Market Research

I. Market AnalysisJohn E. Jeuck

BIBLIOGRAPHY

II. Consumer ResearchDik Twedt

BIBLIOGRAPHY

I MARKET ANALYSIS

Market analysis is concerned with predicting the size and location of markets as measured by sales. The term is sometimes used as a synonym for “market research,” which is more accurately defined to include the entire range of methods and techniques employed not only for the delineation of the sizes and shapes of markets but also for the evaluation of tactics and strategies useful for developing and exploiting markets.

While market analysis and advanced methods of sales forecasting derive from the relatively young fields of statistics, survey methodology, and econometrics, the practice of prediction—commercial prediction no less than any other kind—has a very long history. Before the Delphic liturgy was defined, the auguries were divined from the livers of bulls and the entrails of doves; and the ambiguities of the oracle may not have been less than some of the forecasts that today trumpet tomorrow’s fortunes.

The process and the outcome of sales forecasting command general interest among business managers, since sales volume constitutes the most important source of income for all but a narrow class of (financial) enterprises; and it is basic for corporate planning with respect to plant, personnel, and investment. Marketing plans and the allocation of promotional budgets are, of course, importantly influenced by expectations of the level and distribution of market opportunities.

The methods of market measurement In measuring markets, one is concerned with both industry and enterprise concepts—total generic product sales in an area, loosely referred to as “market potential,” and enterprise or brand sales, which are some proportion of industry sales. This is the “market share” of the firm or brand. The sales forecasts of particular companies normally depend on estimates of market potential (itself sometimes an inference from a general economic forecast), which are then adjusted for the firm’s expected market share.

Methodological discussions of market analysis usually focus on sales or market potential. In practice, of course, it is not always obvious what the most appropriate definition of “market” or “product” is. One speaks of markets in various contexts: the “gasoline market,” for example, of which the “high octane market” is a segment. The “petroleum market” and the “liquid fuels market” are broader concepts—but narrower than the “energy sources market.” Similarly, one often hears of the “high price” and “low price” markets, the “New York,” the “west coast,” and the “export” market. In estimating market potentials it is important to choose the relevant definition, but this choice is often influenced by the data available. [SeeMarkets AND Industries.]

While successful forecasting depends on gauging the effects of changes in the firm’s marketing policy, on competitive behavior, and on the consequences of future events that may transform the environment, business planning often starts from estimates of present (recent) sales. It is common practice to estimate potential market size by taking recent industry volume as the first approximation to the measure of expected market size. Even this is not always easy. Not only are there difficulties in establishing a relevant definition of “industry,” but data on recent industry sales are not as readily available as one might expect. Census data of various sorts offer authoritative figures on production and sales volume of fairly broadly defined product classes in the United States, but these tabulations do not appear at frequent intervals. More frequent compilations of industry statistics are by-products of excise taxes (e.g., gasoline and tobacco in the United States) and/or import duties. Other sources of industry sales-volume data are sometimes found in the publications and the archives of trade associations and of advertising media. Many firms find it useful to subscribe to one or more of the commercial subscription services that collect panel data from households and/or dealers for certain kinds of products. These panels provide much more detailed information on product movement—by brand, price level, and type of outlet, for example —than is available by other means. Finally, of course, it is possible to undertake special surveys of product volume and use.

In the absence of industry sales figures, and in order to avoid the expense of direct data collection, corollary or proxy data may be used to estimate market potentials. Data on complementary goods, for example, can serve this purpose—the market potential of automobile batteries for replacement can be estimated from data on the distribution of cars, sales of electrical appliances will be related to and limited by the number and location of wired homes.

Whatever the data sources of the market potential estimates, they are often available only on a national basis. One of the tasks of market analysis is to make the conversion to local and regional estimates. The conversion is characteristically made by distributing national figures in proportion to such measures as population, income, number of employees, and value added by manufacture, which are available for smaller geographic units.

The technology of sales forecasting It is customary in the literature of forecasting to distinguish between short-term and long-term forecasting, although there is only rough agreement on definitions. “Short term— clearly refers to the monthly or quarterly outlook and many practitioners would include the annual forecast. “Long term” can safely be attached to 10-year forecasts, but many technicians would so classify any predictions whose horizon is greater than one year. Forecasts for three years and five years are sometimes categorized as “intermediate range.” Much (most) of the technology of forecasting is common to predictions of the various planning horizons, although truly long-range forecasts (e.g., for 10 and 15 years) depend heavily on trend analysis of such fundamental variables as population, productivity, income distribution, political developments, and leisure—all of which change slowly and are assumed to be stable in forecasts for any period up to a few years.

While few analysts would enthusiastically endorse Alphonse de Lamartine’s observation that “History teaches everything, even the future,” the fact is that the experience of yesterday and today constitutes all we know about tomorrow. The technology of forecasting includes methods for identifying relevant historical data and for manipulating them in ways which may make the record more meaningful and illuminating. The art of forecasting largely consists in interpreting historical data and in specifying those future events that will condition tomorrow’s performance.

There are numerous ways of categorizing sales forecasting methods. It is convenient here to distinguish among the following: “judgment forecasts,” including sales-force estimates; market surveys; market tests; time series analysis, including curve fitting; and regression analysis. While these various techniques are sometimes presented as mutually exclusive and alternative analytical tools, in practice full reliance is seldom placed on any single method; two or more approaches often supplement each other. Certainly, the forecast ultimately accepted as the basis for company planning almost always will reflect the intuitive estimates of senior management.

Judgmental approaches. Hardly anyone will quarrel with the assertion that intuition and judgment dominate the practice of sales forecasting. Ease, cheapness, and flexibility all combine to support the use of predictions of sales volume which rest unabashedly on the judgment of one or more persons in the firm. Judgment forecasts, whether made by individuals or committees, are dominated by the experience, perception, and intuition of the forecasters. The process cannot be separated from the performer(s) and replication is impossible. However inelegant the method may appear, impressionistic evidence suggests that the more rigorous statistical forecasting methods continue to play a subsidiary role in business planning.

Judgment forecasts are sometimes based on sales-force estimates of future sales. It is hoped that sales personnel (and their managers) are familiar with the needs and circumstances of customers and the relative strength of competitive offerings. In some companies sales representatives systematically interview customers concerning their spending plans. Rarely, however, are sales-force estimates the sole basis of a company’s forecast. Substantial editing is generally contributed by corporate staff and senior management.

While it is often claimed that salesmen (and their supervisors) are unduly optimistic and therefore generate faulty forecasts, evidence is hard to come by. It is not obvious that such a bias should consistently attach to their estimates, since the estimators’ interest will be a function both of the character of the compensation plan and the particular use to be made of the forecast.

Market surveys. Whereas sales-force estimates rely on a survey of the forecasts of sales personnel, a more recent development has been that of the survey of buyer intentions. The market survey rests on a sample of potential buyers whose attitudes and/or intentions to purchase are solicited in person or by mail or telephone.

The survey has enormous appeal as a forecasting aid. It seems plausible that buyers should know (and be able and willing to say) what and whether they plan to buy in the future—especially if the future is not too distant. But this view assumes that consumers can predict what the future will be like as well as what they will choose to do if the future is as they expect it to be. Evidence on the validity of these assumptions is uncertain.

The Survey of Consumer Finances was initiated in 1946 by the Board of Governors of the Federal Reserve System (Juster 1964). These surveys, covering intentions to purchase durable goods (e.g., houses, automobiles, and major household appliances), are among the best-known and most thoroughly analyzed materials in the literature of forecasting. They are in many respects models of methodological nicety and generally far surpass ordinary commercial standards—and costs—of market investigation. Despite the relatively long history of these surveys and the superior methods and skills of the investigators, one must still characterize the method as more promising than certain. What does seem clear is that the survey method is best adapted to products involving relatively large expenditures and a fairly high degree of planning for purchase. In the case of family expenditures, homes and automobiles most clearly meet these conditions. In the case of industrial markets, expenditures for capital goods (e.g., expensive machinery and new plant construction) obviously qualify. Indeed, surveys of spending plans for industrial capital goods have tended to yield more accurate forecasts than those for consumer goods.

Market testing. Another method used to estimate market potential is the market test. Employed primarily for new products and invading brands, the market test is essentially an “experiment” which eschews direct questioning of respondents in favor of measuring their behavior. Market experiments are not feasible for all product categories, and they are expensive even when confined, as they usually are, to one or a few markets. The test is most commonly employed for packaged consumer goods items—the frequently purchased branded food, drug, and household supply products of relatively low price and high turnover. Market tests by and large amount to “tryouts” in one or a few markets—typically cities or metropolitan areas—which are selected as test locations because they are considered to be representative and relatively self-contained and have trade outlets and advertising media which are amenable to experimentation.

Virtually nothing is publicly available on the predictive value of the test market as a forecasting method. Test marketing is above all else a private affair. One infers that it provides useful, if not precise, information from the fact that successful and sophisticated enterprises continue to employ the method. At the same time, some conspicuous commercial failures suggest that test marketing is far from infallible as a basis for estimating market potential. In an article on the predictive power of test marketing, the author concluded:

Despite its shortcomings, test marketing has provided a certain degree of service to marketing management. It has, without question, considerably narrowed the range of new product sales prediction error over forecasts arrived at by judgment, even experienced judgment. From a management point of view, however, for a technique as expensive as test marketing, and one that exposes the company’s future hand to competitors, it would seem that we have a right to expect a much greater measure of reliability and accuracy. (Gold 1964, p. 16)

Time series analysis. Among the most frequently employed statistical forecasting techniques is that of time series analysis, which probes the record of past sales behavior in search of patterns that may be extrapolated. Historical sales, economic, and social data constitute time series which can be analyzed by statistical techniques of various kinds. But when measured against the task, the statistical tools are rudimentary and, some statisticians believe, of dubious value. One persistent approach seeks indicating or “leading” series which signal the direction and, hopefully, the extent of movement in the “following” or “lagging” series. While experience has dampened the hope that many companies can find consistently reliable lead-lag relationships, the search has not abated and empiricism rules.

Past sales patterns are sometimes mechanically projected by methods of curve fitting and trend analysis. The results are extremely sensitive to the particular formulas chosen and are perhaps most often used in combination with the forecasters’ best guesses about the state of key variables in the future and often with a substantial dash of reasoning by analogy.

Forecasting for the month or the quarter ahead is frequently characterized by relatively mechanical projections of recent sales, usually “smoothed” by one statistical device or another. The availability of computers and developments such as exponentially weighted projections of past sales, the weights declining as one incorporates earlier and earlier data, promise better short-term forecasts and improved inventory control in situations of highly complex inventory assortments and relatively strong seasonal demand.

In all forecasting it is well to have some criterion of success and, particularly, a measure of the extent to which forecasts do little but exploit the tendency for sales figures to show inertia, each period tending to be close to the experience of the immediately preceding period. This interest in a criterion underlies the so-called “naive models.” The simplest of such models is the “no-change” model, that is, the forecasted sales for the period ahead will equal the sales of the current period. A somewhat more complex formulation would be that the direction and rate of change, rather than the absolute value, will persist. (These are, of course, simple methods of extrapolation.) Naive models have been used mainly as standards for evaluating alternative forecasting methods. Nonetheless, the notion underlying naive models can be expressed in various autoregressive statistical methods that have some promise of improving judgmental forecasts, even for periods as long as a year.

Regression analysis. Regression analysis, in its many forms, attracts increasing interest and use. The underlying idea is to exploit the statistical relationship evidenced in the past between the variable to be forecast and the so-called independent variables (e.g., income, population, prices) that may be related to it. There are numerous regression analyses “explaining” the sales of various products, in the sense of showing a statistical relationship between sales and the independent variables. The prediction problem is then shifted from the dependent variables (sales) to the independent variable(s). Occasionally, of course, a lagged relationship can be shown to exist. Predicting from a regression analysis assumes stability of the past statistical relationships. As in all other statistical forecasting techniques, intuition and judgment are required to allow for changes in basic conditions that are neither reflected in historical data nor embraced by the statistical model.

The advent of the computer has made it possible to undertake more sophisticated (complex) manipulations of data than ever before. Computer programs make feasible the exponential smoothing of time series, stepwise multiple regression, the solution of systems of multiple regression equations, and simulation. But while the contributions of the computer promise wider use of complex forecasting techniques and offer improved opportunities for experimenting and testing alternative methods, the millennium is not yet.

Status and prospect. Despite the enthusiasm for sales forecasting and the exhortations that it constitutes the basis if not, indeed, the essence of business planning, the record is neither clear nor impressive. Accurate forecasting is still a wish rather than a fact. Distressingly little published information is available on the degree to which actual sales conform to forecast sales. One survey in the United States reported that there was an average deviation of 8 per cent between forecasts and actuals in 1955, but the average performance masks a wide range of deviations within and between industry groups (American Management Association 1956, p. 148).

In an English study Carter and Williams suggest some of the difficulty of making accurate sales predictions in the case of new ventures. In 57 per cent of the cases examined where expected and actual yields were at variance, the explanation was associated with “unforeseen changes of demand, or changes in the price of competing product.” They properly observed:

The importance of changes in demand can be seen. Many of these were changes in the demand for an intermediate product, deriving from a change in the nature of, or the demand for, some final products of other industries. Given the long period which must often elapse between the final decision to go ahead with a new plant and the sale of its first product, it is clearly unreasonable to expect many of these complex changes in demand to be foreseen; market research is not an answer to everything. (1958, pp. 90-91)

Limited information on European experience indicates mixed results, although the record of the “Munich Business Test” on quarterly forecasts of turning points is encouraging (see Theil 1958, chapters 4, 5).

There is an extensive literature on market analysis and sales forecasting, but most of it is preoccupied with technique and much of it is hortatory. And as Lorie observed a few years ago:

Progress comes most rapidly in any pragmatic discipline when adequate testing devices are available for measuring the success of current theories and procedures. Only by such testing is it possible to discard what has not succeeded and to cling to what has succeeded, for the purposes of further elaboration and refinement. These facts are considered self-evident in the field of meteorology, where most of the practitioners make their living by forecasting. As a consequence, a very extensive literature devoted to problems of evaluation has developed during the last seventy-five years. (1957, p. 177)

Unfortunately, the evaluation of sales forecasts is not general in the literature, and it is not common in industrial practice. Much too little is known about the predictive value of alternative techniques. The most sanguine would agree that there is much to be learned and that the task of prediction is to be approached with humility. Performance offers ample opportunity for improvement. The recording of methods and specific quantified forecasts with subsequent comparison with experience is certainly to be encouraged, not only as a means of educating technicians, but also as an opportunity for evaluating them.

John E. Jeuck

BIBLIOGRAPHY

American Management Association, Marketing Division 1956 Sales Forecasting: Uses, Techniques, and Trends. Special Report No. 16. New York: The Association.

Carter, Charles F.; and Williams, Bruce R. 1958 Investment in Innovation. Oxford Univ. Press.

Ferber, Robert 1960 The Railroad Shippers’ Forecasts and the Illinois Employers’ Labor Force Anticipations: A Study in Comparative Experience. Pages 181–199 in Universities-National Bureau Committee for Economic Research, The Quality and Economic Significance of Anticipations Data. National Bureau of Economic Research, Special Conference Series, No. 10. Princeton Univ. Press.

Gold, Jack A. 1964 Testing Test Market Predictions. Journal of Marketing Research 1, no. 3:8-16.

Hummel, Francis E. 1961 Market and Sales Potentials. New York: Ronald Press.

Juster, Francis T. 1964 Anticipations and Purchases:An Analysis of Consumer Behavior. Princeton Univ. Press.

Lorie, James H. 1957 Two Important Problems in Sales Forecasting. Journal of Business 30:172–179.

Mclaughlin, Robert L. 1962 Time Series Forecasting: A New Computer Technique for Company Sales Forecasting. Chicago: American Marketing Association.

National Industrial Conference Board 1964 Forecasting Sales. Studies in Business Policy, No. 106. New York: The Board.

Spencer, Milton H.; Clark, Colin G.; and Hoguet, Peter W. 1961 Business and Economic Forecasting: An Econometric Approach. Homewood, 111.: Irwin.

Theil, Henri (1958) 1961 Economic Forecasts and Policy. 2d ed., rev. Amsterdam: North-Holland Publishing.

II CONSUMER RESEARCH

Consumer research, or marketing research, is the systematic gathering, recording, and analyzing of data and problems related to the marketing of goods and services. The “first law” of marketing has been expressed as “Make what people want to buy; don’t merely try to sell what you happen to make.” This dictum reflects the existence of a society in which consumers can and do exercise considerable freedom of choice in purchasing.

Present interest in the consumer can be viewed against a historical background of concern with two other factors: production and sales. Concern with production grew in importance during the industrial revolution and reached a peak in the most advanced countries in the early part of the twentieth century with the introduction of assembly lines and mass production methods. Then, as manufacturing efficiency increased and unit costs decreased, the need to dispose of the fruits of mass production gradually shifted management’s attention to sales. Later, with the increase in competition for the buyer’s dollar within the market, sellers found that they had to take an interest in the consumer and, thus, turned to marketing research to find out what it is that people want to buy. By telling the seller what people want to buy, the researcher helps him plan more efficiently, avoid failures (thus lowering costs to consumers), and provide consumers with a much wider choice of products.

Definition The first and most important step in any marketing research undertaking is to define the problem. A clear statement of the problem is sometimes more than half the answer. Sometimes, too, after the problem is clearly outlined, it becomes obvious that the answer is already known or may cost too much to obtain. However, the cost of not having the answer must also be considered. If it is high, research may pay off. But if the price of being without the answer is minimal, marketing research may be uneconomical.

Once the problem has been defined and agreement has been reached as to what kinds of measurement will yield acceptable answers, the researcher designs his experiment. There are two basic ways of obtaining information about consumers: one is to ask them directly about their awareness, attitudes, and opinions; the other is to observe their actual behavior (other than verbal behavior). Depending upon the problem and the need for a precise answer, the opinion test (sometimes called “consumer jury”) may be entirely adequate. This approach is often used because it is quick, relatively inexpensive, and often provides useful guidance for marketing decisions. Its major drawback is that consumers do not always do what they say they will, for a variety of reasons.

Motivation research In an attempt to obtain better answers about why consumers act as they do, marketing researchers sometimes engage in motivation research (MR). The basic assumptions of MR are that people often have unconscious motivations and that indirect questioning and various projective devices may be more effective than direct questions.

In a study by Haire, for example, women were asked to describe a hypothetical individual solely on the basis of reading a grocery list prepared by that individual. There were two such lists, differing only in one item. One list included a reference to “regular grind coffee,” and the other list referred to “instant coffee.” Reactions to the second list showed that a significant number of women believed that the use of instant coffee characterized its user as “lazy.” Once the reason for the product’s lack of acceptance had been identified, the next step was to conduct an educational campaign, pointing out to housewives that by reducing food-preparation time, they could be with their families more—a use of the saved time that the experimenter assumed was consistent with the house-wives’values (Haire 1950).

Other examples of such techniques include:

(a) Word association. (“Tell me the first word that comes into your mind when I say ’margarine.’”)

(b) Incomplete sentences. (The respondent is asked to complete a sentence such as “The main reason my family doesn’t have soup more often is———”)

(c) Balloon Cartoons. (An example would be a comic strip drawing of two men in conversation. The balloon above one of the men might read: “John, I’ve been thinking about buying a station wagon.” The balloon above John’s head is blank, and the respondent is asked to suggest John’s response.)

(d) Narrative projection. (Character descriptions are followed by questions about the kinds of attitudes or purchasing behavior that might be expected from the individuals described.)

(e) Involuntary attention tests. (The respondent is asked to look at one of two stimuli, and the one he picks up and examines first is assumed to be more interesting.)

There is little doubt that for certain kinds of problems, and particularly for certain kinds of products and services about which people are sensitive (such as personal-care items or small loans), the techniques of motivation research are appropriate and even necessary. But to claim, as some proponents did during the 1950s, that these techniques are the only way to measure attitudes properly is to go to extremes. The violent arguments of that era between the quality-oriented, small-sample MR specialists and the more traditional, quantity-oriented, big-sample researchers (dubbed “mere nose counters” by the MR group) have largely ceased. It became clear that MR, although it contributed many lasting insights and methods that eventually influenced even the most traditional researchers, was not the ultimate key to a complete understanding of human behavior.

Sampling In most marketing research projects, once the problem has been defined, the next step is to decide how the sample of respondents is to be selected and how large the sample should be. It is important that the sample be free from systematic bias that could influence the conclusions. For example, a pre-election political survey of voter preference in upper-income suburbs would very probably be biased in favor of one political party; hence the sample findings could be quite different from the outcome of the actual election [seeSample Surveys].

In marketing research it is not always desirable to sample from the total population. For example, if we wish to study reactions to a brand of hair tonic, we would be wise to qualify respondents as tonic users before proceeding to interview them about their reactions. Going still further, we might want to talk only to heavy users. For a wide variety of frequently purchased products it has been shown that the 50 per cent of buyers who are heavy buyers account for 80 to 90 per cent of total sales volume (Twedt 1964).

Since about half the population uses hair tonic, a fourth of the total population (the top half of users) accounts for nearly 90 per cent of all hair tonic consumption. There is increasing evidence that this relationship is not limited to packaged goods; the same basic pattern of purchase concentration is shown in sales of gasoline to credit card customers, in tolls for residential long-distance telephone calls, and even in sales of fractional horse-power motors to industrial buyers.

Since consumption varies by individual and by household, it seems appropriate to let each respondent “vote with his pocketbook” by weighting his response according to the amount of the given product he consumes. Suppose, for example, that research has been conducted to determine whether a package for cake mix should have a red or a blue background for package illustration. From hypothetical data in Table 1, one sees that although the total sample gives a majority vote to the package with the red background, the correct marketing decision would be to choose the blue background, since it is clearly more appealing to the heavy users, who account for about 85 per cent of total cake mix purchased.

Questionnaires After the necessary decisions have been made as to the purpose of the survey and who the respondents will be, it is necessary to make up a questionnaire. Questionnaires are often designed with a few general, bland questions at the beginning in order to establish rapport between interviewer and respondent. It is essential to ask the questions in such a way that the answer to a given question does not bias succeeding questions. For this reason many researchers build questionnaires by writing each question on a separate card, moving the cards around until a smooth sequence is achieved, and then transferring the questions to a regular questionnaire format. In major surveys the questions are “pretested” on a small group of 25 to 50 respondents, whose answers are not included in the final tabulations but are analyzed to see if the questions are clear and unambiguous. Revisions are made, and the questionnaire is again pretested. It is not unusual for a survey to go through as many as a dozen revisions before the questionnaire is finally approved (Payne 1951).

Preceding and tabulation The next step is to precode the questionnaires by assigning to, and printing next to, each question the appropriate code for machine tabulation of responses. Questionnaires can now be designed for optical scanning by machine, with the answers converted directly to either punched cards or magnetic tape; the cards or tape can be programmed to produce a page of “print-out” that is used as photographic copy for the final research report. The enormous economies in human time and the accompanying error reduction through elimination of tedious copying and hand tabulation and computing make it obvious that these developments will continue. The great potential economies of the computer in marketing research are not its only value; a beneficial side effect is that in order to use the computer to its maximum efficiency, the entire research project, including the final tabulations, must be thought through in advance, and this usually results in uncovering mistakes in planning at a time when they can be more easily corrected.

Quality control in field interviewing Questionnaire surveys are usually made by interviewing organizations that specialize in this type of work. A list of such firms, and of others that offer a broader variety of services, including consulting, experimental design, and preparation of the final report with recommendations for action, is contained in Bradford’s Directory of Marketing Research Agencies, which lists 350 firms in the United States and abroad (Bradford 1965–1966).

After the interviewing organization has been selected, it is customary to hold a briefing session with the project leader, the interviewing supervisors, and the interviewers in order to give detailed instructions on how the interview is to be conducted. Any questions the interviewers have should be answered at this time. Often they will be told not to proceed after the first day of interviewing until the field supervisor has had an opportunity to review the quality of work done during that day.

As the field work is progressing, or immediately after it is completed, 10 to 20 per cent of the interviews are validated in one of several ways: reinterview, a telephone call, or by mail. During these follow-up procedures, the supervisor verifies that the interview actually took place, and a few key questions may be repeated to check for report consistency. When the work of a given interviewer seems irregular, all the interviews turned in by that person are rechecked. If, as occasionally happens, it develops that the interview was not made and that the questionnaire has been faked, the interviewer is not employed again by the same organization. One of the weakest links in the entire data-gathering and data-processing chain is the failure to validate questionnaires properly [seeQuality Control, Statistical].

Recommendations for action If the objectives of a research survey have been clearly denned, the right questions asked of the right people in order to answer these objectives, and the field work properly supervised, the two remaining tasks—tabulation and analysis, and report preparation—should be fairly routine. There are, however, two major conflicting viewpoints about the extent to which the researcher should make firm recommendations for marketing action.

One group of marketing executives tends to regard the researcher primarily as a technician from whom only “the facts” are wanted, with no interpretation or conclusions about appropriate courses of action. The other viewpoint is reflected by Theodore Levitt of Harvard University, who says,

Expertness [in marketing research] encompasses much more than the elaboration and use of formal techniques in research and analysis. More than anything else it should be viewed as involving imaginative audacity in the interpretation of data and events and in formulating positive action-oriented proposals for management’s consideration. … Too often nothing is permissible in the way of making policy or entertaining ideas unless the data are so unambiguously in favor of proposed policies or ideas that even the elevator operator can see their merit. (1962, pp. 187-190)

Observing behavior Most of the discussion up to this point relates to the gathering of opinions through survey research. There are many other ways in which marketing research can be conducted. Observing and recording consumer behavior takes many forms. Legibility tests of an advertisement, for example, can be made by determining the amount of illumination or length of exposure required to allow reading of the sales message. It is even possible to measure behavior which the respondent is not consciously aware of, such as pupillary dilation while viewing a product or an advertising stimulus. Or suppose that we wish to determine the proportion of consumers that will read the statement of ingredients on the package label of a new food product. It is a simple matter to observe shoppers at the moment of buying and record the number of seconds they examine each package before selecting it.

Another major method of behavioral research in marketing is to conduct controlled advertising and sales tests in selected areas. Two cities may be matched, for example, on the basis of their previous sales of a given product. An advertising campaign may then be undertaken in one city, and a different campaign (or perhaps no advertising) employed in the other city. After a predetermined time period sales results are compared for the two cities. Variations of this method may include much more complex statistical procedures and may involve factorial and Latin-square experimental designs, with many cities or retail outlets included in the test. The measures may be actual sales, the proportion of consumers who are aware of a given brand name, or the proportion who can recall specific sales points about the product.[SeeExperi-Mental Design.]

Subjects for marketing research The most common research activities of 1,660 companies are shown in Table 2.

Advertising research Advertising research is a special application of marketing research but employs many of the same techniques and methods. The task of advertising research usually is to find answers to one or more of the following questions:

(1) What shall we say? (2) How shall we say it?

(3) Where, when, and how often shall we say it?

(4) How well did we communicate the intended message? These questions are investigated, respectively, in motivation research, copy research, media research, and evaluation research [seeAdvertising, article on Advertising Research].

Growth of marketing research departments Of 1,660 companies surveyed more than half reported having a marketing research department (American Marketing Association 1963). Research departments are most common among companies that manufacture consumer goods (62 per cent have them), industrial companies (60 per cent), and publishers and broadcasters (57 per cent). The bigger the company, the more likely it is to have a research department. The research department is a fairly recent addition to corporate staffs; more than half of the departments in the survey had been formed since 1955. Even the companies that do not have formal marketing research departments carry on such research, either through their own personnel or through outside consulting firms. Between 1960 and 1965 marketing research budgets as a percentage of sales increased for both consumer and industrial companies.

It is clear that marketing research has matured greatly since 1955 and that it has gained increased acceptance from business management. It is also true that marketing research has not yet reached its full potential.

Dik Twedt

[Directly related are the entries Advertisingand Consumers. Other relevant material may be found in Interviewing, article on Social Research.]

BIBLIOGRAPHY

Readers interested in more detailed consideration of the various aspects of marketing research are referred to Wales & Ferber 1956, an annotated bibliography of more than 1,600 references covering 28 major areas. For current developments in the United States and abroad, see the three leading professional journals: Journal of Marketing, Journal of Marketing Research, and Journal of Advertising Research.

American Marketing Association 1963 A Survey of Marketing Research: Organization, Functions, Budget, Compensation. Edited by Dik W. Twedt. Chicago: The Association.

Bradford, Ernest S. (editor) 1965–1966 Bradford’s Directory of Marketing Research Agencies and Management Consultants in the United States and the World, llth ed. Middleburg, Va.: Bradford.

Haire, Mason 1950 Projective Techniques in Marketing Research. Journal of Marketing 14:649–656.

Journal of Advertising Research. ? Published since 1960.

Journal of Marketing. ? Published since 1936.

Journal of Marketing Research. ? Published since 1964.

Levitt, Theodore 1962 Innovation in Marketing: New Perspectives for Profit and Growth. New York: McGraw-Hill.

Payne, Stanley L. 1951 The Art of Asking Questions. Studies in Public Opinion, No. 3. Princeton Univ. Press.

Twedt, Dik W. 1964 How Important to Marketing Strategy Is the “Heavy User”? Journal of Marketing 28, no. 1:71-72.

Wales, Hugh G.; and Ferber, Robert (1956) 1963 A Basic Bibliography on Marketing Research. Chicago: American Marketing Association.