Evolutionary approach to the development of decision support systems in the movie industry
Introduction
Over the years, decision support systems (DSS) have been applied and implemented in a variety of companies and organizations. The bulk of the work has concentrated in domains like the manufacturing and processing industry, supply chains, distribution, transportation, and finance. It is much more difficult to find applications of DSS in the so-called creative industries, such as leisure and entertainment. Nevertheless, we propose and demonstrate that even in these intuition-dominated domains, successful implementation of DSS is feasible. The work reported here combines elements of marketing and optimization (scheduling) in an unusual and challenging industry – motion picture – where management is not necessarily predisposed to accept analytical approaches.
Our strategy for implementing DSS in the movie industry was to follow an evolutionary approach. The implementation of decision support system in the motion picture industry concerns a situation of a structured problem (movie scheduling) which is quite amenable to optimization procedures. However, the implementation is in an organizational culture that is dominated by intuition rather than modeling and possibly not a positive a-priori attitude towards DSS. In such a situation, a well thought-out implementation strategy holds the key, where elements such as user involvement, top management support, and communication are important [26]. Resistance to change as a hurdle for the adoption of management support systems has been recognized in the DSS literature for a long time. In this context, the “unfreezing-moving-refreezing model” has often been recommended and used [22]. We believe that for the purpose of getting a DSS adopted in a new area with a potentially skeptical audience, as in the movie industry, it is better not to try to go through the unfreezing-moving-refreezing in one big jump. A step-by-step approach is more effective. It is delineated in this paper.
DSS should evolve over time in response to changing managerial levels of comfort and needs, increased data availability, and research advances [13], [23], [24]. Despite these dynamic aspects, there are relatively few published studies in the marketing of entertainment products reporting how models have actually evolved from both a technical and managerial standpoint (for exceptions, see [25], [26]). We demonstrate this here with a model, SilverScreener [21], developed initially to assist managers of a Dutch movie exhibition chain, Pathé, to schedule movies in a single theater with multi screens. Having established a level of comfort with decision support systems and models, Pathé subsequently asked the modeling team to assist them in scheduling movies in multi theaters with multiple screens, within a single city, each week. We discuss in this paper our experience in addressing this new challenges, impacting decisions, policies, and practices. More specifically, this paper reports how we modified the SilverScreener model for the multi-theater multiplex situation, how we made scheduling recommendations for a period of 26 weeks, and how we did all of this in close interaction with Pathé management. We present the results in terms of both how Pathé used the DSS' recommendations and the performance implications of the DSS implementation. We emphasize issues related to the interface of modelers and management. Therefore, we also pay attention to how the model was used in combination with the judgment of Pathé management, and how this particular multi-theater-multiplex scheduling application represents a specific stage in the adoption process of DSS-methodology by this movie company.
The evolutionary approach followed in this paper builds upon previous research proposing a general framework integrating the elements that determine the success of a DSS [27]. Two critical elements of the framework in [27] are the demand side (characterized by the decision problem, the decision environment and the decision maker) — and the supply side (encompassing the functionality of the DSS and the decision support technology used) In the case of Pathé, we have here a very challenging demand side of the decision support system, with, on the one hand, a relatively structured problem (clear decision variables, predictable outcomes) and rich data, but on the other hand, a decision environment characterized by a heuristic decision style, and heavy reliance on intuition. The challenge is how to develop a DSS fitting with this demand side, and to find an effective implementation strategy so that the DSS is actually adopted and used. Elements of this strategy are: 1) evolutionary model development; 2) combining hard data and the intuition of the manager (e.g., in the classification of new movies; in the option to overrule the recommendation of the DSS); and 3) in providing a quantitative measure of the monetary value of performance improvement through the DSS. Our evolutionary approach can be summarized as shown in Fig. 1.
As shown in the figure, the evolutionary approach involved the following steps. We first demonstrated that there exists a match (1) (room for improvement in decision-making) between the demand side and supply side of the DSS via ex-post analyses. This is based on demonstrating the effectiveness of our model on past data [21]. On the basis of these results, we established a relationship (2) with the organization through an internal champion. This champion is usually from the senior management. Gaining trust was relatively easy because of an earlier interaction when two members of the present SilverScreener team were involved in the successful launch of a DSS for the prediction of the number of visitors for new movies at one of the Pathé movie theaters in Holland [5]. In consultation with management, we agreed on meaningful metrics (3), and experimental setting to demonstrate our results in practice. We built the confidence of the management via implementation of our approach in simpler yet realistic settings, namely, in a single multi screens theater (4). We next proved the effectiveness of the approach in more complex situations, such as multiple theaters with multiple screens case (5). This is the topic of the present paper. The successful experience in this implementation will move us to the next level in complexity, namely, micro-scheduling (6) (i.e., within the theater scheduling movies showings for different hourly slots).
Related research on evolutionary development of DSS has appeared in contexts other than entertainment industry. [17] describes a multi-year effort, which resulted in the implementation of a series of human resource planning DSS applications in the U.S. Navy shipyard community. This paper concentrates on the development and implementation of a DSS in a large organization that is going through a personnel-downsizing process. [1] considers a customer-oriented catalog segmentation problem that addresses the crucial issue of the design of the actual contents of the catalogs. The DSS recommends alternative, satisfactory solutions to the decision maker. Using three algorithms, the DSS provides the decision maker with an easy-to-use, yet powerful tool to examine various catalog design options and their implications on the contents of the catalogs and the clusters of targeted customers. [2] models the constituents of a collaborative supply chain, the key parameters they influence, and the appropriate performance measures in a decision support environment. Their paper shows how the constituents, key parameters and performance indicators are modeled jointly into the environment.
To illuminate the setting of the current project, Table 1 lists the Pathé movie theaters in Amsterdam with their seating capacities (for each screening room) and Table 2 provides, as an illustrative example, the actual weekly schedule of movies in one of Pathe's movie theaters, De Munt, for the first eight calendar weeks of the year 2002. (See Table 3 for abbreviations and corresponding movie titles). The primary goal of the project described here was to provide, each week, recommendations for such movie schedules for each of the Pathé theaters in Amsterdam. This is called a macro-scheduling task. In the implementation section, presented later, we will provide movie theater specific results from three major movie theaters in Amsterdam-Arena, City, and De Munt. These three major movie theaters together comprise 34 of Pathé's 41 screens in Amsterdam. In fact, this covers most of the movie supply in Amsterdam, because Pathé owns all the major movie theaters in Amsterdam (90% of box office sales).
The movie macro-scheduling problem represents an area where DSS has a high potential in helping managers, but an unpredictable chance to succeed. While many of its managerial problems tend to be fairly structured, the decision environment is quite dynamic, contractual arrangements between parties are complex, and the cognitive style of the decision makers is often non-analytical or heuristic [27]. These characteristics represent challenges in developing implementable models for decision makers in this industry. Despite the above-noted challenges, a stream of research that addresses these and related issues in the area of movies is emerging. Forecasting, for example, has received an increasing amount of attention. Work has been reported on forecasting the enjoyment of movies at the individual level [6] and on predicting commercial success of movies at the aggregate level [5], [8], [16], [20], [21]. Other topics that have received research and modeling attention include release timing of movies and videos [10], [14], [10], assessing the impact of advertising on box-office performance of new films [28], and designing contracts in the film's supply chain [18]. However, the above mentioned research has taken an ‘one-shot’ type approach and no study to date has focused on developing and implementing decision support models and systems over time, working closely with managers in the movie exhibition industry and assisting them in their decision-making. This paper reports the implementation efforts aimed at fulfilling this gap.
The remainder of the paper proceeds in the following manner. Section 2 presents a more detailed description of the problem. Section 3 presents the development of the multi screens macro-scheduling algorithm used. Section 4 describes our forecasting process for weekly attendance and our analysis of the accuracy of the system. Section 5 provides an evaluation of the success of the implementation, and the last section (Section 6) deals with the lessons learned for the implementation of DSS in “non-traditional” domains.
Section snippets
Problem description
Every week, movie distributors typically have 3 to 5 new movies available for release into the market and movie exhibitors need to decide which, if any, of those movies to show in their movie theaters, and which old movies to stop showing if necessary. While exhibitors at times make commitments months in advance to show a specific movie, typically for blockbusters such as, Lord of the Rings, most exhibitors have a management meeting every Monday morning to review the past weekend's box office
Development of the scheduling algorithm
We begin the exposition with the formulation of the basic theater programming problem, which was solved every week for the six movie theaters considered in this study. The scheduling algorithm, an integer programming problem which optimizes each theater's net margin over a (rolling) planning horizon of W weeks, is built on a core theater programming model (see Appendix A for a detailed description). System-wide constraints imposed because of the multiple theaters operated by Pathé have been
The demand model and seasonality
To forecast weekly attendance for a movie at the individual movie theater level, we used an exponential decay model. Previous researchers have shown that, to a first degree of approximation, most mass-market movies follow an exponentially decaying pattern once it has been widely released [12]. For a given movie, the exponential model, stated in logarithmic form, is the following:
The values of the two parameters α
Evaluation of the SilverScreener DSS' performance
The performance evaluation of the DSS is an extremely important issue. Clear and operational measures have to be established before the implementation begins. It was decided jointly by Pathé's management and the modeling team that the evaluation of this application of the SilverScreener system would be based on output metrics, such as attendance and net margin, as well as on behavioral measures, such as, the extent to which managers followed the DSS recommendations, and the extent to which they
Decision support systems in new domains
As mentioned earlier, the motion picture industry is an unusual application area for DSS. The organizational culture does not favor mathematical models, and the cognitive models of the decision makers are heuristics rather than analytical. Furthermore, the products, that is, the movies are changing all the time and their demand is highly uncertain and location specific. In such an environment it is not easy to implement decision support systems, and it requires a lot of effort on the part of
Acknowledgement
The authors thank the management of Pathé chain of movie theaters for their help and cooperation. The authors gratefully acknowledge the computational support provided by Sumit Raut.
Sanjeev Swami has joined Faculty of Social Sciences, DEI, Agra, India as Professor and Head, Department of Management. Prior to joining DEI, he was Associate Professor in Industrial and Management Engineering at IIT, Kanpur, India. He holds an undergraduate degree in Production and Industrial Engineering from University of Allahabad, a master in Industrial and Management Engineering from IIT, Kanpur, and Ph.D. in Marketing from University of British Columbia, Canada. He is a doctoral consortium
References (28)
Customer-oriented catalog segmentation: effective solution approaches
Decision Support Systems
(2006)- et al.
A model and a performance measurement system for collaborative supply chains
Decision Support Systems
(2006) Evolution of the strategy and structure of a human resource planning DSS application
Decision Support Systems
(July 1995)- et al.
Database models and managerial intuition: 50% models and 50% manager
Management Science
(1990) Spatial Competition in Retail Markets: Movie Theaters, forthcoming RAND
Journal of Economics
(2008)- et al.
MOVIEMOD: an implementable decision support system for pre-release market evaluation of motion pictures
Marketing Science
(2000) - et al.
Modeling goes to Hollywood: predicting individual differences in movie enjoyment
Management Science
(1994) - et al.
Implementing and evaluating SilverScreener: a marketing management support system for movie exhibitors
Interfaces: Special Issue on Marketing Engineering
(2001) - et al.
Film critics: influencers or predictors?
Journal of Marketing
(1997) - et al.
AMPL: A Modeling Language for Mathematical Programming
(1993)
The last picture show? Timing and order of movie distribution channels
Journal of Marketing
Combining models with intuition to improve decisions
Competitive dynamics and the introduction of new products: the motion picture timing game
Journal of Marketing Research
Building Models for Marketing Decisions
Cited by (14)
Determining the optimal release time of movies: A study of movie and market characteristics
2023, Decision Support SystemsCitation Excerpt :Our paper thus provides a much-needed decision framework for policymakers. In this context, our paper contributes to the study of decision support systems for the movie industry [5,11]. The paper is organized as follows.
Executive functions and decision making: A managerial review
2013, IIMB Management ReviewCitation Excerpt :Third, the experts want the power to override the model in case its use is clearly inappropriate (as during a stock market crash). Therefore, it has been suggested that a 50–50 combination of a bootstrap model and top experts is best (Blattberg & Hoch, 1990; Eliashberg, Swami, Weinberg, & Wierenga, 2009). However, even bootstrap models have their limitations.
Using online search data to forecast new product sales
2012, Decision Support SystemsCitation Excerpt :The first week of release is often the most crucial for motion picture revenues, and is also the most difficult to forecast. Therefore, pre-launch forecasting will provide several useful implications to managers for scheduling, resource allocation, forecasting, etc. [2,8]. Given the characteristics of online search terms that are mentioned earlier and these characteristics of the motion picture industry, we think it is a good area to begin investigating the potential of search term activity as a forecasting measure.
Managerial decision making in marketing: The next research frontier
2011, International Journal of Research in MarketingThe economics of movies (revisited): A survey of recent literature
2023, Journal of Economic SurveysApplying option thinking to value experiential marketing content
2023, Journal of Media Economics
Sanjeev Swami has joined Faculty of Social Sciences, DEI, Agra, India as Professor and Head, Department of Management. Prior to joining DEI, he was Associate Professor in Industrial and Management Engineering at IIT, Kanpur, India. He holds an undergraduate degree in Production and Industrial Engineering from University of Allahabad, a master in Industrial and Management Engineering from IIT, Kanpur, and Ph.D. in Marketing from University of British Columbia, Canada. He is a doctoral consortium fellow of American Marketing Association, a recipient of University Graduate Fellowship at UBC, and career awards from Department of Science and Technology, India and All India Council of Technical Education. His research work has been published in Marketing Science, International Journal of Non-profit and Voluntary Sector Marketing, Journal of Operational Research Society, DEIJSER, M&SOM, Interfaces, Marketing Letters, and Vikalpa.
Jehoshua (Josh) Eliashberg is the Sebastian S. Kresge Professor of Marketing and Professor of Operations and Information Management, at the Wharton School of the University of Pennsylvania. His research interests are in developing models and methodologies to solve business problems. His research has focused on various issues including new product development and feasibility analysis, marketing/manufacturing/R&D interface, and competitive strategies. He has particular interest in the media and entertainment, pharmaceutical, and the hi-tech industries. He has authored numerous articles appearing in major journals such as: European Journal of Operational Research, Group Decision and Negotiation, Interfaces, Journal of Economic Psychology, Journal of Marketing, Journal of Marketing Research, Management Science, Manufacturing and Service Operations Management, Marketing Science, and Optimal Control Applications & Methods. His work in the entertainment industry has been the subject of articles appearing in Businessweek, The Christian Science Monitor, The Financial Post, Financial Times, Forbes, Fortune, Los Angeles Times, The New York Times, Variety, Newsweek, The Wall Street Journal, and The Washington Post.
Berend Wierenga is Professor of Marketing at the Rotterdam School of Management, Erasmus University. His main research area is marketing decision making and marketing decision support. Together with Gerrit van Bruggen he published a book on Marketing Management Support Systems (Kluwer Academic Publishers 2000). He has (co-) authored several articles on decision support systems in the domain of marketing (e.g. Journal of Marketing, Marketing Science, International Journal of Research in Marketing, Management Science), as well as in the general IS/DSS domain (e.g. MISQ, Communications of the ACM, and Decision Support Systems). Berend Wierenga is also the Editor of the Handbook of Marketing Decision Models (2008), published by Springer Science + Business Media.
Charles B. Weinberg is the President of SME Vancouver Professor of Marketing at the Sauder School of Business, University of British Columbia, Vancouver, Canada. In 2008, he was selected as one of the first ten fellows of the INFORMS Society for Marketing Science. His research focuses on analytical marketing, services, and public and nonprofit marketing. His work in the nonprofit sector includes pricing, the marketing of safer sex practices, portfolio management and competition. For more than 30 years, he has studied the arts and entertainment industries. His early work focused on live entertainment and included the ARTS PLAN model for marketing and scheduling performing arts events. More recently, he has focused on the movie industry in which he has studied such issues as competitive dynamics, scheduling of movies into theaters, sequential release of movies and DVDs, and contract terms. He is a former editor of Marketing Letter and former area editor of Marketing Science. He served as chair of the 2008 Marketing Science conference in Vancouver.