Awareness As an Indicator of New Product Performance

David W. Olson, Leo Burnett U.S.A.
[ to cite ]:
David W. Olson (1975) ,"Awareness As an Indicator of New Product Performance", in NA - Advances in Consumer Research Volume 02, eds. Mary Jane Schlinger, Ann Abor, MI : Association for Consumer Research, Pages: 495-506.

Advances in Consumer Research Volume 2, 1975      Pages 495-506

AWARENESS AS AN INDICATOR OF NEW PRODUCT PERFORMANCE

David W. Olson, Leo Burnett U.S.A.

[The author gratefully acknowledges the assistance and support of Dr. Joseph T. Plummer and Mr. Frank Feinberg of Leo Burnett U.S.A. in preparation of this paper.]

[David W. Olson is Associate Research Supervisor, Special Task Force, Leo Burnett U.S.A.]

Since advertising weight for new products does not directly relate to trial levels achieved by those products, awareness is postulated to be a useful intervening variable between advertising weight and trial. Aided brand awareness is found to be a superior measure to unaided brand awareness in this application. At the same time, aided brand awareness is subject to problems of "false" awareness and of bias due to question order, and is a generally unstable measurement. Therefore, care must be taken to avoid designing models whose output is highly sensitive to varying inputs of awareness.

Traditionally, an integral part of any research package measuring performance of a new product in a test market is the A-T-U (Awareness-Trial-Usage) study. Awareness of the new product is generally considered to be one relevant indicator of performance for a new product. And yet, the precise nature of these variables rarely seems to be fully understood by those using awareness to evaluate the new product's performance.

The purpose of this paper is to clarify the nature of the measurement of awareness of new products, to specify the usefulness of such measures, and to caution against their misuse in certain applications.

THE MODEL

The basic form of various models which simulate new product performance is:

FIGURE

There is, of course, no simple relationship between promotion effort and sales for a new product. That is, what happens in the "black box" has a major effect on sales levels for any particular new product, which far outweighs the unitary effect of promotion effort.

The first step in expanding the model is, of course, to break "sales" into its components of trial and repeat purchase (usage). It is quite well accepted that repeat purchase is primarily a function of satisfaction with the product by triers, and that promotion effort only plays a secondary role in generating repeat purchase. Therefore, the most that can be hoped for is some relationship between promotion effort and trial levels achieved by new products.

Let the model be further restricted by considering one important aspect of promotion effort, namely a measure of advertising weight in television. [The reason for choosing TV advertising weight is that its effects are usually large in comparison to other forms of promotion effort (|.e. most of the marketing dollars are spent in TV). Since the model described here includes only one input, it is natural to use TV GRP's as that input. For products which are heavily supported by alternative advertising vehicles (e.g., print or consumer promotion), this formulation of the model is inappropriate.] Does there exist a relationship between TV advertising weight and trial levels achieved by new products? [Trial throughout this paper will be reported among each product's "potential users." This permits looking Jointly at products in many different categories, since their trial levels are "normalized" by basing them on their own potential market.]

FIGURE 1

RELATION BETWEEN CUMULATIVE GROSS RATING POINTS ON TV AND TRIAL FOR NEW PRODUCTS AMONG POTENTIAL MARKET IN FIRST THREE MONTHS.

As Figure 1 shows, little variability in trial levels for different new products is accounted for by a measure of advertising weight (measured in terms of cumulative TV Gross Rating Points since introduction of the new product, confined to its first three months). The explanation is obvious: too many important factors in addition to ad weight (e.g. price, distribution, appeal of product concept) significantly affect trial for the relationship to hold.

The model is, therefore, of no predictive or quantitative use in this form. What is needed to improve it is some intervening variable, which relates well to both ad weight and trial, independently. Some measure of awareness of the new product seems a likely candidate. Intuitively, one should expect awareness to be significantly affected by advertising weight, and similarly, one can postulate that awareness relates, in some sense, to trial (if only as a necessary prerequisite to trial). The model would then look like this:

TV GRP's -> AWARENESS -> TRIAL

It is important to note that awareness, used in this way, is not a behavioral measure. It is nothing more than an artificial variable which, it is hoped, relates quantitatively to television GRP's and to trial. The balance of this paper will explore "awareness" as it relates as an intervening variable within the simple model described above.

MEASUREMENTS OF AWARENESS

Awareness is far too often thought of as part of some sort of bipolar process, where consumers are either "aware" or are not. It is assumed that after repeated exposure to some stimulus, such as advertising, at some point consUmers move into a "state of awareness" from the state of total ignorance they were in previously.

Unfortunately, any thought on the matter at all will soon lead to the conclusion that awareness is, in fact, a continuum. In terms of a new product, awareness of that product may build from a dim recognition of some of the narrative elements in its commercial, through to brand name recognition, through to a detailed understanding of what the product is and how it may be personally beneficial. Section III will discuss a frequently employed method of incorporating awareness into models of consumer behavior which, in essence, violates the assumption that awareness is a continuum.

Given that awareness is a continuum, how does one measure it? One must understand that whatever they are measuring they are, by nature of their measuring tool, artificially dividing the continuous world into those "aware" and those "not aware."

Two measurements of awareness which have traditionally been obtained in studies are unaided and aided brand awareness. These two measurements will be discussed below, and evaluated on the basis of how each works as an intervening variable in the model which has been described.

Unaided Brand Awareness

Unaided brand awareness is the traditional measure of awareness for established brands. Countless studies have shown this measurement to be sensitive to changes in advertising weight for ongoing brands (e.g. Palda, 1969) and other studies have shown some relations between unaided awareness and purchase behavior (e.g. see Axelrod 1968, Assael and Day, 1968, Gruber 1969). It is, therefore, reasonable to explore how well it relates to advertising spending for new products.

FIGURE 2

RELATION BETWEEN CUMULATIVE GROSS RATING POINTS ON TV AND UNAIDED BRAND AWARENESS FOR EIGHT NEW CEREALS.

Figure 2 represents the relationship between cumulative GRP's since introduction and unaided brand awareness achieved by eight new cereals. As can be seen, the relationship is poor. It appears that the unaided awareness level obtained by a new cereal is more dependent on other factors than on the advertising weight spent behind that new product, and hence is not a good measure to incorporate within the model as it has been outlined.

There are three major reasons why the measurement of unaided- awareness is unsuitable for new products as an intervening variable. First, new products generally attain rather low levels of unaided awareness, which makes discrimination between different brand performances somewhat difficult. Second, the numbers obtained are sensitive to how the question is asked. If asked to name all brands of cereal they can think of, a significantly smaller percentage of respondents will name a particular new cereal than if they were asked to name all the new cereals they can think of. Standardization of the question, especially across categories, becomes a hazardous task. [This, of course, leads to another methodological problem with unaided awareness, which is: what does one do with a new product which does not fit cleanly into an existing product category?] Finally, unaided awareness is probably a composite measure of a combination of other performance factors, such as recency of trial or usage. That is, unaided awareness may, to a sizeable extent, be generated by events which occur after trial of the new product, which is contrary to the implied causal structure of the model.

For these reasons, then, unaided awareness appears to be a poor measurement of awareness for new products to use in the model. Unaided awareness is not a useful intervening variable in this context.

Aided Brand Awareness

Aided brand awareness ("Have you ever heard of _____?"), to a large extent, overcomes the three problems of unaided awareness. It can be asked consistently, across product categories, and for products which have no relevant category; the magnitude of the numbers obtained are sufficiently large to provide a possibility for discrimination; and it is not unduly affected by events which occur after trial of the product. [This last point is not strictly true, of course. Some people can and do become aware of a new product (as measured by aided awareness) only after trying the product, which violates the causal structure of the operational model. The magnitude of this effect, however, is usually small enough to be negligible.] At the same time, it has other problems, to be discussed shortly.

But first, the question is: Does aided brand awareness relate to advertising spending? The answer is yes; at least, it relates far better than unaided brand awareness.

Figure 3 is the relation between cumulative GRP's spent behind products and the aided awareness levels attained by them in the first three months after introduction. While the relation is by no means clean, the curve can be reasonably represented by an exponential curve

Yt = a + (1-a) (1-e-bSt)      0 < a < 1    (1)

where

Yt = Awareness at time t

St = Cumulative GRP's at time t

a = Level of awareness attained prior to TV advertising

b = Rate of increase of the awareness function

This curve has a number of interesting features: 1) aided awareness is naturally bounded above by 100%, which products approach very slowly. That is, there are diminishing returns on awareness for increased advertising weight. 2) The Y- intercept is positive, in the range of 10-20%. This means that given zero TV GRP's spent behind a new product, one usually obtains 10-20% aided awareness for that product. This is probably due to two main effects: sell-in to stores before the start of advertising, where some consumers may become aware of the brand by seeing it in the store (probably a small effect); and, more importantly, "false" aided awareness for the new product.

FIGURE 3

RELATION BETWEEN CUMULITIVE GROSS RATING POINTS ON TV AND AIDED BRAND AWARENESS

Anyone who measures aided awareness of brands is familiar with the concept of "false" awareness. It is assumed that a certain number of respondents who say they are aware of a brand are not "really" aware of the brand in question. Operationally, "false" awareness for a brand can be defined as that level of aided awareness obtained for the product with no advertising, no distribution, and no other marketing effort of any kind.

The whole issue of "false" awareness is the biggest problem with aided awareness, and points to the overall fragility of that measurement of awareness. If the amount of "false" awareness were constant from brand to brand, then there would be no problem, since the curve in Fig. 3 would already incorporate this constant "false" awareness. Data at hand, however, suggest that many brands differ significantly in their levels of "false" awareness.

FIGURE 4

RELATION BETWEEN CUMULATIVE GROSS RATING POINTS ON TV AND AIDED BRAND AWARENESS FOR ALL PRODUCTS AND FOR PRODUCT CATEGORY A.

Figure 4 displays the curve (generated by equation [1]) which best fits the relationship between cumulative GRP's and aided brand awareness, both among all products as a whole, and among products within one particular categorY. This category is composed solely of new products made by a well-known food producer which are all similar in nature to existing products made by that producer, and which all carried the producer's widely familiar corporate name. Under these conditions, it might well be anticipated that potential exists for substantial "false" awareness of these new products. Figure 4 shows that these specific products do achieve higher levels of aided awareness than most new products (given their spending), and it is reasonable to believe that this difference is primarily due to unusually high levels of "false" awareness for these products. A second indication of the fragility of the measurement of aided awareness is that results can be very dependent upon question order for certain products, especially when "false" awareness exists for those products.

An experiment on question order was conducted while tracking the introduction of a new personal care product. It was known, going into the tracking study, that "false" awareness would probably be a major problem in interpreting the brand's performance, since its name was very similar to two existing major brands in the category.

The experiment involved rotating the order in which respondents (male product category users) were asked aided awareness of four brands. Two of the brands(A and B) were well-established, high-volume products (with product A the bigger seller), while the other two (C and D) were new products which had similar names to products A and B. It was hypothesized that the levels of aided awareness measured for brands C and D would be higher, due to higher "false" awareness, when they were asked first, before the well-known brands A and B, than when they were asked after brands A and B. It was assumed that consumers who confused the new brands with the old brands would be "helped" by hearing the names of the old brands first.

Table 1 shows the results of this experiment. As expected, the highest levels of aided brand awareness were obtained for the two new products when they were asked first, before the well-known brands were asked, and this difference was significant in magnitude (10-12%3. Interestingly, the established products did not show a difference in aided awareness depending on whether they were asked first or last.

TABLE 1

AIDED AWARENESS, BY ORDER OF ASKING

This experiment casts great doubt on the usefulness of the measurement of aided awareness, since it appears to be a loose, unstable measurement. Given these problems, how useful is it? The relationship in Figure 3 suggests that it is a useful measurement despite these problems. The effects of these problems do not appear to be so large that general predictive relationships between advertising spending and aided awareness cannot be made. At the same time, it must be understood that aided awareness is not a firm, concrete measurement, and should not be treated as such. The second question to be answered in evaluating aided brand awareness as a relevant measurement is how well does aided brand awareness relate to trial of new products?

Figure 5 shows the relation between aided brand awareness of new products and their trial levels ("normalized" by being computed among potential market for those products) during their first three months after introduction. While again the relationship is by no means perfect, it does appear that a definable relationship exists. Blattberg and Golanty (1974) have quantified this relationship to be

Tt = Tt-1 + < (At-At-1) + B (At-1 - Tt-1)

where Tt is trial in period t, At is aided brand awareness in period t, and < and B are parameters.

FIGURE 5

RELATION BETWEEN AIDED BRAND AWARENESS AND CUMULATIVE TRIAL FOR NEW PRODUCTS IN FIRST THREE MONTHS.

It thus appears that the measure of aided brand awareness is a useful intervening variable between marketing effort (TV GRP's) and trial levels attained by new products. Those relationships, however, are not predictive. Any specific new product may get much higher awareness levels than the "average" new product (because it has, for example a commercial with more attention-getting power than-the "average" new product commercial). But the relationships are useful in predicting potential ranges in performance, given varying levels of advertising weight. When used in this way, aided brand awareness is very definitely a relevant indicator of a new brand's performance, relative to other new products in the past.

MISAPPLICATIONS OF AWARENESS

A whole genre of models designed to predict sales for new products prior to test market involve showing respondents a stimulus (usually a commercial or concept board for the new product) and then measuring, in various ways, trial of the new product. That measure of trial is then factored down for awareness. It is assumed that by showing the stimulus to all respondents, 100% awareness is generated, and that in the real world, 100% awareness will not be obtained for the new Product.

While it is clear that any measured trial obtained under these conditions must be factored down to correct for lower real-world awareness, there are some major theoretical problems with the ways this adjustment is typically made.

Most frequently, the adjustment is of the linear form

TA = T x A   (2)

where TA is the adjusted trial estimate, T is the measured trial, and A is some measure of expected awareness (usually aided brand awareness).

There are two basic problems with the use of awareness within this adjustment procedure. First, as has been seen, every measurement of awareness has some significant limitations. Unaided awareness does not seem to be very predictable in terms of advertising weight, and aided awareness is a quite fragile and unstable measurement, subject to problems of "false" awareness and question order. It seems at best unwise to design a model whose prediction of sales varies linearly with variables as uncertain as these.

The second problem with this procedure is philosophical in nature. Which measure of awareness should be used in (2) -- unaided or aided, brand or advertising? Vastly different levels are obtained on each measure for any given product. Which one is proper? In most models of this nature, aided brand awareness is chosen. On what grounds?

Theoretically, the answer is that one must estimate the percent of the population which will be in a state of awareness equal to that state of awareness generated in the artificial forced-viewing laboratory situation. What might this measurement be?

It could be aided brand awareness. Certainly there is 100% aided brand awareness generated in the laboratory test. But it might also be aided advertising awareness, which is also 100% in such tests. Or perhaps unaided awareness, which might be 100% in such tests. Or even an awareness of "poorer quality" than aided brand awareness, such as recognition of narrative elements in the commercial. The central question is why should one apply aided brand awareness estimates to correct for 100% "awareness"7 If awareness is indeed a continuum, as suggested earlier, then any measurement designed to divide the world into those "aware" and those "not aware" is entirely arbitrary. No one measurement is "magical" in capturing the essence of awareness.

If one is going to use "awareness" in these pre-test models, the measurement of awareness used must meet the following criteria:

1) It must be estimatable from the marketing plan, within a narrow error range;

2) It must have a definable relationship to trial in the real world;

3) It must be relatively stable, and not sensitive to variability due to noise (necessary for 1) and 2) to be fulfilled); and

4) It must not be highly correlated with those variables which one is trying to ultimately forecast.

The available data on unaided and aided awareness suggest that neither measurement fulfills 1) - 4) above.

The ultimate test of models which hang on this use of the concept of awareness is, of course, whether they work. On theoretical grounds, at least, their potential appears highly doubtful.

SUMMARY

Awareness has been shown to be a useful intervening variable between advertising weight and trial for new products. Aided brand awareness is superior to unaided awareness as an intervening variable. At the same time, it must be remembered that awareness is a continuum, and that any measure of it is an artificial dichotomization of reality. Given this fact, as well as the difficulties in actually measuring awareness, the use of awareness to factor down simulated trial measures to correct for real-world awareness is a highly hazardous procedure.

REFERENCES

Assael, H. & Day, G.S. Attitudes and awareness as predictors of market share. Journal of Advertising Research, 1968, 8 (4), 3-10.

Axelrod, J.N. Attitude measures that predict purchase. Journal of Advertising Research, 1968, 8 (1), 3-17.

Blattberg, R. & Golanty, J. An early test market forecasting model for new product planning. Working paper, 1974.

Gruber, A. Top-of-mind awareness and share of families: An observation. Journal of Marketing Research, 1969, 4, 227-231.

Palda, :K.S. The measurement of cumulative advertising effects. Englewood Cliffs, W.J.: Prentice-Hall, 1964.

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