globally distributed r&d work in a marketing management ... · globally distributed r&d...

Journal of Advances in Management ResearchVol. 5 (I) 2008 (pp. 7-20)

Globally Distributed R&D Work in a Marketing Management SupportSystems (MMSS) Environment: A Knowledge Management Perspective

Jehoshua Eliashberg1

[email protected]

Sanjeev Swami2

[email protected]

Charles Weinberg3

[email protected]

Berend Wierenga4

[email protected]

ABSTRACT

Globalisation, liberalization and rapid technological developments have been changing business environments drastically inthe recent decades. These trends are increasingly exposing businesses to market competition and thus intensifying competition.In such an environment, the role of marketing management support systems (MMSS) becomes exceedingly important for thelong-term growth of an organisation�s marketing expertise and success. In this paper, we discuss the evolution of a globallydistributed R&D project spanning three continents in developing an MMSS for the motion picture industry. We first provide theconceptual background of the MMSS and knowledge management systems relevant for our work. We then provide a detailedcase study of our MMSS implementation. We specifically focus on the following elements of our work: globally distributedR&D efforts, knowledge elements, and fit between demand and supply sides of MMSS. We conclude with a discussion ofimplications for future research in this area.

Keywords: Globally Distributed Work, Marketing Management Support Systems, Knowledge Management

INTRODUCTION

Deregulation (liberalization), globalisation, and rapidtechnological developments have been changing businessenvironments drastically in the recent decades. Thesetrends are exposing businesses to market competition ata greater extent. The intensified competition inducesfirms to aim for a more targeted product line, greaternumber of merger and acquisitions, and experiment witha variety of delivery channels. As firms race toward morecompetitive positions, knowledge becomes a significant

1 Jehoshua Eliashberg is Sebastian S. Kresge Professor of Marketing,and Professor of Operations and Information Management, WhartonBusiness School, University of Pennsylvania.2 Sanjeev Swami is Associate Professor of Marketing and OperationsManagement, Department of Industrial and ManagementEngineering, Indian Institute of Technology, Kanpur, India (currentlyon leave as Professor & Head, Department of Management, DEI,Agra, India)3 Charles B. Weinberg is Alumni Professor of Marketing, SauderSchool of Business, University of British Columbia4 Berend Wierenga is Professor of Marketing, Center for InformationTechnology in Marketing (CIT/M), Rotterdam School ofManagement, Erasmus University, The Netherlands.

8 Journal of Advances in Management Research vol. 5(I) pp. 7-20, 2008

GLOBALLY DISTRIBUTED R&D WORK IN A MARKETING MANAGEMENT SUPPORT SYSTEMS (MMSS) ENVIRONMENT

driver of competitive advantage under a globallyderegulated business environment.

With the easing of national barriers, managingknowledge has become essential to accessing timelyinformation about international competitiveenvironments, regional growth rates, and economic andcultural issues for building a global business portfolio.The penetration of the Internet has been one of the majorcatalysts that are speeding up this process. With theincrease in virtual collaboration and remote teamingamong highly distributed teams across the globe, thepartnering firms need both explicit and tacit knowledgesharing. Businesses that were once organized alonggeographic lines are now reorienting themselvesaccording to markets, products, and processes.Companies such as Lotus, Verifone, and Microsoft areusing this phenomenon to their advantage by shifting�mental labor� intensive software development to theirprogrammers in India and Russia (Tiwana 2000). Forexample, in year 2000, 185 of Fortune 500 companiesoutsourced software development to India alone and theamount of outsourcing grew at a 53% yearly rateaccording to report by the National Association ofSoftware and Service Companies (Mockus and Herbsleb2001).

In such a dynamic and competitive environment, themanagement of R&D as well as marketing functionsbecomes very important. It is challenging to managediverse R&D teams spread across geographicalboundaries. One effective way to increase the efficiencyof the marketing function is to make good use ofmarketing management support systems (MMSS) for thelong-term growth of the organisation. In this paper, wereport the evolution of a globally distributed R&D workspanning three continents in developing an MMSS forthe motion picture industry.

The R&D work reported here is collaborative in naturebetween industry and academia, in which a company (tobe discussed in detail in later sections) identified a seriesof scheduling problems whose solutions were developedby the university-based scientists. From a knowledgemanagement perspective, therefore, we make a distinctionbetween academic knowledge (R&D work) and practice.Moreover, the type of knowledge generated by thecollaborative work is a special kind of knowledge froman organizational perspective. This is in contrast withthe R&D knowledge that is generated completely �in-house� or �outsourced� by the organization.

We first provide a theoretical background in the area

of developing analytical MMSS tools in a knowledgeenvironment. We then provide a detailed account of thecase study of MMSS implementation. In discussing thecase study, we focus on the following perspectives: (i) aglobal resource generation perspective, (ii) fit betweenthe supply and demand sides of the MMSS, and (iii)knowledge management. We conclude with a discussionof future research implications in this area.

CONCEPTUAL BACKGROUND

MMSS Implementation and Challenges

A marketing management support system (MMSS) isdefined as any system combining (1) informationtechnology, (2) analytical capabilities, (3) marketing data,and (4) marketing knowledge, made available to one ormore marketing decision maker(s) to improve the qualityof marketing management (Wierenga and van Bruggen2000). Since this definition is at a fairly general level ofspecification, it encompasses a variety of tools andsystems, such as, models, information systems, DSS,expert system, and knowledge-based systems. A list ofthe types of MMSS�s and their summary characteristicsis provided in Table 1.

Knowledge Management

What is Knowledge Management?

Knowledge is a fluid mix of framed experience, values,contextual information, expert insight and groundedintuition that provides an environment and frameworkfor evaluating and incorporating new experiences andinformation. It originates and is applied in the minds ofknower. In organizations, it often becomes embedded notonly in documents or repositories, but also inorganizational routines, processes, practices, and norms.Knowledge management, therefore, enables the creation,communication, and application of knowledge of all kindsto achieve business goals.

Types of Knowledge

There are two major types of knowledge in anorganization: tacit and explicit knowledge. Tacitknowledge is personal, context-specific knowledge whichis difficult to formalize, record, or articulate, whereasthe explicit knowledge is well-documented, codified andtransmitted in a systematic and formal language throughdocuments, databases, web, w-mails, and so on.

Journal of Advances in Management Research vol. 5(I) pp. 7-20, 2008 9


What Can Be Leveraged in Knowledge Management?

There are a number of useful leveraging elements inknowledge management, such as (Tiwana 2000): (i)collaboration and collaborative knowledge sharing, (ii)conversation as a medium for thought, (iii) easy to findsources of the technical know-how, locate people andexpertise, and reuse what exists either in a tangible form,or in people�s minds, (iv) ability to provide decisionsupport, (v) flexibility and scalability, (vi) pragmatism,(vii) user-orientation, and (viii) ease of use. We elaboratebelow on another important leveraging element, namely,the Internet.

Leveraging the Internet

A quick glimpse of existing technology in mostcompanies reveals a transformative addition � theInternet. It provides integration of the islands ofinformation across the organizational landscape. Certaincharacteristics of the Internet have made it an inevitablechoice for globally distributed R&D projects. The globalreach of the Internet has the following key implications:(i) a cost-effective global network backbone, (ii) ubiquity,(iii) distributed connectivity (i.e., distributed resourcesand databases can be interconnected cost effectively andreliably, using virtual networks), (iv) robust global data

Table 1: Characteristics of Marketing Management Support Systems

Type of MMS Characterizing Keywords

Marketing Models (MM) · Mathematical representation· Optimal values for marketing instruments· Objective· Best solution

Marketing Information Systems (MKIS) · Storage and retrieval of data· Quantitative information· Registration of �what happens in the market�· Passive systems

Marketing Decision Support Systems (MDSS) · Flexible systems· Recognition of managerial judgment· Able to answer �why� questions (analysis) and �what-if� questions

(simulation)

Marketing Expert Systems (MES) · Centers on marketing knowledge· Human experts· Rule-based knowledge representation· Normative approach: best solution

Marketing Knowledge-Based Systems (MKBS) · Diversity of methods, including hybrid approaches· Structured knowledge representation, including frame-based

hierarchies· Model-based reasoning

Marketing Case-Based Reasoning Systems · Similarity with earlier cases(MCBR) · Storage of cases in memory

· Retrieval and adaptation· No generalization

Marketing Neural Networks(MNN) · Training associations· Pattern recognition· No a priori theory· Learning

Marketing Creativity Support Systems (MCSS) · Association through connections· Idea generation· Endorse creativity in problem solving



path, (v) cheaper, faster, and directly usable globalcompetitive intelligence, and (vi) platform independencein terms of a number of data formats and platform-dependent forms.

CASE STUDY: AN EVOLUTIONARYDEVELOPMENT OF AN MMSS IN MOVIEINDUSTRY

Building on the discussion so far, we now discuss theevolution of a globally distributed R&D project spanningthree continents in developing an MMSS for the motionpicture industry. The motion picture industry representsan area where MMSS have a high potential for helpingmanagers, but an unpredictable chance to succeed. Thisis because the decision environment is quite dynamic,contractual arrangements between parties are complex,management turnover appears to be high, and, perhaps,most importantly, the cognitive style of the decisionmakers is often non-analytical or heuristic (Eliashberget al. 2001; Wierenga, Van Bruggen, and Staelin, 1999).These characteristics represent challenges in developingimplementable models for decision makers in thisindustry. In this section, we present an evolutionaryaccount of the development of an MMSS in this industry,starting with a scheduling model, namely, SilverScreener(Swami, Eliashberg, and Weinberg 1999), and evolvinginto varied applications (Eliashberg et al. 2001;Eliashberg et al. 2006a).

The Problem Hierarchy

We first present the major entities of a typical supplychain in motion picture industry (refer Figure 1). Movies

are, in general, produced and distributed by major studios(e.g., Paramount, Warner Brothers), and played by anexhibitor (e.g., United Artists). An exhibitor�s schedulingproblem can be summarized as: which movies to chooseto show each week and for how long to play them.

This exhibitor�s problem can be addressed at twolevels: macro-scheduling (a weekly scheduling problemat the entire theater chain level) and micro-scheduling (adaily scheduling problem at an individual theater level).Figure 3 shows these views for a multiplex chain. Themacro level problem, when solved by SilverScreenermodel, recommends a weekly list of recommendedmovies for each theater. A Senior Manager of theaterprogramming at the chain�s Central Office uses thismodel�s recommendations to decide on a list of moviesto be played for the coming week for each theater screen.The micro level problem is then solved for within theweek, or daily planning for individual theaters.Pictorially, Level 1 or macro planning view can bepresented as shown in Figure 2. Based on a considerationset of movies available from distributors in the market,the planner selects a subset of movies for scheduling inthe coming weeks. The output of macro problem plussome additional parameters such as screen capacities,and demand forecasts, are used as inputs to the Level 2problem of micro-scheduling. The output from thisprocess provides a screening schedule of movies (screen,time slot, movies) during different times of the day.

Knowledge Elements

In the MMSS implementation project in the movieindustry, we have used several knowledge elements aslisted below.

Fig. 1: Motion Picture Supply Chain

Motion Picture Supply Chain

DistributorDistributor ExhibitorExhibitor AudienceAudienceAudience



Knowledge about the movie industry

A comprehensive review of the state-of-the-art researchin the movie industry appears in Eliashberg, Elberse andLeenders (2006). A rapidly emerging stream of researchin this field addresses various research aspects, such as,scheduling, forecasting, contract design, consumerbehavioral aspects, and online recommendation systems.Movies are considered as one of the most dynamicindustries in many countries/cultures due to theemergence of various innovations and technologies, suchas, digital production and exhibition, and newer retailformats, such as multiplexes or megaplexes (sci-tech-today.com 2005, Jardin 2005). Movies are interestingbecause of their experiential nature, which requires co-creation by both producer and consumer to consummatethe consumption. This aspect raises distinctive challengesbecause the same product (i.e., movie) may appealdifferently to different people making the task of fore-casting movie�s demand or success probability a difficultone (Eliashberg, Jonker, Sawhney, and Wierenga 2000).

A key property of the demand pattern of majority of

movies, that renders complexity to several moviescheduling problems, is exponential decay of demandover time (Krider and Weinberg 1998). This demandperishability view for multiple movies, coupled with asliding scale box-office revenue-sharing contract betweenthe distributor and the exhibitor, makes moviesscheduling problem a non-trivial one (e.g., Swami,Eliashberg, and Weinberg 1999). As a contrast, ininventory control problems, perishability is typicallyconsidered in terms of physical deterioration of a product(e.g., a grocery item with an expiry date). This viewcomes from the supply side of the product. The moviescheduling problem adopts a �demand side view� inwhich the physical product (i.e., a copy of the movie)remains the same, but its demand perishes over time.

Modeling in Operations Research and Implementation

Several operations research models were developed andimplemented at different locations of the Pathe theatersas part of this project. A brief description of these isprovided in the sections below.

Fig. 2: A Schematic View of Macro and Micro Versions of the Exhibitor�s Problem

Micro-SchedulingTheater �1�

Number of Screens = 14

Micro-SchedulingTheater �n�

Number of Screens = 6

Theater Chain Central Office: Senior Manager, Theater Programming

Macro-Scheduling Weekly Recommended List

(Theater Wise)

List for coming week for each theater screen

Level I: Macro Scheduling Problem

Total Availability Set

(Distributors)

INPUT: Availability set during the period

OUTPUT: Selected set for screening in the multiplex for the planning

period

SelectedSet

Selection and Scheduling? ! SilverScreener Model (Swami, Eliashberg and Weinberg, Marketing Science, Spl. Issue: Managerial Decision Making, 1999)

CONTRACT

�Output of Macro-Scheduling Problem�

Screens

T

I

M

E

M2

M1M4M3

M5

OUTPUT: Daily Schedule (Within Week)

INPUT: Capacity of screens, Demand forecasts

SelectedSet



Macro Planning: SilverScreener Model

SilverScreener is aimed at helping exhibition managersmake better decisions by using a mathematicalprogramming approach and a fast, but readily accessiblealgorithm (Swami, Eliashberg, and Weinberg 1999). Themathematical program used is an integer linear program.The major objective of the program is to help select andschedule movies for a multiple-screens theater over afixed planning horizon in such a way that the exhibitor�scumulative profit is maximized. Swami, Eliashberg, andWeinberg (1999) propose a two-tier integrated applicationof the model. The first tier involves development of amovie selection plan to help the manager plan an entireseason and bid for movies before the start of that season.The second tier � adaptive scheduling � of the integratedapproach helps the exhibitor in weekly decision makingduring the season. This application involves �rolling,�and updating data, from one week to the next. Eachweek�s attendance data can be used in the model. Swami,Eliashberg, and Weinberg (1999) provide variousanalyses of normative versus actual decision making,based on publicly available data.

Implementation of SilverScreener Model at a Single-Screen Theater

SilverScreener was implemented and used by Holland�sleading movie theater company (Pathé) for a singlelocation (Eliashberg, Swami, Weinberg, and Wierenga2001). The model was implemented on a weekly basis

using the adaptive scheduling approach of SilverScreenerfrom Week 45, 1999, to Week 8, 2000.

Multi-Theater Application of SilverScreener Model atPathe, Holland

Pathé management, having gained confidence in themodel and the modeling team on the basis of the single-location implementation, asked us to develop a systemthat would be available every Monday morning to makerecommendations as to which movie should be scheduledfor each screen across all their movie theaters inAmsterdam (refer Eliashberg, et al. 2006a). For this morecomplex assignment, the following additional issues hadto be taken into account in the mathematical model: pre-commitments (e.g., contractual agreements withdistributors to show a particular movie in a particularscreening room of a particular theater during a specifiednumber of weeks); special treatment for kid�s movies andmatinee movies; cannibalization of a movie�s sales inone theater by its revenues in another theater; andcomplexity rendered by different numbers of screens inthe different theaters, different screening room capacities,different demand situations (translating into differentprognoses for box office sales for the same movie indifferent theaters, and different consideration sets (setsof possible movies to be shown). The algorithm was basedon solving the above multi-theater problem as a separableset of single movie theater problems. A conceptual viewof our implementation plan, as explained above, is shownin Figure 3.

Week

1 . . . . . . . 8

1 . . . . . . . . ScreenSt

Artificial screens for kids� movies, matinee movies

Pre-commitments

Single-theater SilverScreener algorithm applied to empty cells

Movie Theater 1 Movie Theater T

Fig. 3: Conceptual View of Multi-Theater Screen Scheduling Implementation



Micro-Scheduling Algorithm: From SilverScreener toSilverScheduler

As movie theaters have evolved from single screentheaters to multiplexes (sometimes involving 20 or morescreens), the problems of scheduling the movies ontoscreens has become increasingly complex (refer Figure4). Some of the key reasons for that complexity are asfollows: (i) there is a large number of different movieswith different running times that the theater wants to showin a typical week, (ii) the number of seats per screeningroom differs, (iii) there is variability in demand formovies, with respect to the movie type, time of the dayand day of the week, (iv) there are different genres ofmovies, which has implications for their scheduling, (v)constraints posed by the logistics of a movie theater, suchas, opening and closing times and cleaning times, and(vi) management specified constraints (e.g., the timebetween two movie starts should never be more than 20minutes). The problem considered in the micro-scheduling algorithm is how to make a schedule of thedifferent movies on the different screens during a givenweek, which obeys all the constraints and maximizes thenumber of visitors.

Based on the mathematical programming techniquesof integer programming, and column generation, we havedeveloped a procedure that produces an optimal movie

program for a given week, given the list of movies to beshown, their running times, demand, the capacities ofthe different screening rooms, information aboutcontractual agreements, and any other constraints, asapplicable. This system is labeled as SilverScheduler.

Use of Statistics and Forecasting Approaches

SilverScreener Model

The SilverScreener model makes use of an ex anterevenue prediction scheme for the first tier, movieselection plan. This is based on the attributes of themovies, such as, genre, stars, distributors, etc., and alsomakes use of the analogical reasoning, such as, a�matching� movie from previous seasons. The secondtier, the adaptive scheduling approach, makes use of theactual data available about a movie to use it in a two-parameter exponential revenue prediction model.

Implementation Applications

In the implementation cases, there was the issue ofmaking forecasts for newly released movies, for whichno box office data are yet available. In our one-theaterstudy with only 6 screens, we simply asked managementto make judgmentally numerical predictions, based on

1 61 4 :0 0

H P O V1 51 3 :5 0

1 41 3 :4 0

Z L1 31 3 :3 0

S N D TO TD S A W1 21 3 :2 0

L O T R1 11 3 :1 0

M N S 1 01 3 :0 0

A S H O 1 191 2 :5 0

L O T R81 2 :4 0

H P N L 71 2 :3 0

P D D O H 61 2 :2 0

Z L 51 2 :1 0

S N D T 41 2

O T 31 1 :5 0

21 1 :4 0

H P O V11 1 :3 0

9 09 63 8 21 7 71 7 51 7 21 6 31 6 11 0 21 1 33 4 02 2 22 2 2C ap acity

1 31 21 11 0987654321S creenT im e

Fig. 4: A Representative View of Micro-Scheduling Problem Solution



their experience, or to provide a �matching movie,� amovie that is similar to the new one, and for whichhistorical data are available. In the second application,for the larger number of theaters and screens, we deviseda movie classification scheme, which utilizes managerialjudgmental information, without the manager having tomake numerical forecasts for individual movies.

A key input to the micro-scheduling algorithm is thedemand information about the movies in the movie list.For each movie in the movie list we need to have aforecast of the number of visitors that this movie willgenerate in a showing. This number has to be availablefor each day of the week and for each different startingtime of the movie (using intervals of one hour). For thispurpose, we developed a forecasting procedure with twomodules. The first module is for movies that have beenrunning already. The observed numbers of visitors areused to estimate a forecasting model. The second moduleis for newly released movies, where the numbers ofvisitors are forecasted in an indirect way, using thecharacteristics of a movie as determining variables.

Computational capabilities

The model developments, as well as implementation

approaches, made extensive use of the state-of-the-artcomputing resources, such as fastest PC�s and UNIXsystems, various software (e.g., AMPL/CPLEX, C, C++,MS-Office), and broadband internet.

Fit between Demand and Supply of MMSS

As mentioned earlier, the fit between demand and supplyis essential for a successful MMSS. First, by globalsourcing of R&D talent, we have been able to providethe best fit possible, both in terms of expertise andtimeliness. Then, by the local presence of one of the teammembers near the potential users helped facilitation ofdirect and personal interaction. Further, ourimplementation strategy was guided by Wierenga, VanBruggen and Staelin�s (1999) framework, which relatesthe success of a MMSS to a number of factors: demandside factors, supply side factors, the match betweendemand and supply, design characteristics of the MMSSand characteristics of the implementation process (seeFigure 5).

The demand side of the MMSS is movie programmingat Pathé, The Netherlands. As noted earlier, Pathé is thelargest movie theater company in the Netherlands Witha very high market share. The Programming Department

D ecision S itu ation C h aracteristics

D ecision P roblem• S tructuredness• D epth of know ledge• A vailability of data• M arketing instrum ent

D ecision E nvironm ent• M arket dynam ics• O rganizational culture• T im e constraints

D ecision M aker• C ognitive style• E xperience• A ttitude tow ards D S S

M P S M

M ark etin g M an agem en tS u pp ort S ystem (M M S S)

C om ponents

• Inform ation technology• A nalytical capabilities• M arketing data• M arketing know ledge

F unctionality

• O ptim ization• A nalysis & diagnosis• S uggestion & stim ulation

T ypes

• M M M K B S• M K IS M C B R• M D S S M N N• M E S M C S S

M atch b etw een D em and & S u p ply of D ecision S u p port

D esign C h aracteristicsof the M M S S

• A ccesibility• S ystem Integration• A daptability• P resentation of outputand user interface

• S ystem quality• Inform ation quality

C h aracteristics of th eIm p lem en tation P rocess

• U ser involvem ent• T op m anagem ent support• C om m unication about the M M S S• M arketing orientation• P resence of an M M S S cham pion• A ttitude of the IS departm ent• In-com pany developed versus purchased• T raining of the users

S u ccess M easu res

T echnical V alidity

A doption and U se

U ser Im pact V ariables

• . U ser satisfaction• . P erceived usefulness• . D ecision confidence

O rganizational Im pact V ariables

• P rofit• S ales• M arket share• T im e saved• P ersonal productivity• C ost reductions

D em andSideofD ecisionSupport

SupplySideofD ecisionSupport

M P S M

M M S S2

3

5

6

4

1

Judging Implementation Success Judging Implementation Success (WVS 1999)(WVS 1999)

Fig. 5: Integrative Framework of the Factors that Determine the Success of a Marketing Management Support System(Wierenga, Van Bruggen, and Staelin 1999)



of Pathé chooses which movies to play in which theater(s)and on which screen, in a given week. In the Netherlandsa new movie week, starts every Thursday. Every Mondaymorning a team of three people meets and jointly preparesa movie allocation schedule for all the Pathé screens inthe country for the following movie week, effective thenext Thursday.

Using the implementation framework, the followingstrategy was determined citation:

� An environment was chosen with a relativelyfavorable initial attitude towards a marketingmanagement support system.

� An excellent technical match was made betweenthe decision problem and the marketingmanagement support system.

� Given the low experience with models (andanalytical methods in general) in the organizationalenvironment, the MMSS was operated externally.

� The SilverScreener�s recommendations wereaccessible instantly through the Internet to thedecision makers, and presented in a very user-friendly way.

� Constant involvement of the users was maintainedduring the process.

Role of Enabling Technologies

The timing of events is of immense importance in theadaptive scheduling applications of the type discussedhere (see Figure 6). The coordination of theimplementation involved activities spanning threecontinents � North America, Asia and Europe. Theresearch collaborators on this project are based in theU.S., India, Canada, and The Netherlands. While theempirical analysis, execution of the model and the finalrecommendations were conducted each week at theIndian Institute of Technology, Kanpur, India, theimplementation site is in The Netherlands. At the sametime, frequent discussions were occurring among theresearchers via e-mail across the three continents.Therefore, proper coordination of activities in this projectrequired continual use of the information and Internet-based technologies. Figure 4 shows the occurrence ofdifferent events in a Movie Week for Movie Weeks 45,46, and 47 in 1999.

In our projects, the analytical solutions were literallycarried out nearly half way around the world from theimplementation site. With communication costs virtuallyzero, and therefore geographic proximity not a

requirement, a diverse team of researchers has beenassembled to address a challenging problem. Thus, oneimportant role of the internet is to make it easier toincrease the size of a team with less co-ordination effortthan is required otherwise. Further, the internet facilitatednearly 24 hour work cycles, because the time differencebetween India and Holland allowed us to wait for latestweekend data before transmitting the schedules toHolland just-in-time for the manager�s Monday-morningdiscussions.

The Information and Communication Technologies(ICT), therefore, open new possibilities for marketingmanagement support systems: the option of centralizedexpert centers with decentralized applications. Thevarious documents formats, and their accompanyingtechnologies used in our projects are shown in Figure 7.As our project implementations have shown, it is notnecessary to have highly qualified model builders or evenhigh-level software at the physical location of theapplication. Even ongoing optimization can take placefrom a very distant place. A company does not necessarilyhave to own or operate an MMSS itself, but maysubscribe to an MMSS service from a place elsewherein the world. This brings very sophisticated MMSS withinthe reach of even small companies who do not have theresources to acquire the expertise and the softwarethemselves.

Success Measures

Our R&D efforts provided different results at differentstages of the project. These are explained below.

SilverScreener Model Results

Our results suggested the following: (i) the exhibitor canachieve substantially higher cumulative profit, (ii) theimprovement over actual decisions in terms ofprofitability appears to result from a combination of bothbetter selection and improved scheduling of the movies,and (iii) the general structure of the optimal normativepolicy in a representative setting is a prescription to focuson fewer �right� movies and run them longer (Swami,Eliashberg, and Weinberg 1999).

Implementation Results

In the case of single-theater implementation, the facevalidity of our recommended schedules to Pathe was very



Fig. 6: Time-Line of Transmission of Data and Recommendations

high with at least 5 of the 6 weekly recommended moviesmatching the actual schedule every week. Theprofitability analysis of the model, which was done on aweekly basis, showed that the change in attendance atthe model-aided theater was much higher than that forthe other two theaters. Using Schultz and Slevin (1975)scale, we also measured the attitude towardsSilverScreener among the members of the PathéProgramming Team, at three different times: before theimplementation, after the first effective implementationperiod, and after a change in management. Our results

showed that, for all the managers involved, the attitudestowards SilverScreener became more favorable over time.

In the case of multi-theater implementations, ourresults were projected to an annual improvement inoverall net margin of approximately � 700,000. Toexamine the extent to which SilverScreenerrecommended movies played in the same capacity screenand movie theater as that actually scheduled by themanager, we examined the metric of weighted capacitymatch (weighted by screen capacity) for theimplementation period. The average match percentages



Fig. 7: Various Data Formats and Technologies Used in the Project

were of the order of 60% across movie theaters, andweeks. As theater managers have knowledge which isnot incorporated in the MMSS, we believe this is areasonable outcome.

Global and Evolutionary R&D Perspective

In our project, we dealt with a very local problem, namely,scheduling problems of Pathe in Amsterdam, but througha global perspective, we mobilized resources andknowledge from around the world. The expansion of ourR&D network from a three-member team to eightmembers, across several countries (USA, Canada, The

Netherlands, and India) and specializations (marketing,operations, management science, mathematics) is shownin Figure 8. An evolutionary view of our applicationefforts in this domain is shown in Figure 9 (the last namesof the research collaborators form the acronym mentionedfor each work).

From our project implementations, it became clearthat, as the work became more complex, additionalexpertise was needed. With an 8-person team, it is literallytrue that some of the people have never met each other.We anticipate that, in such an organization of R&Defforts, the role of delegation and smaller team formationwill become important.



Fig. 8: A Geographical Network View of R&D Collaboration in the Project

SilverScreenerAlgorithm (SEW 1999)

Single Site in The Hague (ESWW 2001)

Multiple Locations in Amsterdam (ESWW 2006a)

Micro scheduling (Work in progress at Pathe, EHHHMSWW 2006b)

Fig. 9: An Evolutionary View of Implementation Efforts

CONCLUSIONS

Current Status

In this project, the current status of the work involvesimplementation of the micro-scheduling algorithm atPathe. Additionally, a USA theater chain has shown keeninterest in the forecasting and scheduling applicationsof SilverScreener. A pilot implementation of our macroMMSS is currently underway at this chain.

Future of MMSS

We are optimistic about the future of MMSS in variousindustries. We believe that Pathé management�s requestfor several extensions of the original model indicates that

modeling is now seen as a way to help address difficultissues. We are also optimistic about the effect of thecurrent ICT possibilities on the further adoption and useof MMSS. They will enable the global rollout of newtools and systems in a very short period of time.

There are several options to be considered for futurelarge-scale dissemination of these MMSS�s. Oneattractive alternative is to develop a user-friendly desktoptool for the managers in this industry. This would beginwith getting users� specifications for how exactly theywould like the tool to be designed for them. Althoughthe managers in this industry do not use model-buildingas a routine exercise, they extensively use spreadsheetsoftware for internal reporting purposes. If a tool couldbe designed that appears to be no more complicated than



using Excel, then the ensuing interactivity and diagnosticswill provide the user a greater sense of control andownership.

Another possibility is to use an application serviceprovider (ASP) business model that provides computer-based services to customers over a network as on-demandsoftware. This could be done in the most limited senseusing a standard protocol such as HTTP. The need forASPs has evolved from the increasing costs of specializedsoftware systems. Also, the growing complexities ofsoftware business have led to huge costs in distributingthe software to end-users, providing them upgrades, andmanaging service support.

The current research work opens up some importantresearch issues and implications for theory-building.Building on Wierenga, van Bruggen and Staelin (1999),our research work asserts that there is a need for morestudies in real-life company environments in this area.This would raise even more interesting issues when theimplementation is in the context of globally distributedwork environment. One of the key issues in this contextwould be the documentation of improvements in animplementation setting. As pointed out by Wierenga, vanBruggen and Staelin (1999), �what are the best ways todocument improvements if there are no logical controls,e.g., the firm either fully implements the proposedstrategy or does not? How can the researcher best predictwhat would have happened if the decision aid was notused?� Another interesting research issue from atheoretical standpoint would be the development of ataxonomy of various application settings, which wouldprovide a prescription of which setting is more amenableto MMSS implementation. It would be interesting todevelop a continuum of problem settings from relativelystructured to more complex problems (refer Van Bruggenand Wierenga (2001) for a representative work in thisarea).

As far as the learning from the current application isconcerned, it is evident that the motion picture industryis an unusual application area for implementation ofMMSS. The organization culture does not favormathematical models, and the cognitive models of thedecision makers are heuristics rather than analytical. Insuch an environment, a lot of effort is required on thepart of the model developers to gain the trust of themanagement. As is clear from the series ofimplementations discussed in this paper, trust was builtin an evolutionary way. Another crucial aspect of thedevelopment process is to maintain managerial

involvement in the model development process andaccommodating user needs. Finally, we learned that theMMSS should play only a decision support role, as itshould. There are simply too many judgmental elementsin most practical settings to leave 100% of the decisionsto the model. Thus, the model and the intuition of themanager together should constitute a most powerfulcombination.

Our experience shows that it is possible to implementMMSS in newer application domains. The present workcould readily be extended to other parts of theentertainment industry, such as facilities for plays andmusic performances, and the scheduling of sportingevents. While there is less information available aboutthe demand for entertainment products than for frequentlypurchased consumer goods, substantial progress can bemade in applying DSS in such fields. The theoreticalconcepts developed in the current research efforts couldbe applied to other shelf-space management applicationdomains, such as super-market retailing, e-retailing andbanner advertising. Some preliminary work in thisdirection is available in Raut, Swami, and Moholkar(2006), and Hansen, Raut and Swami (2006).

REFERENCES

1. Eliashberg, J., Elberse, A. and Leenders M.A.A.M., 2006,The Motion Picture Industry: Critical Issues In Practice,Current Research & New Research Directions, MarketingScience, Vol. 25(6), 638-661.

2. Eliashberg, J., Swami, S., Wierenga, B. and Weinberg, C.B.,2001, Implementing and Evaluating SilverScreener: AMarketing Management Support System for MovieExhibitors,� Interfaces, Vol. 31 (3), May-June, S108�S127.

3. Eliashberg, J., Swami, S., Wierenga, B. and Weinberg, C.B.,2006a, Decision Support in the Movies: Scheduling theMultiplexes, Working paper: Rotterdam School ofManagement.

4. Eliashberg, J., Hegie, Q., Ho, J., Huisman, D., Miller, S.,Swami, S., Wierenga, B. and Weinberg, C.B., 2006b,Micro-Scheduling of Movies at Theaters, Working paper:Rotterdam School of Management.

5. Eliashberg, J., Jonker, J., Sawhney, M.S., and Wierenga,B., 2000, MOVIEMOD: An Implementable DecisionSupport System for Pre-Release Market Evaluation ofMotion Pictures, Marketing Science, Vol. 19 (3), 226-243.

6. Hansen, J., Raut, S. and Swami, S., 2006, Heuristic andMeta-heuristic Approaches for Retail Shelf AllocationProblem, In: Proceedings of AMS-ACRA Conference,Florida, USA.



7. Jardin, X., 2005, The Cuban Revolution, www.wired.com/wired/archive/13.04/cuban.html, Issue 13.04 [April 2005]

8. Krider, R. E. and Weinberg, C.B., 1998, CompetitiveDynamics and the Introduction of New Products: TheMotion Picture Timing Game, Journal of MarketingResearch, Vol. 35(1), 1-15.

9. Mockus, A. and Herbsleb, J.D., 2001, Challenges of GlobalSoftware Development, IEEE Metrics, pp. 182-184.

10. Raut, S., Swami, S. and Moholkar, M., 2007, Heuristic andMeta-heuristic Approaches for Multi-Period Shelf-spaceOptimization Model: The Case of Motion Pictures,Working Paper: Department of Industrial & ManagementEngineering, IIT, Kanpur, India.

11. Schultz, R.L. and Slevin, D.P., 1975, Implementation andOrganizational Validity: An Empirical Investigation, In:R.L. Schultz and D.P. Slevin (eds), ImplementingOperations Research/Management Science. New York:American Elsevier Publishing Company Inc.

12. �Star Wars� Is Mother of Technology, www.sci-tech-today.com/story.xhtml?story_id=35013 [May 16, 2005]

13. Swami, S., Eliashberg, J. and Weinberg, C.B., 1999,SilverScreener: A Modeling Approach to Movie ScreensManagement, Marketing Science, Vol. 18 (3), 352-372.

14. Tiwana, A., 2000, The Knowledge Management Toolkit,Pearson Education, New Delhi.

15. Van Bruggen, G.H. and Wierenga, B., 2001, MatchingManagement Support Systems and Managerial Problem-Solving Modes: The Key to Effective Decision Support, European Management Journal, Vol. 19(3), June, 228-238.

16. Wierenga, B., and Van Bruggen, G.H., 2000, MarketingManagement Support Systems, Kluwer AcademicPublishers, Norwell, MA.

17. Wierenga, B., Van Bruggen, G.H., and Staelin, R., 1999,The Success of Marketing Management Support Systems,Marketing Science, Vol. 18 (3), 196-207.

globally distributed r&d work in a marketing management ... · globally distributed r&d...

Documents