This Document Contains Chapters 20 to 25 PART IV TEACHING NOTES FOR CHAPTERS CHAPTER TWENTY DISCRIMINANT, FACTOR, AND CLUSTER ANALYSIS Learning Objectives • Discuss the concept of discriminant analysis. • Discuss the concept of multiple discriminant analysis. • Discuss how to interpret the results of the discriminant and multiple discriminant analysis techniques. • Discuss the concept of factor and cluster analyses. • Describe business applications of factor analysis. • Discuss the potential limitations (through violations of the assumptions) of the Factor analysis technique. • Describe business applications of cluster analysis. • Discuss the potential limitations (through violations of the assumptions) of the cluster analysis technique. Teaching Suggestions The text has been written to be as much accessible as possible. However, owing to the nature of the techniques, it may be difficult to reach out to the students. The instructor can begin by discussing the need for Discriminant Analysis. Emphasis can be on the need for a method that considers the influence of a number of variables simultaneously. This can be used to motivate the students to learn this technique. The instructor can go through the basics using the example provided in the text. The class discussion can center around the following: the objectives of the technique, interpretation of the discrimination coefficients, classification of objects using the discriminant function and the use of the confusion matrix. Considerable time should be spent on the use of discriminant analysis in the marketing applications. The part on cluster and factor analyses is written to be very accessible. It really requires only an understanding of what a correlation is. Thus, it could be the only chapter assigned beyond Chapter 18 or one of two or three. Despite the fact that the chapter is accessible, factor analysis will be new to everyone, even those who have had some statistics including regression. There are several concepts that will be new and slippery such as factor, factor score, factor loading, variance explained, communality and varimax rotation. It’s usually very worthwhile to go through the chapter presenting the concepts in lecture form and making sure the students are comfortable with them. The discussion questions can be used in this context. The factor analysis presented in Figure 20-5 is interesting and the study on which it is based might be worth additional discussion. It should be emphasized that factor analysis is a highly subjective process and the various steps involved in a factor analysis involve subjective judgment. Also, it is to be made clear to the students that there are no statistical tests employed in factor analysis and hence it is difficult to know if the results convey anything meaningful. Therefore, it would be in the best interests of the researcher to divide the sample randomly into two or more groups and independently run a factor analysis with each group. If the same factors emerge, then the researcher can be confident of the results. Cluster analysis will also be new to the students. It is better to go through the chapter in a lecture format and then supplement the class discussions with discussion questions. Cluster analysis can be a practical tool. In a segmentation study, for example, the segments need to be defined on the basis of homogeneity of benefits sought, perceptions held, or lifestyles. Second, it gets the reader thinking about combining variables to form constructs. It differs from factor analysis in that the context is usually people or objects instead of variables and because the method is more direct and therefore transparent. Third, by introducing the concept of quick clustering and by providing opportunities to practice it, the reader will get a feel for correlation (or association) matrixes and with looking for relationships instead of just looking for the nearest computer. Among the important cluster analysis concepts are those of similarity measures and the level of clustering (the number of clusters). Another point to be made is the relationship between duster and segmentation research. Segments, like benefit segments, can be created via cluster analysis and used as dependent variables (and independent variables) in subsequent analyses. Questions and Problems 1. With the number of groups (m) being three and three independent variables (p), the number of discriminant functions that can be generated is given by the minimum of (m-l, p) and which is two. However, since all of the (m-l) axes may show statistically significant variation among the groups and hence fewer than m-l discriminant functions may actually be needed. 2. Discriminant analysis involving deriving the linear combination of the two or more independent variables that will discriminate best among a priori defined groups. This is achieved by the criteria of maximizing the between group variance relative to within group variance. The linear combination for a discriminant analysis is derived from an equation that taxes the following form: Z = b1 X1 + b2 X2 + b3 X3 + ... + bn Xn where Z = Discriminant score b = Discriminant weights X = independent variables The value of the discriminant weights or coefficients for a particular predictor depends on the other predictors included in the discriminant function. The signs of the coefficient are arbitrary but they indicate which variable results in large and small function values and associate them with particular groups. Predictors with large standardized coefficients contribute more to the discriminating power of the function. The relative importance of the predictors can also be obtained by examining the structure correlations also called as discriminant loadings. These simple correlations between each predictor and the discriminant function represent the variance that the predictor shares with the function. However, both the loadings and the weights need to be interpreted with caution. Unless the sample size is large, both the loadings and the coefficients are unstable. 3. The validation of discriminant function is necessary to avoid sample specific conclusions. The U method or cross validation makes use of all the available data without serious bias in estimating error rates. This method holds out one observation at a time, estimates the discriminant function and classifies the held out observation. This process is repeated until all the observations are classified. Most discriminant analysis programs estimate a classification matrix based on the estimation sample. Because they capitalize on chance variation in the data, such results are better than the classification obtained on the hold out sample. 4. The assumptions underlying the discriminant function are: a. The p independent variables must have a multivariate normal distribution b. The pxp variance-covariance matrix of the independent variables in each of the two groups must be the same. The assumptions of the discriminant analysis have to be tested as it is possible that the assumption of equal variance-covariance matrix of the independent variable in each group is not met. In such cases, alternate techniques such as the logit analysis have to be resorted to. 5. This question can be used to stimulate classroom discussion and can be used to explain to the students the various facets of discriminant analysis. 6. The factor loading is interpreted as a measure of association between the factor and the variable (just as a regression coefficient is a measure of association between an independent variable and a dependent variable). In the chapter it has been assumed that the variables are standardized and the factors are independent (orthogonal). Under these assumptions the factor loadings are correlations. This question like the next six are answered fairly well in the chapter. 7. A communality is the percent of variance of a specific variable that is explained by the factors. A low communality indicates that the variable involved is not represented by the factors. If it is believed that the relevant dimensions are reflected by the factors, then it can mean that the variable is not of importance. However, it can also mean that the variable was alone in measuring an important underlying construct. Because there was only one variable representing that factor, the factor probably explained only a small percent of the total variance and because of the rules of thumb in selecting the number of factor to consider, it never emerged. 8. In the rotation discussion the criteria of the two approaches are described. Some discussion might be warranted on the various rotation schemes and how you choose between them. There is no concept of a variable being a function of a few factors and an error terms. Its purpose is to select a factor that will explain the largest amount of variance possible. With one factor selected, principal components will then try to select a second factor (holding the first one fixed) that will maximize the explained variance of the second factor. The process continues until there are as many factors as original variables or until the analyst stops the process. In many analyses, there is a strong halo effect and in fact there really is only one factor, a general like-dislike factor. In that case, it may not be appropriate to use a varimax rotation which because of its objective function will not allow a general like-dislike factor to emerge. The varimax rotation maximizes the variance of the loadings for each factor. The maximum would occur for a factor if half the loadings were zero and half were either -1 or +1. Another rotation is quartermax which attempts to find factor loading patterns in which a variable has a high loading on as few factors as possible. A problem with quartermax is that it has a great tendency to result in a general factor as was found in the principal components solution of Table 20-1. 9. Factor analysis is normally performed to identify underlying constructs and to reduce the number of variables or scales to a more manageable set. One rotation might accomplish this objective better than another. In fact, there are infinite number of rotations possible. A persistent analyst could try thousands. (If thousands were tried, hopefully the data set would be divided so that a validation sample would be available.) The most valid rotation can really only be determined by evaluating the resulting factors using some theoretical judgment. 10. When there is a general like-dislike dimension and all variables are really tapping that dimension, one could easily argue that the data of principal components analysis of Table 20-1 fit this description and the principal components solution was the most theoretically valid. 11. The rule of thumb is to include only those factors that explain one fifth or 20% of the variance. Thus, only one factor would be included. The drop on variance explained from factor 1 to factor 2 would support that judgment. However, the first factor might well be a general factor that under rotation might be decomposed into interesting and interpretable factors such as happened in Figure 20-2. Thus, it might be worthwhile to include two factors and to subsequently look at varimax rotation to see if the resulting factors were indeed helpful and interpretable. 12. A routine approach for items b through e is to run a factor analysis on them. It would be appropriate to divide the sample in half to make sure that the factors that emerge are not just an accident. The factor analysis would be used to generate variables or constructs to use on subsequent analysis. Variables might be selected to represent factors or factor scores might be obtained to use in subsequent analyses. The question really is what type of factor analysis should be conducted for the data described in item a. One approach would be to conduct six different factor analyses, one for each of the six sets of image ratings. Another would be to combine the data for the three bands (giving 6,000 ratings on each image scale). Under this approach a factor analysis could also be run for the two savings and loan firms (giving 4,000 ratings on each image scale). The final possibility is to combine all six rating sets together (giving 12,000 ratings on each image scale). Actually, such a large sample is not necessary; comparable results would split and two analyses run). The selection of the appropriate base will depend upon whether it is felt that different dimensions would be used by respondents when evaluating the different objects. The “Store image Study Revisited” case addresses this question as well. 13. There is plenty of room for a variety of interpretation of the varimax rotation factors. It could easily be argued that the principal components solution though less interesting is more valid. 14. This question can be used to stimulate class discussions. 15. See the teaching notes for PG & E (B) case. 16. There are researchers who will never use hierarchical clustering for the same reason they will not rely upon step regression. The final solution is sensitive to the first few steps which themselves might be rather unstable. Thus, splithalf reliability is low. However, if the data does have structure to it and splithalf runs are possible, a hierarchical program is much easier to work with and to interpret. 17. The quick clustering techniques will generate clusters of four clusters: 3-8-11 2-7-1 4-6 5-9-10-12 PART IV TEACHING NOTES FOR CHAPTERS CHAPTER TWENTY-ONE MULTIDIMENSIONAL SCALING AND CONJOINT ANALYSIS Learning Objectives • Discuss the concept and need for multidimensional scaling and conjoint analysis. • Discuss the different types of input to create perceptual maps and partworth utilities. • Describe how to interpret the solution of multidimensional scaling and conjoint analysis. Teaching Suggestions Although some students will have been exposed to MDS in other marketing courses, for others this material will be all new. Thus, it will be useful to make sure that the student understands the basic concepts of a dimension and the positioning of an object on the dimension. If the factor analysis chapter was covered the students should have a good feel for this core concept. Another important element of MDS is the concept of a similarity based input scale and how it differs from an attribute scale. It should be noted that a preference based scale in this context is really just providing a measure of similarity. Objects that are ranked fourth and fifth should be more similar than objects that are ranked fourth and tenth on a preference scale. Considerable time could be spent on how you can use MDS to make positioning decisions, new product decisions and to define and analyze markets. However, one view is that the research course could really focus upon the methodology, its inputs, outputs, assumptions and limitations, and let the management and strategy courses (and other courses on the functional aspects of marketing) have the primary responsibility to apply MDS. The location of conjoint analysis at this point is somewhat arbitrary. Some of the material may be best understood in terms of the data collection issues in Chapters 7-9. However, with the emphasis on conjoint analysis, the chapter fits more readily with the data analysis topics. There are at least three ways this chapter may be used depending on the instructor’s objectives. (1) The chapter can be used as an extension of the attitude scaling discussion in Chapter 10. The emphasis could be on the special problems of measuring attitude importance. (2) The topics of this chapter could also be used in a discussion of consumer behavior with special emphasis on decision-making choice models. (3) In a more methodologically-oriented course, the instructor may choose to use the material (as positioned) as an introduction to the analysis of dependence techniques. Questions and Problems 1.a. One approach would be to ask Indiana students what other schools they applied to, what other schools were second and third choices, or what other schools are “similar” to Indiana. Another would be to ask high school seniors similar questions. An issue is from what high schools would the sample be drawn. Still another source of respondents might be from other colleges mentioned by the Indiana students. (b) One professor made this the dominant question of a whole course. One approach is to ask respondents “why are the two objects similar?” or “why are they different?” It is usually necessary to use factor analysis to remove the redundancy for the attribute list. (c) One approach is to ask whether object A is more similar to object B or object C. Another is to ask whether the pair of objects, A and B. are more or less similar than the pair of objects C and D. Still another is to ask which of the objects is most similar to Object A. A very different approach utilizes free association data linking objects to image components and is reported by Jain and Etgar (Journal of Retailing, Winter, 1976-77, pp. 61-70). A variety of approaches can be based upon attribute data. (d) This question should lead to a useful discussion of exactly what perspective provides the bases for similarity judgments. 2. Basically, ideal object clusters should identify “holes” in the space for which there is demand for a new product. For a review of the problems see Alvin J. Silk’s article, “Preference and Perception Measures in New Product Development: An Exposition and Review,” in Sloan Management Review, Fall, 1969, pp. 21-37. 3. This question identifies a very practical problem of scaling and of product development. It is possible to link changes in product composition and taste, in packaging, in advertising, etc., to changes in image on psychological dimensions but it is not at all trivial or obvious. 4 The Hustad et al. article in the Journal of the Academy of Marketing Science (Winter, 1975, 34-47), indicates that ideal objects were obtained from preference data. The interesting fact and the point of the question is that preference judgments and the ideal object locations were made conditional on context. Thus, for use contexts such as: During the Summer When friends come for dinner During the Winter When you are thirsty For breakfast When you wish to relax For lunch When you need a pick-me-up Hustad et al. determine context specific ideal beverage dusters and attributes most closely related to them. 5. The statement is a bit strong. The use of similarity data is motivated more by a desire to improve validity than by a hope to gain insight into “unconscious” motives a la motivation research. A purpose of the question is again to get the student to think in terms of what underlies a perceptual map. 6. MDS has been used to evaluate advertising (e.g., Lautman, Percey and Kordis’ article in JARS June, 1978, pp. 35-40.) Of concern are questions such as: Will a few exposures change a position, especially for an established brand like Pepsi? Will the map really represent the position? Will the key dimensions and relative competitive objects be represented? 7. A family may buy two brands for two different family members or for two different use applications. In either case, there is no reason to assume that the brands are in any way similar or competitive. However, the potential value of getting maps from scanner data is so large that it is probably worth pursuing. 8. These questions are addressed by John Summers and David McKay in their article, “On the Validity and Reliability of Direct Similarity Judgments”, in MR. August, 1976, pp. 289-295. They got students to provide similarity judgments of automobiles and distances between campus buildings. They then asked them to indicate which of three perceptual maps matched their view: their own, an aggregate map or a third map. Such a test plus a test retest measure paints a rather discouraging picture of the validity and reliability of similarity judgment data. 9. Such an exercise is useful to show the students that the dimensions in practice are not so easy to identify. 10. Before a study, cluster analysis would be used to cluster respondents with respect to the similarities measured(or whatever perceptual data was involved) so that the MDS is conducted on groups who are homogeneous with respect to their perceptions. After the perceptual map is generated, we may want to cluster objects to see what competitive groups exist or we might want to cluster people to identify segments. 11. The advantage to using a rating scale is that it can be administered via a mail questionnaire. A ranking task is more complex and usually requires a personal interview. 12. Attribute importance is measured by the difference in utility between the highest and the lowest level of an attribute. The greater the difference in utility, the more important the attribute. If all levels of an attribute have the same (or nearly the same) utility, the attribute has no influence on overall attitude. The levels chosen for consideration may influence the resulting importance rating. In the credit card example (Figure 21-6), the respondent appears to be relatively indifferent between marketing support at the 1.0% level vs. at the .75% level (there is little utility difference). However, if marketing support were to increase to 1.25% or decrease to .50% the respondent’s utility may be quite different, 13. Full-Profile Advantages - Description of concepts more realistic since all aspects are considered simultaneously. - Can employ either a ranking or rating scale Trade-Off Advantages Dealing with only two attributes at a time makes respondent's task easier (though longer) Researcher is assured that respondent has considered each attribute Yields somewhat higher predictive validity for large numbers of attributes, which are environmentally independent. In the battery case, only three factors are involved which are not environmentally correlated, therefore, the trade-off approach would not be tedious or difficult. Moreover, the manufacturer is especially interested in the maintenance question and would not want to take the risk that the respondent ignore the variations in this attribute. Since quality is not being asked about explicitly, the danger with “price” being included is not acute. There is no necessary reason why price would be correlated with either of the other two attributes. Normally, price and quality attributes should not be considered separately in a trade-off analysis. Both the Marden SUMM model and trade-off analysis assess importance from the standpoint of a respondent’s willingness to give up some amount (level) of a less important attribute to get as much as possible of a more important attribute. The SUMM model uses a compositional approach coupled with a constant sum rating scale. The trade-off or conjoint models use a decompositional approach which starts with overall judgment of descriptions of concepts or objects. Both types of models require explicit, prior specification of the levels of all possible attributes. As such, both encounter difficulties with attributes which are difficult to categorize into objective and comparable levels. Because the SUMM model can, in principle, ask about an almost unlimited number of attributes, it may be better suited to very complex choice situations with more than seven or eight distinct attributes. One limitation which the Marden model shares with trade-off, but not conjoint analysis, is the problem of environmental correlation among attributes. 14. One can answer this question with both a “yes” and a “no”. The reason lies in the nature of the purchase decision process for major appliances and in particular, in the lack of relevant product experience that a prospective purchaser brings to the initial stages of search and evaluation. When these factors are coupled with the large impact of the cost of these products on the family budget, we typically encounter a two-stage evaluation of alternatives. In the first stage, price is used (along with size and availability of financing, perhaps) as a screening attribute. The set of those brands/products to be seriously considered is determined by the price range the buyer can afford. Those that are either too high or too low in price are not considered further. This can be tested by asking either recent buyers or current prospects to define their consideration set, and describe the difference between those brands inside versus those outside the set. One would expect only a description. A different evaluation procedure, employing a compositional model of preference, is likely used to choose the specific brand and features from within the consideration set. This may only entail a comparison of two or three possibilities. Here price assumes an incomplete sentence. Different role, as an indicator of value for money, which involves a judgment about the price versus perceived quality and features provided. However, it is only one of a number of attributes, and may not even be the most important. Indeed, if all the refrigerators, for example, are at the same price point, it may not even be determinant. This could, of course, be readily tested with a multi-attribute compositional model. One of the implications of this two-stage process is that a manufacturer must first be aware of the variables used by consumers to define consideration sets, for these will dictate how the product line is to be designed, i.e., size and price ranges. Then the feature combinations, plus guarantees and serviceability, must be superior within the probable consideration sets. The choice of airline presumes there is more than one carrier serving the route. Among the attributes that are likely to be considered are: (a) Type of aircraft (b) departure time relative to ideal (c) punctuality of arrival (d) passenger load (e) number of stops en route (f) attitudes of flight attendants (g) entertainment (h) quality of food (i) amount of leg room These and sixteen other service factors, including on-ground services, decor of cabins and others were reportedly used by one airline to develop utility functions on a “route and a purpose of trip” basis. See Paul E. Green and Yoram Wind, “New Way to Measure Consumer’s Judgments,” Harvard Business Review (July-August 1975). 16. While this question is designed as a class field exercise, it should be noted that these questions are precisely those asked by Xerox in researching the plain-paper copier market. Indeed Xerox managers are likely to be among the largest users of conjoint and trade-off analysis, and have developed considerable proprietary software. Among the attributes they use are: speed copies per minute time to get the first copy capacity/volume plain versus coated paper loading time rent vs. purchase size process type size of image form of copy paper PART IV TEACHING NOTES FOR CHAPTERS CHAPTER TWENTY-TWO PRESENTING THE RESULTS Learning Objectives • Discuss the fundamentals of research presentation. • Discuss how to prepare the research report. • Discuss issues related to successful oral presentation. • Discuss the importance of continued relationship with the client. Outline Guidelines to Successful Presentations ▪ Communicate to a Specific Audience ▪ Structure the Presentation ▪ Create Audience interest ▪ Be Specific and Visual ▪ Address issues of Validity and Reliability The Oral Presentation ▪ Don't Read ▪ Use Visual Aids ▪ Make sure that the start is positive ▪ Avoid distracting the audience ▪ Involve the Audience The Written Presentation Questions and Problems 1. The criteria should draw upon those covered in the chapter. Communicate to a specific audience. Well structured. Motivated. Interesting to read. Well written at the sentence and paragraph level. Specific rather than being too general. Figures and charts used well. Understand the main points after reading. A results document would likely have numbers and a quantitative analysis. If so, there is a special need to avoid getting buried in the numbers and instead to highlight those of most interest and relevance. There will also be a need to present statistical tests in a fair and understandable manner. They usually should be in footnotes in any case, they should not intrude. Some students let the statistical significant issues dominate. 2. These follow one evaluation form. 3. The point of this exercise is to sensitize the students to the problems of oral presentations Evaluation of an Oral Presentation Rate each item from 1 (very poor) to 10 (very good) or NA (not applicable). Content 1. Introduction 2. Central idea / first statement 3. / development 4. / re-statement 5. Pace & density of information 6. Level of detail: low/high 7. Conclusion 8. Suitable for the audience Delivery 9. Poise 10. Mannerisms 11. Eye contact 12. Handling of notes or script 13. Voice/pitch 14. /audibility: soft/loud 15. Rate of speaking: slow/fast 16. Conviction & enthusiasm 17.(a) gestures (b) bearing (c) movement Visual Aids 18. Decision to not use visual aids 19. Legibility 20. Handling of aids Suitability for the intended Audience 21. Content 22. Delivery 23. Length: short/long PART V TEACHING NOTES FOR CHAPTERS CHAPTER TWENTY-THREE MARKETING-MIX MEASURES Learning Objectives ● Describe some major applications of marketing research. ● Discuss the applications of marketing research in pricing of a product. ● Discuss the various distribution decisions that require marketing research inputs. ● Describe the techniques used in the actual industry to obtain the measures used to evaluate advertisements. ● Describe the concept of total quality management. ● Describe the methodologies used to measure the different dimensions of total quality management. Teaching Suggestions New product research is often the easiest research to justify since the purpose is usually clear and the value obvious. Thus, much of marketing research does involve new products. In fact, many of the cases in the book are actually new product cases. The purpose of this chapter is to structure and organize this important application area. In this regard, Figure 23-1 provides a very useful structuring function. The concept evaluation and development is organized by the questions: ● How are the concepts exposed? ● To whom are the concepts exposed? ● To what are they compared? ● What questions are asked? The validity of the concept test often is dependent upon the nature of the exposure which necessarily has to be somewhat artificial since the product is not yet fully developed. The point that it is easier (and thus more valid) for respondents to make relative judgments about concepts than absolute judgments might be emphasized. Also, intention responses are nearly always inflated and this source of bias needs to be considered carefully when designing concept tests. Students might be asked if they have ever been respondents in concept tests. Or if they know of any product that has bombed, do they think a well-designed concept test would have detected the product’s problems and caused the cost of its market entry to be saved. Product Evaluation and Development differs from concept testing in that the product and parts of the marketing program have been developed and product use tests are usually possible. The chapter discusses use tests, and laboratory test markets. The laboratory test market is useful in itself and makes a nice contrast to the use of test markets. The next section of the chapter discusses test marketing. One issue is how to design a good marketing test—the considerations like selecting the test area, controlling the test, the length of the test, and the types of measures used. Another issue is the limitations of the market. Students often ascribe more power to a test than it warrants and believe that a scientific manager should always test the market, and that, in contrast, a laboratory test, for example, is rather invalid and crude. The fact is that a test is expensive in time and money, is difficult to control (and thus may generate biases results), contains substantial variation (product acceptance often varies widely among cities), and requires that the marketing program be essentially complete (a test market is not the place to refine all elements of the marketing program). Forecasting is a subject that can be bypassed in a research course. We included it in this book because it has such practical value. Many students will seldom be exposed to a focus group, a survey, or an experiment when they get on the job, but most will have to forecast. Consequently, a feel for forecasting and the role of research techniques to designing and executing forecasting systems will be very beneficial. The final section in this chapter deals with pricing research. It is worthwhile for the instructor to go through each and every pricing strategy and the informational requirements for each strategy, in detail. Wherever needed, examples have been provided to explain the various pricing concepts. A feel for the marketing research applications in distribution will be very useful to the students. The chapter provides an overview of the marketing research applications in distribution decisions. The role of marketing research in distribution decisions is generally overlooked but due to the practical value of applications in this area, we have decided to include it in our book. Advertising is an excellent context to illustrate and discuss the concepts of marketing research because it is inherently interesting to students and because it is a well-developed and focused research area. To motivate students it might be well to start with an open-ended questions—what is advertising supposed to do anyway? Hopefully, some feel of the variety of tasks (and thus, response measurements) that can be involved will quickly emerge. Another question that might add motivation is which test is the best? The students may well benefit from a review of four or five representative copy test methods. There are a lot of tests and it gets somewhat confusing. It will be helpful if they are comfortable with a few and those can then serve as a frame of reference. A discussion of just a few in-depth—can serve as an excellent review of practically the whole course. For each the instructor should first get it described accurately (through discussion or lecture) and then discuss its limitations and assumptions. Question 4 and the discussion of it below can be a point of departure. Questions and Problems 1. In all these situations several key issues will emerge. Which segments will be the most fruitful? The demand estimate will nearly always need to be made by segment. What will be the marketing program? Distribution channels and price are particularly crucial to most estimates. The estimate will usually need to be based upon some assumption of these two critical elements. For example, consider the first two products. (a) One market is travelers. Airline travelers could be sold either through vending machines or over-the-counter. In either case the package design will be important because it will have to communicate the concept. The airlines themselves have, in fact, bought quantities of these for their passengers. Another market is truckers. In this case, focus groups might be appropriate to get reactions to the product and to various distribution alternatives. Another approach is to visit airline, bus and train waiting rooms and truck stops, expose the concept and get reactions. (b) One market is restaurants and other commercial eating establishments. One approach would be to obtain a list of a variety of restaurants such as fast foods, chains and company and school cafeterias. The concept could be exposed to them and their reaction obtained. (c) Another market is the home. To explore the market, “experts” like gourmet cooks and buyers in the relevant departments of department stores might be contacts. Use tests might be highly appropriate to see if it actually works. A knowledge of how much lemon is used on food (as opposed to in drinks or in cooking) might be helpful in assessing the total market potential. Secondary sources or a survey could provide such information. 2. Benefit Structure Analysis is described in the cited article by Myers (IM, October 1975). It starts with 25-50 in depth interviews or several focus groups. Specific use occasions are probed. One objective of this phase is to obtain the list of benefits and product characteristics. Specific questions that can be used include: (a) Why is that brand preferred? (b) is Brand A more similar to Brand B or Brand C? Why? In what respects? Describe Brand B. What are its characteristics? (c) The second phase involves 500 or so respondents obtained from 12 metropolitan areas. Among the available outputs are: a. Benefits Wanted b. Benefits Received c. Product Characteristics Wanted d. Product Characteristics Received e. Use Occasions f. Products Used g. Brands Used h. Respondent information A variety of analyses are possible. Two key variables are benefit deficiencies (benefits wanted-benefits received) and product characteristic deficiencies. For a detailed discussion, see Myers’ article. 3. Douglas Jordan, of Schrello Associates, puts forth the following set of criteria: (a) is it real? Is the market real? o Is there a need? want? o Can the customer buy? o Will the customer buy? Is the product real? o Is there a product idea? o Can it be made? o Will it satisfy the market Can the product be competitive? o On design/perf. features? o On promotion? o Is the price right? (b) Can we win? Is the timing right? Can our company be competitive? o In Eng./production? o In sales/dist.? o in management? o in other considerations? Will it be profitable? o Can we afford it? o is the return adequate? o is the risk acceptable? (c) Is it worth it? o Does it satisfy other company needs? o Does it support company obj.? o External relations? 4. One approach is to compile a list of relevant words, break them down into syllables and develop a computerized list of all possible combinations and permutations of those syllables. They are screened for reasonableness and then scaled as to desirability. 5. The point of this question is to illustrate that some new products are easily communicable even over the telephone or via a mail questionnaire and others must be demonstrated visually via television or in person. The second question should serve to introduce the trade-offs that all new product managers must make. On one hand, there is enormous pressure to move quickly and “quietly” before competitors can react to it. On the other hand, testing can avoid introducing losers or using ineffective marketing programs in the introduction phase. When the pressures are severe (although they always seem intense), then it is very tempting to bypass testing. 6. There are several approaches. One is to determine if the respondent has the capacity to purchase and the need for the product. Capacity to buy a car might depend upon financial resources. The age of present car might indicate the need. Another approach is to ask for a commitment. For example, a respondent might be asked whether he or she would be interested in receiving information about the product when it is available. A negative reply might indicate that the intentions interest was inflated. Another approach is to scale the intentions questions so as to obtain a measure of intensity. For example, the respondent could be asked if the concept was appealing even at a higher price. The best way, however, is to obtain an agreement to actually make a purchase commitment. Sometimes that is feasible. 7. The main set of assumptions relate to the artificiality of the laboratory setting. It must be assumed that the advertisement exposure and the simulated shopping expedience have enough realism so that the resulting trial purchase is a valid indicator of a trial purchase in the market. There is non-response bias (many recruited will not cooperate), before measurement effects, and a testing effect all affecting internal as well as external validity. In particular, the home use of a new brand is rather artificial because of the testing effect and the before measurement effect. It is also assumed that brand awareness and distribution achieved can be forecasted. Finally, there is the assumption that something will not happen in the market (competitive reaction or channel development) to alter radically the context in which the brand will be introduced, 8. The CDSM basically has two problems, the inability to test the product’s ability to gain distribution and the possible lack of projectability of the available test cities. Some firms may have so much channel power and/or experience that they can forecast their ability to gain shelf space and retailer cooperation for displays, etc. Another consideration is the product type. A new cereal’s ability to get shelf space is more predictable than some product that is opening up a new product class. 9. The question was stimulated by the Beecham suit against Yankelovich Clancy Shulman because the YCS laboratory test market provided an inflated forecast of the sales of Delicare Cold Water Wash because of alleged YCS mistakes. Some possible responses/positions are as follows: (a) The research firm is a professional organization and it should be responsible for this unprofessional error to the degree that this error affects the result of the project. (b) A research firm is simply supplying information. The client has some responsibility of understanding what the research methodology is and how the research output gets transferred to recommendations. The client is not a naive man off the street but in fact an experienced user of marketing research information. The model used is not a blackbox—the basic structure and assumptions are exposed to those who choose to make the effort. (c) The research firm is a professional organization that should be held responsible for incompetence just as a physician is held responsible when a mistake causes a patient to be permanently harmed. 10. Periodic discounting is followed by firms that possess information on the reservation prices of consumers. When some consumers in the market have differential reservation prices, then firms can start at a high price and periodically discount their prices in order to draw consumers with low reservation prices. Companies follow random discounting when they perceive the customer differentiation in terms of the search costs. Consumers with high search costs tend to buy at undiscounted prices and the consumers with low search costs tend to wait and buy at the lowest price. 11. This would be considered as a case of Second market discounting. 12. This question has been framed so as to enable the student to think of the various possibilities. This question can also be used to stimulate classroom discussion. 13. There is a basic hierarchy notion that attention to the ad must precede communication. Recognition is an indication that the ad was at least seen. If recognition is low, that tells you something useful about the ad. Further, there is the argument that some ads, namely those use an emotional appeal, do not attempt or even want the audience to engage in information processing. Thus, a recall measurement would be completely inappropriate. The FCB study in the text in which emotional ads did relatively badly with recall scores makes the point. 14. The BRC uses a large (500) sample mail survey and is inexpensive (under $1,000). The large sample provides high reliability. The issue is the stimulus. The BRC experience is that the stimulus is, in fact, adequate and it can even tap feelings of warmth and irritation that could only have been present when the ad was first exposed. Communicus involves costly personal interviews even with mail samplings, However, the stimulus, namely exposure to an excerpt of the ad is much more realistic. 15. Again, the hierarchy notion is that attention should be followed by interest/processing of a communication to be effective. DAR is a measurement of interest/processing that has high face validity. If a message was processed, it should be recalled. In contrast, you can recognize all sorts of nonsense material that had no meaning to you. The “day after” feature is also highly supportable. How useful are the measurements taken just after an exposure? After all, you don’t buy right after right after an exposure. The controversy over DAR may be an overreaction to it simply because it was so established that it began to dominate the creative process. You could not get a commercial through (even with an emotional appeal) if it did not Burke out well. The power of DAR needed to be cut back and a fullscale attack was about the only way. 16. The McCollom/Speilman version does use multiple geographic locations and a different persuasion measurement. It also features two exposures and the A/C measurement. (a) The Mapes and Ross Approach ● also involves a before measurement ● probably has less of a selection bias, but still requires cooperation and draw only from cable homes ● the artificiality of the viewing context is much less ● the lottery might be a more involving valid measurement of preference, but it could also stimulate a variety-seeking bias—l’ll try something new—it’s free (b) The Apex Approach ● uses a before measurement, although it is different—how much does that help? The control group provides a comparison point. ● uses a brand preference measurement that should be more sensitive and valid than a single choice while retaining the “prize” motivation ● the artificiality of the viewing context is the same as the Mapes and Ross test ● the control group helps to control for some elements of sample composition (c) The Tele-Research Approach ● has no before measurement—the results are compared to a control group ● there is a nonresponse bias and a selection bias (caused by the selection of the shopping center(s) ● the forced exposure is similarly artificial—perhaps more as there may be more time pressure to get out of the test and resume shopping ● the coupon requires perhaps more commitment than a prize selection—further, the shopping is done in natural surroundings (d) The Sherman BUY Test ● has a fairly artificial setting and sample ● probably has a good dependent measurement. The Guttman scale concept that someone has to go through all steps provides a nice theoretical background to the test ● there is no before measurement (e) The Behavior Scan approach eliminates nearly all threats to validity. The only problem is the representativeness of the sample. The nonresponse bias remains and the issue as to whether the test towns are not representative. IRI has added 12 “neighborhood” samples from New York, Los Angeles, and Chicago so they recognize the problem. 17. A host of factors other than advertising can cause sales such as price, distribution, and most of all competitor actions. Tracking studies focus in on the advertising performance and allow you to address the key questions. Is the advertising performing satisfactorily? Or do we need a new campaign or a new agency? As the example in the text illustrates, it can also provide guidance as to what is the problem. Ad awareness (and sometimes brand name awareness) is usually measured. A low awareness level has obvious diagnostic value. 18. An adjective check list will help to determine if the advertising is perceived as being warm, irritating, entertaining/humorous or informative. Similar devices are widely used. In particular, Leo Burnett and Y&R use a similar phrase check list extensively. Instead of a warm factor, they have an empathy factor. “I can see myself doing that.” “I can relate to that.” etc. Some believe that unless some ads can achieve a degree of empathy they will not perform well. 19. Such a field experiment should be concerned with: (a) External contamination such as competitor actions, weather factors, seasonal effects, etc. Some thought and information gathering would be worthwhile. (b) A big concern would be the carry-over effects, although since this is a new product (a new version), there should be an immediate impact. Nevertheless, the teflon image, long term, is very important and we should be concerned about what happens to it. If the same test were repeated two years later, the carry-over issue would be even more crucial. Agencies sometimes argue that it may take three years to detect carry-over effects. (c) The experiment seems costly. Could fewer cities be used if they were matched instead of randomly selected? Could the response be measured in terms of audited sales of key retailers instead of calling 1,000 people? Could a store-based experiment be conducted using the store’s retail advertising effort instead of spot buys of broadcast media? 20. Mediamark uses the recent reading techniques, while Simmons uses the through-the-book. Obviously, the recent reading has more potential for inflated readership scores. Someone who normally reads it may incorrectly assume they read that issue when they didn’t. (On the other hand, someone who saves up issues may be incorrectly noted as not reading an issue.) The through-the-book is more certain to record only those people who actually read it. The more serious concern is to make sure the comparison is consistent across vehicles within a given method. The text mentions the case where Newsweek does better at out-of-home readership which is detected easier by the recently reading method. Time. in contrast, will get more serious, thorough readership which is best measured by the through-the-book approach. 21. Respondents would generally tend to exaggerate their readership of prestige magazines to build their own self-esteem, to impress interviewers, to conform to perceived social norms, etc. The use of an unobtrusive measurement might help to remove or estimate the bias. One technique might be to check the refuse of a home or office to see what magazines are actually being thrown out. Another might be to have some organization, like the Salvation Army, collect magazines from a home and see what magazines emerge. Still another might be to sit in a public lobby and simply observe which magazines are picked up. 22. The audimeter cannot provide information as to who is doing the viewing. The diary requires considerable effort. Some may neglect to fill it out at times and others may try to do it by memory and make mistakes. Nonresponse bias is another problem. A camera is considered by many as too Orwellian. Is it reasonable for marketing research to ask anyone to give up that much privacy? Further, there is the practical question of nonresponse bias. Are not those who would agree to such a request different than others? The switch, people use to indicate entering or leaving a room is a good possibility. It could have each person’s name on it. Or everyone could have their own p switch. It could be made portable like a remote control, however, with the VCR remote control and the TV remote control, can a respondent keep track of them and avoid losing them? 23. There are three approaches to this issue. The first approach called the sales effort approach is logical and straightforward. In this method, the number of sales calls to be made to prospects are determined is divided by the average number of calls that can be made by a representative and the number of sales representatives is estimated. In the second method, the territory sales history is examined. An analysis of the sales versus market potential is done to assess the manpower requirements. In the third method, two areas are considered. While more frequent calls are made in one area increasing the number of representatives and the frequency of calls, the intensity of sales effort is reduced in the other area and the effects studied. Also, there are a number of computerized models for determining sales force size and for allocating the sales force by market and product line. 24. Trading area data is useful in creating mailing lists, evaluating a store’s or shopping center’s market positioning, measuring competitive customer bases, determining potential of new locations and evaluating regional retail chains and acquisition plans. Formal models have been designed for predicting the trading area of a given shopping center or retail outlet based on relative size, travel time and image. An analysis of the credit card customers can give a useful estimate of the trading area. However, the best method is to conduct surveys to determine the trading areas. Shopping center surveys are conducted along with a survey of the non shoppers to arrive at the appropriate trading area. 25. The point of this question is to get the students to review the various models available for location of warehouses and salespersons and to recall the pros and cons of each model. The students should identify the key capabilities of each model. The review should provide the students with the opportunity to appreciate the various types of computerized models available. 26. Total Quality management (TQM) is a process of managing complex change in the organization with the aim to improve quality. The characteristics of an organization that has successfully implemented TQM are as follows: ● A TQM company continually strives for quality. To help in this, all employees must be trained and educated continuously. ● Formation of cross functional teams is a must. ● Quantifiable measures of progress must be established and rewards should be based on these measures. ● Use of the state of the art techniques, processes and tools to achieve total quality. In order to achieve Total quality, it is important to have the measures for the TQM programs. It should be remembered that the measurement be specific. All the relevant information pertaining to the customer should be converted to suitable measures so that customer expectations can be met. It is also important to have internal goals and measure the achievement periodically. PART V TEACHING NOTES FOR CHAPTERS CHAPTER TWENTY-FOUR BRAND AND CUSTOMER METRICS Learning Objectives • Discuss the agenda for marketing research in the twenty‐first century. • Discuss the concept of competitive advantage and the various ways of measuring it. • Discuss brand equity and the various techniques used to measure it. • Discuss customer satisfaction and the different methods of operationalizing it. • Discuss the importance of a customer satisfaction measurement process. • Describe the concept of the “Wheel of Fortune Strategies.” • Discuss the strategies to maximize customer profitability Teaching Suggestions In a sense, this chapter represents the trend of the marketing research today. The role of marketing research is all too pervasive, encompassing all spheres of marketing. With so much of uncertainty in the business environment, more and more corporations are seeking research help to provide insight into matters of critical importance. In this context, this chapter will be very beneficial to students in that they are exposed to some of the very latest in marketing concepts. The instructor can start off with an introduction of the corporate environment and explain how competitive advantage plays a vital role. The various methods of assessing competitive advantage can be gone through, in detail. Then the concept and measurement of brand equity can be done. Emphasis should be placed on this topic, as an exposure to this topic will be very beneficial to the student, as it has very high practical value. Next, the instructor can discuss how “Buyer –Centricity” is emerging as an important concept. The instructor could define “Buyer- Centricity” and explain how current IT tools and technologies can enable organizations to analyze existing data and pin-point specific buying patterns, preferences of their individual customers. One key point to be emphasized is that buyer centricity benefits both the sellers and the buyers because organizations produce only those products and services sought by the customers. With so much of importance placed in customer satisfaction, it is obvious that measurement of customer satisfaction will play a big role in the nineties. This topic can provide the instructor with the launch pad to move over to topics like relationship marketing, integrated marketing communications and total quality management. Finally, conclude with a discussion of various strategies that companies can use to maximize customer profitability. Questions and Problems 1. The type of data to be collected depends upon the nature of business and the clientele covered. In case of a service oriented organization, specific measurements to measure the service quality indicators must be developed. Let us assume that an airline wants to design a customer satisfaction measure. After assessing the customer expectations and the relative importance of each event to the customer, the airline may decide that timely arrivals/departures and efficient baggage handling are the keys to enhanced customer satisfaction. The airline should focus on developing quality measures in order track the customer satisfaction measures. This information should not be made exclusive to the top management but this information should percolate to all the employees. Similarly, information related to various suppliers, employees should be collected so as to achieve quality in all spheres. 2. Customer satisfaction research is a natural corollary to the quality movement of the 90s. Satisfaction research can be studied in the context of an interrupted time series quasi experimental design. Over time, management can improve its processes and evaluate at regular intervals to see if the programs worked. The major guidelines to a customer satisfaction measurement process are: * Define goals and the use of the information * Discover what is important to the customer and employees * Measure the critical needs * Act on the information * Measure performance over time. The students should be made to go through the strengths and weaknesses of the methodology discussed in the text and ascertain the validity. 3. The objective of this question is to stimulate a healthy classroom discussion on the issue of brand equity. There is plenty of room for varied interpretation of the various issues involved. 4. Measuring competitive advantage is central to strategic thinking. Businesses seeking advantage are exhorted to develop distinctive competencies and achieve differentiation through lower costs or higher customer value. Measurement of competitive advantage can be done using a market based approach or a process based approach. The text contains a summary of all of the market and process based approaches. 5. Integrated Marketing Communications (IMC) is in the process of developing and implementing various forms of persuasive communications programs with customers and prospects over time. Goal of IMC is to influence or directly affect the behavior of the selected communications audience. The process starts with the customer and then works back to determine and define the forms and methods through which persuasive communications program can be developed. The costs involved in an IMC approach is the creation of an infrastructure to assess the needs and wants of the customers. The scope of the team is focused. Research regarding customer needs must go beyond features and benefits to determining the best way the products and services are best delivered. Post search measures how well the organization did in improving customer perceptions and behavior and which areas need improvement. The benefits are focused communication, enhanced customer satisfaction and increased shareholder value. The advertising agency should conduct a cost-benefit analysis along the above mentioned lines and convince the management that IMC is a worthwhile strategy. 6. Students should highlight the benefits of the buyer-centric approach over the traditional seller-centric approach. Some points to consider are: ∗ Buyer-centric approach focuses on the consumers and creating value for them ∗ It provides services to help consumers make informed purchase decisions ∗ It focuses on helping consumers to buy, rather than enabling sellers to sell ∗ It offers products/services that are relevant to the needs of the consumers by recognizing consumers’ unique needs ∗ Buyer-centric approach aids consumers, while benefitting sellers as well SECTION II TEACHING NOTES FOR CHAPTERS CHAPTER TWENTY-FIVE NEW AGE STRATEGIES Learning Objectives • Describe the concept of database marketing research. • Discuss the role of marketing research in relationship marketing. • Discuss the key elements of e‐commerce. • Discuss the key elements of mobile marketing. • Discuss the key elements of social marketing. • Discuss the key elements of experiential marketing. • Discuss how market research can be used to measure word‐of‐mouth value.• Discuss the concept of Internet of Things. • Discuss the concept of bots and artificial intelligence and their application in marketing. • Discuss the concept of blockchain. Teaching Suggestions This chapter focuses on the emerging trends and strategies in marketing, that are aimed at gaining a better consumer understanding, and improving the delivery of product, services and marketing communications to consumers. Students must be made aware that with the current technological advancements, there is greater sophistication and depth of information available to identify satisfied and profitable customers, their buying patterns, preferences and short-term and long-term business needs. After the above introduction, the instructor can explain how database marketing and customer relationship marketing (CRM) concepts are connected. The students may be exposed to the utility of relationship marketing. The utility of adding information (e.g. demographics and lifestyle) database to enhance a list can be stressed. The power of statistical models (choice models) can be illustrated by discussing the possible use of the information in a database. For example, how a computer firm can use their customers' profile on spending habits to target potential customers who have a tendency to spend more in a given time period. Finally, the customer database allows a firm to identify their profitable customers, thereby providing the ground for relationship marketing. The growth of relationship marketing has been made possible with the use of databases. Instructors may introduce the importance of assessing the value of individual customers to the firm, and the wealth of information made available by loyalty programs. The keys to relationship marketing should be followed by a discussion of the recent developments in relationship marketing research like calculation of customer equity or customer lifetime value, which can have a huge impact on marketing resource allocation of firms based on profile analysis, and prediction of purchase sequence and timing of next purchase. Metrics that are used to measure customer value can be discussed, and the CLV metric can be introduced and discussed. E-commerce is another popular application of marketing research today. The instructor must spend time in discussing various e-commerce applications and trends. Various tables and examples are given that reveals the revenue implications of e-commerce. Mobile marketing is meant to describe marketing on or with a mobile device, such as a mobile phone. This is a rapidly growing application of marketing methods wherein a mobile device is used as the marketing vehicle. The chapter initiates a discussion on what this type of marketing is about, its advantages over traditional marketing methods, and its potential applications going forward. This topic should initiate class discussions on the current trends in this area. Social marketing refers to the use of social media such as Facebook, Twitter, MySpace, YouTube among others to host marketing campaigns. This form of marketing has attracted large amounts of marketing spend and is expected to increase manifold in the future. This discussion provides some of the popular means of adopting social marketing and the viral nature that it holds. Experiential marketing proposes to connect consumers with brands in an effort to make the consumers as advocates. This is yet another development in the field of marketing and many companies are providing experiential treatment to their customers in an effort to grow customer advocacy for the brands. Word-of-Mouth marketing is the art and science of building active, mutually beneficial consumer‐to‐consumer and consumer-to‐marketer communications. It generates a positive influence on the product for consumers, depends on consumer-to-consumer communications and carries greater credibility. The rise of online social media means that the spread of information is not subject to any boundaries and the implications of this on marketing should be discussed. Finally, the instructor can introduce the concepts of IoT, bots, AI and blockchain – all of which have been buzzwords, and grown in popularity in recent times. These developing technologies have been defined and described, with the purpose of stimulating discussions among students about the advantages, disadvantages and the implications of these evolving areas for marketing. Students may also discuss potential applications of these technologies in various areas of marketing and business. Questions and problems: 1. With available data about its customers in hand, AT&T can segment and target customers for Universal Credit Card from its existing customer base. It can send information and introductory offers for the credit card along with the monthly telephone bills to existing customers. Current customer profiles can be used to identify and communicate with prospective customers who exhibit the characteristics of profitable customers. 2. Students can research and discuss their findings in class. 3. With a marketing database, marketers have information about customers, their demographic characteristics and their response characteristics. With a database, marketers can use past actions by customers to predict their future preferences or profile customers for effective market segmentation. With knowledge about current customers’ tastes and preferences, marketers can effectively target new customers with the same characteristics, and even predict the lifetime value of these newly acquired customers. 4. The database should be updated monthly, which is in line with the cycle of fees collection. With monthly update, the company can identify active and inactive customers and thereby update its customers contact plan. 5. Modeling customers serves several purposes: It helps to profile a typical customer and so become more effective in prospecting. It helps to identify the best customers, another aid to prospecting. It helps to identify niche markets to add to the marketing universe. It helps to develop more effective marketing tools (materials and media). It is costly to try to solicit every member of the database. The rule is to solicit only those segments whose expected response rates are above the breakeven rate. 6. Students can research the basic concepts and foundational aspects of these topics and discuss the possible privacy implications in class. Instructor Manual for Marketing Research V. Kumar; Robert P. Leone; David A. Aaker; George S. Day 9781119497493
Download
Close
Close
