A “PROTOTYPING” APPROACH FOR INCORPORATING LARGE DATA BASES INTO MEDIA PLANNING SIMULATIONS: AN EXAMPLE USING MAGAZINE MEDIA Developments In Business Simulation & Experiential Exercises, Volume 22, 1995 103 A “PROTOTYPING” APPROACH FOR INCORPORATING LARGE DATA BASES INTO MEDIA PLANNING SIMULATIONS: AN EXAMPLE USING MAGAZINE MEDIA Hugh M. Cannon, Wayne State University Laura C. McGowan, Wayne State University Sung-Joon Yoon, Wayne State University ABSTRACT Media planning simulations are commonly used in both industry and marketing education to estimate the frequency distribution of media exposure for a given advertising campaign. Typically, these simulations rely on media exposure data from syndicated research services, such as MediaMark Research (MRI) or Simmons Market Research Bureau (SMRB). However, when media are not included in these syndicated studies (i.e. they are “unmeasured”), they must either be excluded from the simulation, or artificial data themselves must be simulated so that they can be included. This is usually done through a process of “prototyping,” or using a measured medium as a prototype for developing the unmeasured parameters. This study describes and tests a method of prototyping in which prototypes are selected through a judgmental process. INTRODUCTION Over the years, media planning has provided a fertile ground for the growth of simulation research. For instance, in 1961, Agostini published his classic formula for estimating the reach and frequency of a media schedule from single medium and media pair audience data. Gensch (1973) and Rust (1986) provide good reviews of this research as of the early 1 970s and mid-i 980s, respectively. However, the research continues with no apparent diminution (e.g. see Ju, Lee, and Leckenby 1994). The literature has also spawned numerous comprehensive planning models, using simulation as a basis for predicting the results of alternative decisions (Little and Lodish 1966,1969; Aaker 1968, 1975; Gensch 1973), as well as simulations that seek to model the decision making process actually used by planners (Fleck 1973). All of these models depend on audience data as a basis for the simulation. These data are readily available from a host of syndicated services, such as Nielsen, MediaMark Research (MRI), and Simmons Market Research Bureau (SMRB). Furthermore, they are relatively economical for industry applications. But not for educational simulations. This is particularly true if the simulation is based on “single-source” data, relating media to target markets that reflect product usage instead of conventional demographics. The expense of data under current licensing arrangements would be exorbitant. And even if this problem were resolved, the sheer bulk of the data make them very unwieldy. The purpose of this paper is to describe an alternative approach to constructing a media database for use in educational simulations. It draws on the principle of “prototyping” of media. While there are several different approaches to the “prototyping” process, they all use a “prototype” media vehicle as a basis for estimating the audience parameters of another in the media simulation. In our application, a simulation can incorporate data from a relatively small number of magazine “prototypes” into the simulation, and from these synthesize a much larger media audience data base. MAGAZINE PROTOTYPING Regardless of source of data used to drive a media simulation, the breadth of media it can accommodate is limited by the availability of data. Some relevant media are bound to be omitted from the base. This is particularly true of magazine data. There are countless numbers of small magazines that may provide useful advertising media, but which are excluded from large-scale syndicated studies due to economic constraints. These are known as “unmeasured” magazines. The usual remedy is to synthesize the missing data through a process known as “prototyping” (Baron 1990/1). This process uses magazines whose audiences are believed to be similar to those of the unmeasured magazine as a basis for simulating the unmeasured magazine audience parameters. Three parameters are of particular interest: (1) the audience size; (2) the target market concentration; and the (3) the audience duplication, or overlap, with other relevant magazines in which the advertiser might choose to place messages. In order to derive the audience parameters, the process uses two types of data. The first are syndicated product-media data, such as MRI or SMRB. These, of course, are the data to which the process seeks to add the unmeasured magazine. But they also provide the measured magazines whose audiences provide the basis for simulating the unmeasured audience. The second type of data are circulation figures obtained from the magazine itself. These represent the number of copies of the magazine in circulation, as opposed to the number of people reading an average copy. The prototyping process begins with the selection of a magazine to serve as a “prototype” for the unmeasured magazine. The traditional approach -- what we will refer to as the “editorial similarity” approach -- is to divide magazines into different editorial classifications -- “women’s service”, “business”, “fashion”, and so forth. Unmeasured magazines are then classified in a similar manner, and prototypes are assigned from the same classification. However, they can also be selected through more quantitative approaches, such as Baron’s (1991/2) profile-distance method. Developments In Business Simulation & Experiential Exercises, Volume 22, 1995 104 Baron’s profile-distance approach has achieved the status of an industry standard for practitioners, in large part because it accommodates any title, regardless of whether it fits neatly into any preconceived magazine classification. However, research suggests that it performs no better than the traditional judgmental approach for magazines that do fall into an editorial classification (Cannon and Boglarsky 1992). APPLYING THE PROTOTYPING PRINCIPLE TO MEDIA PLANNING SIMULATIONS Recall that the application of the prototyping concept to educational media simulations is to synthesize a relatively large database from a much smaller set of media data. The end result need not be a comprehensive set of advertising media, but rather, one that represents a reasonable number of alternative vehicles from each of the major media types. Applied to magazines, this means that the simulation should include alternative titles from each of the major editorial classifications. Given this objective, using the editorial similarity approach to prototyping makes sense. Furthermore, the profile-distance does not make sense. First, compiling the magazine readership studies needed to make the method work is cumbersome for developers of educational simulations. Second, storing the required demographic profiles requires the simulation to incorporate a very large database, which is one of the things our application of prototyping is designed to avoid. Following this logic, this study will use the traditional editorial similarity approach, adapting the magazine categories developed by Cannon, Williams and Doyle (1992), summarized in Table 1. The magazine audience parameters will be estimated as described below. Developments In Business Simulation & Experiential Exercises, Volume 22, 1995 105 Conversely, readers per copy can be estimated where audience and circulation data are available as follows: As we noted, the circulation data are available from the magazines themselves. Readers per copy can be obtained by dividing audience data by circulation. Of course, audience data are not available for unmeasured magazines. But there are at least some magazines for which audience data are available in each editorial classification. A practical estimate of readers per copy can be obtained by applying formula (2) to all the magazines available, and then averaging the resulting RPC figures for each editorial classification. The average RPC figures for the editorial classifications given in Table 1 are presented in Table 2. Adapted from Cannon, Williams and Doyle (1992) .Estimating Audience. Size. Given circulation and readers-per-copy, audience can be estimated by the following formula: Developments In Business Simulation & Experiential Exercises, Volume 22, 1995 106 Estimated RPC are an average derived from audience data taken from MRI (1990/91) and circulation data from MediaWeek’s Guide to Media (1991). Circulation figures are available from every magazine, and are summarized by Standard Rate and Data Service, as well as a host of published media guides. Thus, to estimate audience data, one need only classify each magazine into one of the available editorial classifications, use the RPC figures for that classification, and apply formula (1). For instance, World Tennis has a circulation of 525,000. It falls into the “Tennis” editorial classification, thus giving it an estimated RPC of 2.15. Therefore, the estimated audience is (525,000 x 2.15 = 1,128,750). Estimating Target Mark Selectivity Target market selectivity is generally expressed in terms of an index, where a value of 1 .0 represents a non-selective medium, or one for which the likelihood of a media audience member being a member of the target market is no greater or less than it is for a randomly selected member of the population as a whole. The index is computed using the following formula: The data required to estimate the index can be taken from any syndicated product media service, such as MRI or Simmons. These need only be averaged across the available magazines in each magazine category. Target markets are typically defined by product usage categories. Table 3 presents average indices for each magazine category and ten selected product categories that might be used as a basis for a media simulation. As the formula suggests, the key to estimating duplication is the index (112). The value of U is readily available from MRI or Simmons. R1 and R2 can be obtained by applying the following formula, again using data available from MRI or Simmons: Estimating Magazine. Audience. Duplication Audience duplication is analogous to target market selectivity. The greater the selectivity of one media audience to another, the greater the duplication. The actual duplication can be estimated (as a percentage) by the following formula: Developments In Business Simulation & Experiential Exercises, Volume 22, 1995 107 The logic of prototyping suggests that the selectivity index for magazines within a given magazine category should be very similar to each other. For purposes of developing a media simulation, then, one need only know the index of magazines on other magazines within a given category and the index across categories. The index within categories can be estimated by averaging the index of each magazine on every other magazine within the same category. Ideally, the index across categories would be developed by averaging the index of each magazine within a particular category on each magazine within a second category. For instance, the index of fashion magazines on business magazines would be the average of each magazine within the fashion category on each magazine within the business category. That is, it would be an average of Glamour on Business Week, Glamour on Forbes, Glamour on Fortune, Cosmopolitan on Business Week, Cosmopolitan on Forbes, etc. In practice, however, the number of different possible combinations makes this approach impractical. An alternative is to select a single magazine to represent each category. This produces a single index for each combination of categories. In order to develop a more stable estimate of the index, the process may be repeated, changing one or more of the magazines. In the case of values presented in Appendix I, the process was repeated three times. In the first iteration, the magazine with the strongest loading on a particular category (based on the data provided by Cannon, Williams, and Doyle 1992) was used to represent the category. In the second iteration, the magazine with the second strongest loading was selected. In cases where only one magazine was available to represent a category, the same magazine was used in the second analysis. In the third iteration, the magazine with the third strongest loading was used, or, if the category did not include a third magazine, the magazine with the strongest loading was used once again. After three iterations, the result was three indices for each pair of categories. These three indices were then averaged to get the final estimate. The results of this process are shown in Appendix 1. Diagonal values of the matrix represent the average indices within a given magazine category, and the off-diagonal values represent the average of three indices relating two different categories. In practice, the media planning process would not rely on simple audience duplication data, but duplication within a given target market category. However, several methods exist for estimating this without the benefit of additional data. Cannon (1982) suggests a relatively simple approach, based on the assumption that there are no three-way audience-audience-market interactions. Incorporating this into equation (4), we get the following equation: Note that the equation is identical to equation (4), except that it represents target market rather than total population data. The value of TI12 is assumed to be identical to 112 (Appendix I), given our assumption that there are no three-way audience-audience-market interactions. TU is readily available from MRI or Simmons. The number of target market members in the media audience (TM) is also available from MRI or Simmons. This enables us to estimate value of TR1 and TR2, as follows: SUMMARY AND CONCLUSIONS The objective of this paper has been to describe a process through which a large number of different target markets (product usage categories) and magazine vehicles might be incorporated into an educational media simulation. Such a simulation faces three problems: Incorporating a large number of target market categories and alternative media into a simulation requires a potentially very large database. This makes the simulation program very large and cumbersome. The date contained in the base involves proprietary information. Not only must it be obtained from the companies who sell it, but the designer must also obtain releases to use it in a commercial product. The base is necessarily limited to “measured magazines” -- those included in the syndicated research study from which the data era being taken. Developments In Business Simulation & Experiential Exercises, Volume 22, 1995 108 These problems are not as difficult for simulations that are used in actual media planning situations. Typically, an agency will already have access to the data upon which the simulation feeds. The key to design, then, is simply to make the system user friendly and powerful. For educational simulations, however, the simulation package must generally be complete with data. Furthermore, it must be relatively economical, since students will usually have to buy other classroom materials (textbook, etc.) in addition to the simulation. The process described in this paper is particularly attractive for educational simulations because it addresses the economy issues, both in program size and licensing costs. No licensing is needed, because the database is synthesized rather than copied from an existing source. Indeed, a viable simulation could be created using the data contained in this paper. Furthermore, it could be customized to include any magazines that can be classified into the categories discussed in tables 1 to 3, and in Appendix I. Finally, any errors that might result from the process of synthesizing the media data are not likely to be significant for educational purposes. Educational simulations are concerned with general patterns of media usage, not the accuracy of actual audience figures. Consistent with this logic, future research should include efforts towards identifying better and more comprehensive media types. Eventually, these should be expended to include additional media classes. Again, the key will be to do this without overloading the program with data storage requirements. REFERENCES Aaker, David A. (1968) “A Probabilistic Approach to Industrial Media Selection,” Journal of Advertising Research 8:4 (September), 46-54. Aaker, David A. (1975) “ADMOD: An Advertising Decision Model,” Journal of Marketing Research 13:1 (February), 37- 45. Agostini, Jean-Michael (1961) “How to Estimate Unduplicated Audiences,” Journal of Advertising Research 1:1 (March). Baron, Roger B. (1990/1991) “Using the Profile-Distance Method to Select Unmeasured Magazine Prototypes,” Journal of Advertising Research 30:6 (December/January), 11-18. Cannon, Hugh M. (1982), “A New Method for Estimating the Effect of Media Context,” Journal of Advertising Research 22: 5 (October/November), 41-48. Cannon, Hugh M. and Cheryl Boglarsky (1992) “Matching Unmeasured Magazines to Target Markets: A Test of Two Methods,” Proceedings of the 1992 Conference of the American Academy of Advertising, 95-99. Cannon, Hugh M., David L. William, and Richard L. Doyle (1992), “Toward a Scheme for Selecting Judgmental Magazine Prototypes,” Proceedings of the 1992 Annual Conference of the American Academy of Advertising, 107-116. Fleck, Robert A., Jr. (1973). “How Media Planners Process Information,” Journal of Advertising Research 13:2 (April), 14- 18. Gensch, Dennis H. (1973) Advertising Planning: Mathematically Models in Advertising Media Planning. Elsevier Scientific Publishing: Amsterdam. Ju, Keun-Hee, H.K. Lee, and John D. Leckenby (1990), “A Further Test of a Simple Reach/Frequency Model,” 1990 Conference of AAA. Leckenby (1994) “A Dirichlet Model Applicable to Asymmetrical Media Schedules.” Proceedings of the 1994 American Academy of Advertising Conference Little, John D. C. and Leonard M. Lodish (1969. “A Media Planning Calculus,” Operations Research 17 (January- February), 1-35. Little, John D. C. and Leonard M. Lodish (1966). “A Media Selection Model and Its Optimization by Dynamic Programming,” Industrial Management Review 8 (Fall), 1 5- 23. MRI (1991) Study of Media and Markets 1990/91 Doublebase. New York: MediaMark Research. MediaWeek’s Guide to Media. (1991). Battaglia Michael J. (ed.) New York, NY. Rust, Roland (1986). Advertising Media Models: A Practical Lexington, MA: D.C. Heath. Developments In Business Simulation & Experiential Exercises, Volume 22, 1995 109 Table of Contents Volume 22, 1995 Simulation Performance, Learning and Struggle Are Good Simulation Performers Consistently Good? Cognitive and Behavioral Consistency in a Computer-Based Marketing-Simulation-Game Environment: An Empirical Investigation of the Decision-Making Process Chalk & Cheese: Executive Short-Course vs. Academic Simulations Revisiting Personality Bias in Total Enterprise Simulations Are Good Strategies Consistently Good? Investigating the Use of a Computer Simulation as an Effective Pedagogical Tool for the Application of a Strategic Model The Problem of determining an Individualized Simulation's Validity as an Assessment Tool A Simulation Based Analysis of the Value of Information in the Hrebiniak Joyce Typology of Adaptation Relative to Porter's Generic Strategies The Impact of Sales and Income Growth on Profitability and Market Measures in Actual and Simulated Industries A Comparison of a Stand Alone Version of a Simulation with the Traditional Competitive Version Computer-Assisted Gaming of International Business Analyzing Simulations with Computer-Based Programs and Applying the Experience to a Real-World Business A Preliminary Investigation of the Use of a Bankruptcy Indicator in a Simulation Environment Graduates' Views on the Use of Computer Simulation Games Versus Cases as Pedagogical Tools An Analytical Advertising Model Approach to the Determination of Market Demand Dealing with the Complexity Paradox in Business Simulation Games A Prototyping Approach for Incorporating Large Data Bases into Media Planning Simulations: An Example Using Magazine Media Through the Looking Glass, Inc: Superior-Subordinate Personality Type and the Leniency effect Evaluating the Effectiveness of Role Playing Simulation and Other Methods in Teaching Managerial Skills Student and Teacher Perceptions of a Management Simulation Course Performance Evaluation: The Effect on the Propensity to Create Budgetary Slack Management Team Formation for Large Scale Simulations Comparative Static Analysis with the Complete PPA Package: A Strategic Market Planning Tool Consistency in Intent: Learning Objectives at the 1994 Intercollegiate Business Policy Competition A New Twist on an Old Game: The Business Strategy Game: A Global Industry Simulation 3ed Evaluation of Performance in Management Simulation: A Management Coefficients Model Building SimuWorlds: Strategic Management Games of the Future Bulls and Bears: A Stock Market Simulation A Systems Thinking Paradigm and Think Computer Simulation Model of Broadcast and Cable Television Industry Competition Jacket Factory The Sales Management Simulation The Marketing Management Simulation A Demonstration of Promodel Demonstrating A New, Cross-Functional Business Simulation: Vision+ A Cost Chain For The Business Strategy Game Simulation Special Session On Experiential Teaching Compensation Dilemmas: An Exercise In Ethical Decision-Making Organizational Storytelling: Telling Tales In The Business Classroom Evaluating Experiential Training: Case Study And Recommendations The Internship Portfolio: An Innovative Tool For Experiential Learning, Critical Thinking, And Communication An Ethnographic Analysis Of The Pedagogical Impact Of Cooperative Communicating Consumer Behavior: A Long-Term Integrated Exercise Using Personal Consumption Journals And Consumer Analysis Papers An Experiential Paradigm For Teaching Business Problem Solving Developing Leadership Skills The Spss® Student Assistant: The Integration of A Statistical Analysis Program Into A Marketing Research Textbook Negotiating With Your Students Using TQM Principles To Transform Accounting Systems Into An Experiential Exercise Enhancing The Effectiveness Of Outdoor-Based Experiential Training Using Virtual Reality Concepts Case Writing In A Developing Country: An Indonesian Example Experiential Learning Using Focus Groups How Real Should Experiential Pedagogy Be? A Viewpoint From Our Students Reengineering The Internship: A New Approach To Experiential Learning Utilizing The Cosmopolitan/Local And Marginal Man Constructs To Measure Students' Propensity For Creativity Developing Experiential Processes For Teaching Quantitative Techniques For Business Team Learning Roles: A Cooperative Learning Technique Creating the Ultimate Small Business Student Experience: Melding Score/Ace with SBI The Development of Trust in Work Teams: The Impact of Touch Some Outcomes of Experiential Learning: How the Cultural Dynamics of Different Countries are reflected in Workplace Norms & Values Incorporation of Job Analysis Results in Various Forms of Selection Interviews Chudesno, Inc.: An Evaluation of an Experiential Training and Development Simulation An Exercise for Exploring the Relationship between Jungian Psychological Types and Organizational Dynamics Partnership: A Radical Approach to Experiential Learning Partnership: A Nice Idea, But How Do I Get Started? Recognizing Discrimination at Work Using Critical Incident Skills Questions to Help Students Become More Successful at Job Interviewing The Video Project Introduction to Psychological Type Theory Come On Down Understanding Facilitation for Development and Continuous Learning: A Micro-Workshop Leadership and Empowerment: An Experiential Exercise in Decision Making Experiential Exercises and Pedagogy Track Workshop: Selecting a Manager for Maquiladora, Inc. Experiential Training In Multi-Cultural Corporate Settings The Role of Facilitation as an Aid to Complete Learning An Experiential Exercise to Illustrate Difference in Information Processing Behaviors and Styles How to Deliver Accessible Survey Results Age Diversity in the Workplace- Family Feud Style Nafta Standoff: A Cross-Cultural negotiating Role-Play Three Strikes and You're Out!: A Downsizing Experiential Exercise