SYNTHESIZING DATA FOR MEDIA SIMULATIONS Developments in Business Simulation and Experiential Learning, Volume 25, 1998 SYNTHESIZING DATA FOR MEDIA SIMULATIONS Hugh M. Cannon, Wayne State University Karthekayan Ramachandran, Wayne State University Edward A. Riordan, Wayne State University ABSTRACT Media planning simulations offer enormous po- tential for increasing the efficiency of training professionals in the area of media planning. How- ever, the simulations themselves can be very complex, requiring large quantities of data. Even with the vastly improved capacity of electronic storage devices, the requirements can be over- whelming, both in terms of space and the physical requirements of locating and transferring data. This paper discusses practical procedures for syn- thesizing the large data sets required for a media simulation using data that are readily available to simulation developers. These procedures are illus- trated through examples from television media. INTRODUCTION The data requirements of any simulation are driven by the nature of the model around which the simulation is based. Our model suggests that EXHIBIT 1: EVALUATING MEDIA ALTERNATIVES media alternatives can be evaluated on two levels. The first is at the level of individual media vehi- cles, where the criterion of effectiveness is the cost per thousand effective target market rating point (CPRPETM). The second level involves the evaluation of com- binations of media vehicles, resulting in a distri- bution of advertising exposures). The relative ef- ficiency of the schedule can be expressed in terms of the frequency value per dollar spent on adver- tising (Exhibit 2). EXHIBIT 2: FREQUENCY VALUE ANALYSIS % of target receiving % of the theoretical optimal response E i ( T p Cost per advertisement Media Audience Target Market Selectivityx Effective Advertising Exposure Cost CPRPTM = Target Market Rating x Exposure Effectiveness T n 1 2 3 187 Level of Exposure 0 3 1 2 Frequency Distribution 36% 49% 14% 1% Response Effectiveness 0% 40% 50% 55% Exposure Value 0.00% 19.60% 7.00% .55% Total 27.15% each level of exposure achieved by this level of advertising exposure % of the theoretical optimum delivered by the schedule xhibit 3 combines these two levels of analysis nto a comprehensive media planning process See Cannon, Leckenby and Abernethy 1996). his will provide the basic model for which the aper seeks to address the data requirements. DATA REQUIREMENTS he evaluation presented in Exhibits 3 suggest the eed for several different kinds of data: . cost of ads placed in alternative media vehicles, including various media options (formats and size ads within each vehicle); . media audience (the number of people exposed to a particular media vehicle); . target market selectivity of alternative media vehicles (including both product-usage and de- Developments in Business Simulation and Experiential Learning, Volume 25, 1998 mographically defined targets); 4. media vehicle exposure effectiveness relative to advertising objectives; 5. within and between vehicle duplication. EXHIBIT 3: THE MEDIA SCHEDULING PROCESS If the simulation were limited to a relatively few alternatives, these data could be simply entered into the simulation in the form of data matrices. However, in order to create an educational tool that offers a high level of creative flexibility, we would like a simulation that includes a much broader range of alternatives. Indeed, we would like the flexibility of incorporating a virtually unlimited number of alternatives. Even more im- portant, our simulation sould allow students to select media that might not have been anticipated by the simulation designer. Media Vehicle Costs Media cost data can be accommodated with sim- ple list of costs. However, the data problem be- comes more complex when we consider the multi- tude of alternative options available. For instance, in television, these include 30-second versus 60- second commercials, station-break commercials versus those embedded within an actual program, commercials aired during different seasons of the year, and so forth. As it turns out, most of the variance in cost is ex- plained by distinctions among the options them- selves rather than variations across media vehi- cles. That is, the relationship (i.e. ratio) between the cost of a 30-second and a 60-second television commercial will be relatively constant across pro- grams and time slots, even though the actual cost varies considerably. This suggests that the cost data problem can be simplified by using cost es- timates, starting with a basic cost-per-rating-point (CPRP) figure, then modifying it based on media attributes. For instance, a standard 30-second commercial would have a cost adjustment index of 1.0, while a 60-second commercial would have an index of 1.8, and so forth (Exhibit 4). EXHIBIT 4: USING ATTRIBUTE COST INDICES Media Vehicle Selections Between Vehicle Duplication Within Vehicle Duplication Adjustments for Exposure Frequency Distribution Schedule Development and Evaluation Media Objectives Advertising Budget Allocation x =Base cost per rating point Product of attribute-based cost adjustment indices Estimate cost per rating point Exhibit 5 provides a summary of indices for dif- ferent media attributes. Note the versatility of the approach. They can be used to represent every- thing from different formats within a particular medium (30- versus 60-second or four-color ver- sus black & white) to different media classes (television versus radio, etc.). In each case, we use the attribute indices to adjust the cost per rat- ing point. Therefore, any variation in price must be due to factors other than the size of the audi- ence. To get the estimated cost, we simply multi- ply the CPRP by all of the relevant indices. For example, if the basic CPRP is $6,000, the daytime CPRP would be ($6,000 x 0.50 =) $3,000. A day- time, 60-second, imbedded commercial, during the third quarter of the year, in a prestigious, mass-appeal program would be ($6,000 x .5 x 1.8 x .8 x .7 x 1.5 x .7 =) $3,175. The indices shown in Exhibit 5 provide a useful set of attribute weightings for developing a media simulation. Indices such as these are often pub- lished in pocket media planning guides developed by various advertising agencies. However, they can also be developed for special applications by simply taking a sample of values from published sources, such as Standard Rate and Data Service, 188 Developments in Business Simulation and Experiential Learning, Volume 25, 1998 and averaging ratios of attribute costs relative to the base. EXHIBIT 5: SAMPLE MEDIA COST INDICES The implications of this analysis is that a users can enter a new media vehicle and the cost will automatically be calculated, using the indices pre- sented in Exhibit 5. The user need only specify the required media characteristics Media Audience Media audience estimates, like media costs, are relatively straight-forward, with data readily available. Unlike media costs, they do not vary with media options. However, there are seasonal and regional variations, for some media, at least. Again, these can be handled as attributes. For in- stance, the incidence of television viewing is much lighter during the summer months, when people typically spend more of their time out of doors. It is also lighter in southern climates, where the climate lends itself to outdoor activities – especially during the winter. For mainstream network programs, the problem is simple. The modeler need only extract ratings data from syndicated sources, providing a national average. From a very practical perspective, these data can be easily obtained from Simmons or MRI, which provide a comprehensive source, and are widely available. While these sources are not generally used for television media planning, this is primarily because they are gathered nationally and averaged over a two-year period. Hence, they do not account for regional or seasonal variations. However, this problem can be addressed by in- cluding a regional and/or seasonal adjustment in- dex. 10-second: 0.5 30-second: 1.0 60-second: 1.8 Embedded: 1.0 Station-break: 0.8 Daytime: 0.5 Primetime: 1.5 Late evening: 1.0 Weekend: 0.7 1st Quarter: 1.0 2nd Quarter: 1.1 3rd Quarter: 0.7 4th Quarter: 1.2 Spot: 1.0 Magazines:Television: 4-color: 1.0 Black & white: 0.7 2nd cover: 1.3 3rd cover: 1.3 4th cover: 2.0 Half page: 0.6 Quarter page: 0.3 Business: 2.0 60-second: 1.0 30-second: 0.6 1st Quarter: 0.4 2nd Quarter: 0.5 3rd Quarter: 0.5 4th Quarter: 0.4 Radio: Vehicle Quality: Prestige: 1.5 Specialty: 2.0 Mass appeal: 0.7 Other Media: Cable: 0.3 Newspaper 0.4 Supplements: 0.3 Out-of-home: 0.2 Dir mail (pkg): 2.0 Dir mail (ind): 60.0 The greater problem arises for non-network pro- grams. For instance, what about reruns? The pro- grams are often the same as those carried by net- work stations, but they can be aired during any number of different dayparts, on different stations or different networks. It does not make sense that the ratings would be the same as it is for prime- time network viewing. While the program mate- rial might be the same, the ratings would depend on the reruns’ dayparts and network/station placement. E e s t c g t t f s 189 EXHIBIT 6: HUT 50.0% Seasonal adjustment x .8 Share x 30.0 Estimated national rating 12.0% Regional adjustment x 1 .1 Local population adjustment x .0170 US Population base (000s) x 185,000 Local TV audience (000s) 415 ESTIMATING LOCAL TV AUDIENCE xhibit 6 illustrates a structure for establishing the stimates. It begins with households using televi- ion (HUT). HUT varies by daypart. Estimates of hese variations are readily available from syndi- ated data sources. Among the most accessible to ame developers are Simmons and MRI. While hey are limited to national and biannual averages, hey nevertheless provide a useful starting place or simulation development. Exhibit 7 provides et of useful values. Developments in Business Simulation and Experiential Learning, Volume 25, 1998 The first adjustment is to address seasonality. As we have noted, audiences vary with the time of EXHIBIT 7: HUT ESTIMATES BY DAYPART y a in te v ti se se H v fo In p te p is of program popularity and (2) the average share of the station or network. As with HUT, station and network share is readily available from syndicated sources. Program popularity is more difficult to estimate. It can be developed using the judgment of the simulation developer, or students who are adding a media alternative to the media plan. Or it can be estimated from syndicated data by compar- ear, a fact that reflects the impact of competing ctivities and variations in personal interests. For stance, summer provides a host of outdoor al- rnatives that distract people from watching tele- ision. These are important to consider when es- mating national ratings. Hence the need for a asonal adjustment index. Exhibit 8 provides asonal adjustment indices for television media. UT times the seasonal adjustment index pro- ides an estimated of the effective national HUT r any given daypart. EXHIBIT 8: SEASONAL ADJUSTMENT INDICES* ing the alternative’s rating (from prime time, or other slot where it is normally viewed) with the average rating for other programs. Thus, a pro- gram whose average share is half again that of other programs would have a popularity index of 1.5. Typically, popularity indices will vary be- tween .5 and 2.0 for non-prime-time program- ming. Monday-Friday Monday-Sunday 6-7 AM 12% 7-8 AM 18% 8-9 AM 21% 9-19 AM 22% 10-11 AM 22% 11-noon 23% Noon-1 PM 26% 1-2 PM 29% 2-3 PM 29% 3-4 PM 30% 4-5 PM 33% 5-6 PM 40% 6-7 PM 49% 7-8 PM 53% 8-9 PM 59% 9-10 PM 60% 10-11 PM 56% 11-midnight 41% Midnight-1PM 25% 1-2 AM 15% 2-3 AM 10% 3-4 AM 7% 4-5 AM 5% 5-6 AM 5% EXHIBIT 9: ESTIMATE MEDIA AUDIENCE INDICES While the preceding steps are sufficient for na- tional media, we would also like a simulation to address local media as well. Exhibit 6 uses a re- gional adjustment index to modify local HUT. Exhibit 10 provides a set of adjustment indices. HUT by daypart HUT by daypart Program share Program share Seasonal adjustment Seasonal adjustment Regional adjustment Regional adjustment Local population adjustment Local population adjustment Index of popularity Index of popularity Station/ network share Station/ network share XX XX XX XX order to estimate ratings, one need only multi- ly the effective HUT by the share of available levision watchers who are viewing the target rogram. Exhibit 9 suggests that the share, in turn, a function of two media attributes: (1) an index that reflect the variation from the national average due different region/season interactions. 1st Quarter 1.2 2nd Quarter 0.9 3rd Quarter 0.8 4th Quarter 1.1 *Adjustments do not apply to early morning (1:00-7:00 AM) and weekday daytime (10:00AM-4:00PM) dayparts. The final factor in local adjustment is the local population adjustment. The model assumes that the rating will be the same for the local market situation as it is for the national market, subject to the urban viewership and regional adjustments. However, the base is different. The actual audi- ence can be obtained by multiplying the rating by the proportion of the total United States popula- tion in the local market. This can be converted to 190 Developments in Business Simulation and Experiential Learning, Volume 25, 1998 an actual audience figure by multiplying it by the total U. S. population. Note that the effect of re- gional and season does not have a significant im- pact on early morning and daytime dayparts, as suggested by the footnote in Exhibit 10. EXHIBIT 10: REGIONAL ADJUSTMENT INDICES From a practical perspective, this discussion sug- gests that a user can add media to the simulation at will by simply estimating the relevant attrib- utes. In the case of television, these would include HUT, seasonality, and share. If it is a local pro- gram, they would also need to include the sta- tion/network share, program popularity, MSA classification, region and market size, expressed as a percentage of the total US. Target Market Selectivity Target market selectivity can be expressed in an index of selectivity, as shown in Exhibit 11. As the exhibit suggests, the index is based on the ra- tio of target market concentration in the media audience, as compared to the concentration in the population as a whole. As a rule, media planners generally begin their analysis by defining target market membership in terms of product usage., assuming that past usage is the best indicator of the kinds of people who are likely to use the prod- uct. Assuming that target market membership is defined in terms of product usage, if 20% of the media audience were product users, whereas only 10% of population as a whole used the product, the index would be 2.0, or 200%. EXHIBIT 11: TARGET MARKET SELECTIVITY Rating Index of selectivity Target market ratingx = Proportion of audience who are in target market Proportion of population who are in target market Quarters 1st 2nd 3rd 4th New England Middle Atlantic West Central South East South West Pacific 0.9 0.9 0.8 0.9 1.2 1.1 1.1 1.2 0.9 0.9 0.8 0.9 0.9 0.9 0.8 0.9 1.0 1.1 1.2 1.0 1.0 1.1 1.1 1.0 *Adjustments do not apply to early morning (1:00-7:00 AM) and weekday daytime (10:00AM-4:00PM) dayparts. In order to get an index of selectivity, one needs a data set that contains both product usage and me- dia audience membership. This, of course, is de- manding indeed, not only for the simulation de- signer, but for actual media planners as well. Con- sider the thousands of media alternatives avail- able, and the even greater number of products. Where would one get access to a data matrix con- sisting of thousands of media crossed with thou- sands of products? As it turns out, a considerable amount of research has been done to identify various approaches for addressing the data problem. The most elegant solution, of course, would be single-source data, where the data base actually contained both prod- uct and media usage. Such data bases have existed since the early 1960s, first in survey form (Gar- finkle 1963), and later through the combining of electronically gathered data (Garrick 1984, 1986). We have already mentioned Simmons and MRI as modern successors to the survey data approach. However, these sources involve enormous amounts of data, and even so, they are by no means comprehensive. Much of the motivation for this article is to provide principles for developing simulations that overcome these limitations. Alternative approaches generally seek to synthe- size the information provided in single-source data by linking product and media-usage data sets using demographics. (See Cannon and Seamons 1995 for a review). While the approaches are ex- tremely popular, none of them performs consis- tently well. Furthermore, they are not useful for 191 Developments in Business Simulation and Experiential Learning, Volume 25, 1998 our data problem, since we are synthesizing both our product and media usage data. Therefore, they do not have any demographic correlates. EXHIBIT 12: ESTIMATING FROM “PROTOTYPES” A ( 1 i n c v t a s s m a s u s W d r s a v s c w g e t The most broadly used demographic categories in media planning are males 18-24, males 25-34, etc. through females 65 and over. There is also some work done on categories of prototypic media, par- ticularly magazines (e.g. Cannon, Williams and Doyle 1992). However, there is virtually no work done on the categorization of products, and the Prototype Data Base n alternative uses the logic of “prototyping” Baron 1990/1; Cannon, Williams and Doyle 992; Cannon, McGowan and Yoon 1995). That s, it draws on the characteristics of a limited umber of variables (media and product usage ategories), using them as “prototypes” for other ariables of interest. According to this method, he selectivity of a given media vehicle relative to product usage category can be predicted by the electivity of the media and product prototypes, as uggested by Exhibit 12. The simulation, then, ust include a matrix of representative product nd media prototypes. As the exhibit suggests, the electivity indices for any synthesized product sage or media variables we be imputed to be the ame as those of corresponding prototypes. hile direct matching of product-usage and me- ia-usage variables is the preferred approach to esolving the targeting problem, demographics are till the accepted language of media planning. As result, the simulation should also include rele- ant demographic variables (generally age and ex). This suggests that the simulation should in- lude a matrix of prototype products and media ith the age and sex demographics. The demo- raphic profile of product users and media audi- nce members will be imputed to be the same as hat of the prototypes. classification work on media falls far short of what we need for purposes of the simulation. Pending better research, game developers will be left to create their own systems for prototyping. Product User Profile Media Audience Profile Media x Product Selectivity Imputed Demographics Imputed Demographics Imputed Selectivity Synthesized Product Synthesized Media Vehicle Media x Product Selectivity Synthesized Data Base Media Vehicle Exposure Effectiveness Media vehicle exposure effectiveness is by far the most difficult type of data to estimate. As a first step, many media planners try to estimate the the proportion of people who are media audience members who are likely to be exposed to a given advertisement. One survey suggests the figures suggested in Exhibit 13 (Kreshel, Lancaster and Toomey 1985). We can use the same numbers, adjusting them up or down, using the cost indices shown in Exhibit 5. Thus, if the rating of a televi- sion program were 20%, the estimated proportion of the population actually seeing a 10-second sta- tion-break commercial would be (20 x .5 x .8 =) 8%. EXHIBIT 13: AD/VEHICLE EXPOSURE RATIOS Direct Mail 65.0% Magazines 52.5% Television 32.0% Outdoor 16.0% Radio 16.0% Newspaper 16.0% Rather than try to estimate effective exposure, we simply settle for advertising exposure. We then estimate effective exposure rate through the ad- vertising response curve implied by the third col- umn of Exhibit 2. Based on the of evidence of published studies, we can assume these to be con- cave in shape (Cannon and Riordan 1995). We can estimate the actual response levels by plotting the curve using a modified exponential function (Cannon, Leckenby and Abernethy 1996). In fact, 192 Developments in Business Simulation and Experiential Learning, Volume 25, 1998 the function can be embedded in a simulation game, thus enabling the simulation to automati- cally determine the exposure value of any given frequency distribution. Within and Between Vehicle Duplication The problem of within and between vehicle dupli- cation is similar to that of target market selectiv- ity. The question is what proportion of a media vehicle audience is also being reached by that of another media vehicle. The only difference be- tween this and target market selectivity is that we are substituting a second media vehicle for prod- uct users. The same problems and solutions also apply. That is, there is no way we can include du- plication data for all possible media vehicles, nor can we even use demographics EXHIBIT 14: USING PROTOTYPES TO ESTIMATE AUDIENCE DUPLICATION to match them. We can, however, use prototypes. Cannon, McGowan and Yoon (1995) provide a detailed description of how this might be done for magazine media. They develop a matrix of selec- tivity indices for prototypic media types (Exhibit 14). To get duplication, they estimate random du- plication, and then multiply it by the appropriate selectivity index. In oder to develop the initial matrix, duplication data are readily available for magazines from sources such as MRI and Simmons. Lancaster, Lee and Katz (1988) suggest a method by which they can be estimated for other media, using a re- gression model. SUMMARY AND CONCLUSIONS The objective of this paper has been to describe a process through which a large number of different target markets (both product usage categories and geographic markets) and media might be incorpo- rated into an educational media simulation. Our solution has been to assume the that variance in the values of and relationships among media and target market variables can be explained in terms of key attributes, such as daypart, station, and program type in the case of television media, and product type in the case of target market variables that are defined by product usage. By including the effects associated with the attributes in a data base that accompanies a simulation, the user may add new media vehicles are markets at will by simply specifying their attributes. REFERENCES M1 M2 M3 M4 M5 M1 M1,1 M2 M2,1 M2,2 M3 M3,1 M3,2 M3,3 M4 M4,1 M4,2 M4,3 M4,4 M5 M5,1 M5,2 M5,3 M5,4 M5,5 Key: M1 = Audience for media vehicle 1 M1,1 = Within vehicle 1 cumulative audience M2,1 = Between vehicle 1 and 2 cumulative audience Note that the references have been omitted for the sake of brevity. For a complete manuscript, please contact: Hugh M. Cannon Adcraft/Simons-Michelson Professor Department of Marketing Wayne State University 5201 Cass Avenue, Suite 300 Detroit, MI 48202-3930 (313) 577-4551(o) (313) 577-5486(f) hughcannon@aol.com http://cannon.busadm.wayne.edu 193 Table of Contents Volume 25, 1998 Marketing Goes to the Movies Bringing Experiential Learning to a Principles of Marketing Course Investment Analysis Application Using In-house Spreadsheet Models SugarCoated Statistics: An Exercise for the First Day of Class Improving Undergraduate Student Involvement in Management Science and Business Writing Courses Using the Seven Principles in Action Establishment and Funding for Interuniversity / Multidisciplinary Student experiences The Prospects of Creative Teaching: A Discussion with Patricia Sanders The Simulation and Classroom Assessment Techniques Developments of Management Skill Assessment Games as Instruments of Assessment: A Framework for Evaluation The Role of Artificial Intelligence in Business Curricula Threshold Solo Competitor: A Management Simulation (V1.0) a Windows-Based. Play Alone, Total Enterprise Simulation and Assessment Instrument Toward An Understanding of One's self-concept Total Enterprise Simulations and the Internet: Improving Student Perceptions and Simplifying Administrative Workloads The Expatriate an Assignment Orientation Game An Expatriate's Nightmare: An Experiential Exercise in Coping with Overseas Assignments Analyzing Experiential Exercise: Using the Scientific Method for Problem Solving The Second Component to Experiential Learning: A Look Back at How ABSEL has handled the Conceptual and Operational Definitions of Learning Predictive Models of Learning: Participant Satisfaction of Experiential Exercises in Business Education Accelerating Moral Development through Use of Experiential Ethical Dilemmas Ethical Dilemmas to use with Business Simulations to Teach Ethics The Class Approach in Behavioral Simulation in a Business Policy/Strategic Management course: A Progression toward Greater Realism An Exploration of the Emergence of Process Prototypes in a Management Course Utilizing a Total Enterprise Simulation How Organizations Are Improving Their Performance Utilizing Electronic Commerce: Examples From The Internet The Market Access Planning System (Maps): A Computer-Based Decision Support System For Facilitating Experiential Learning In International Business An Excel Workbook For Student Planning And Interface With A Simulation Game Design Of Multi-Media Based Pedagogy For Leadership Training Panel Discussion On Using The Internet For Courses Valuing And Enhancing Teaching: Sharing Tips Via The Web The Buddy Project: A Semester Long Project Aimed At Developing An Appreciation For Diversity Enhancing The Excitement And Learning Retention In The Classroom: The Power Of Magic The Supervised Management Internship: A Job Or Learning Experience Team Ware™ An Online Moderated Class Discussion Facility And Beyond A Neophyte Distance Educator's Experience Learning Management By Practicing Management: A Report Of Significant Student Service In 1997 Integration Of Academic And Service Learning: Students' Perceptions About Its Effects And Outcomes The Value Of Incorporating A Service Learning Component Into Course Content: A Presentation And Roundtable Discussion Business Games Teach: Thoughts on the Sources of Conflicting Conclusions on their Effectiveness Antecedents Of Learning In The Simulation: A Replication Using Student Journals To Enhance Learning From Simulations Technological Change And Intertemporal Movements In Consumer Preferences In The Design Of Computerized Business Simulations With Market Segmentation Integrating The Marketing Curriculum Using Collaborative Learning Teaching Time Management In A Sales Program: The Application Of A Computer Simulation Game Adapting Interactive Computer Simulations For Content Based Esl Instruction Multimedia And Student Expectations Synthesizing Data For Media Simulations Composing A Team Health Promoting Behaviors-A Decision Making Exercise Does it really Work? An Application of the Group Interaction Framework Administering the MIT Beer Game: Lessons Learned A Paperless Economy? Instructing Students on the Aspects of Successful Electronic Commerce Maximizing Learning Gains in Simulations: Lessons from the Training Literature Observing General Ability in a Total Enterprise Gaming Simulation FReach Teach: A Computer-Based System for Teaching Advertising Media Planning An Integrated Business Instruction System An Experiential Exercise you can Tinker With Experiential Exercises or Computer Simulations? Cash Flow Statements: Are They Important in Business Simulations? Holistic Cognitive Strategy in a Computer-Based Marketing Simulation Game: An Investigation of Attitudes Towards the Decision-Making Process Barnga: A Game on Cultural Clashes The Many Faces of Culture: Understanding Country and Corporate Culture Students' View of the Use of Business Gaming in Hong Kong Assessing General Management Interest What is the Future of Business Gaming? Starting a Small Music Trivia Business Exercise and Other Innovative Icebreakers An Integrated Approach to Behavioral Skill Development Career Focus: A Student and Business Learning Experience The Use of Concept Mapping in Teaching Strategic Management An Experiential Approach to Developing Mission Statements Business Games in Brazil-Learning or Satisfaction A Simulation within a Simulation: Job Layoff's and Emotional Reactions