relieving ingredient in headache remedies, or even the engine-boosting qualities Of gasoline additives. Too often, it 18 argued, production and marketing executives guess at the outcome Of these factors and pay the toll Of falling products or missed opportunity in a competitive market.

Also, the attention Of academics in marketing research is drawn to the pervasiveness Of decisions involving single product attributes. Not everything is "complex and multidimensional", at least not to the manufacturer and not to the consumer. When many executives' technical statistical backgrounds need far too much polishing to read the average research article, for example, "cook book" descriptions of how to apply psychological methodology to answer basic but paramount questions of product performance should not be by-passed in the quest for more sophisticated but less readily interpretable solutions. Knowledge of the upper and lower limits Of acceptable product performance, the magnitude Of perceivable and demonstrable differences ln performance, and the physical cost function underlying the performance curve can be achieved quite readily with straight forward traditional psychological procedures.

REFERENCES

Blankenship, A. B. Let's bury paired comparisons. Journal of Advertising Research, 1966, 6, 13-17.

Day, R. L., Systematic paired comparisons in preference analysis. Journal of Marketing Research, 1966, 2, 406-412.

Greenberg, A. and Collins, S. Paired comparisons taste tests: Some food for thought. Journal of Marketing Research, 1966, 3, 76-80.

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