To Main Proceedings Document
The system dynamics approach and the methodology for
multicriteria decision aid as tools for organizational learning
Rudolf ROSAS FLUNGER
Booz: Allen and Hamilton
Praia de Botafogo 300, 5to Andar
22259-900 Rio de Janeiro, RJ, Brazil
Phone ++ 55 (21) 237 8409 Fax ++ 55 (21) 553 0103
email: flunger rudolf@bah.com
Abstract
The general objective of the present workis to demonstrate that the joint utilization of the system dynamics approach
and the methodology of multicriteria decision aid enable the organizational learning process improving therefore the
decision processes in the organizations.
A scientific approach was developed based on real experiences grouped according two categories: laboratory
experiences and real world organizations. The laboratory experiences are simulations done with university students.
Tt was created a controlled experimental environment to use the ‘‘Beer Distribution Game’. The original rules to
evaluate the learning process were modified to prove our hypothesis. The real world organizations experiences are
based on a pharmaceutical company that is passing a transformation process.
From these experiences we concluded that the joint utilization of the proposed tools present specifics advantages as:
The system dynamics approach helps the formalization and communication of knowledge due to the perception of
causal links, exchanges and the simulation tools
The system dynamics simulations make evident the counterintuitive feedback impact in complex systems.
The methodology of multicriteria decision aid allows to make evident and to address scientifically the desires and
beliefs of organizational agents (stakeholders) in order to formalize them into the organizational strategy and
objectives.
Key words. organizational leaming, system dynamics approach, methodology of multicriteria decision aid, Beer
Game, pharmaceutical companies, Venezuela.
Introduction
The present work is a brief summary of a doctoral dissertation presented at the LAMSA DE, Université PARIS IX -
Dauphine (ROSAS FLUNGER 1999). This research is the result of a dedicated effort looking to improve the
decision processes in organizations. A long several years of professional and acaclemic practice, the system
approach (SDS) (WOLSTENHOLME 1990), and the multicriteria decision analysis (MCDA) (POMEROL &
BARBA ROMERO 1998, ROY 1985) emerged as especially effective tools to improve the decision processes in
organizations. The combined use of the two techniques uses the problem understanding advantages brought by the
system dynamics approach (SDS) and the objectiveness to support the decision process of the multicriteria decision
analysis (MCDA).
The paper is structured in 4 pieces: The presentation of the research problem, the conceptual framework, the
experiences and the research conclusions. The first part set the soope, objectives and drivers of the research. The
second sets the theoretical basis. A fterward, the experiences context are described, the decision framework explained,
an analysis identifying and evaluating the evolution of the articulated knowledge is described and finally the
télationship between the knowledge in use and the resulting performance of the organization under study is analyzed.
At the end, a set of conclusion related to the types of knowledge identified and the impact of the suggested techniques
aremade.
1. Organizational Leaming problem definition
Organizational leaming has been studied from different perspective and through all related fields: sociology, cognitive
sciences, organizational behavior, and others (HUBER 1996, MIDLER 1992). Specific concepts have been defined
like, knowledge, organizational memory, deuteroleaming, and organizational cognition. However, major issues still
requiring futher research (ARGY RIS & SCHON 1996 pp. 200-2001):
= Atwhatlevel of aggregations does it make sense to talk about productive organizational leaming?
= What does it means productive leaming?
= Howto manage the inherent bamiess for organizational leaming in real world organizations?
= Whet kind of interventions can enable the organizational leaming process?
From a decision aid sciences perspective! the organizations can be conceived as entities who's behavior is driven by
the decision processes of its stakeholders (MARCH 1994). Adding the decision perspective to the organizational
Jeaming problem brings three major research directions:
= The organizational action as a result of a decision process based in the articulation of available knowledge
= Knowledge acquisition process inside the organization and types of knowledge articulated in the
organizational decision process
= Organizational memory, problem representation and language
‘These issues should be address in two dimensions:
= Aggregation level at which knowledge is articulated to solve an organizational problem.
= Impact of knowledge on the different decision attributes
In order to build real world interventions that enable organizational leaming, a set of experimental tools can be
assembled from the system dynamics approach (SDS) and the multicriteria decision analysis (MCDA). From the
SDS it canbe taken the simulation and representation tools to understand the decision context (WOLSTENHOLME
1990). From the MCDA, the ctiteria building and aggregation procedures can be taken to enable the final steps of the
decision process” (POMEROL & BARBA ROMERO 1993, ROY 1985).
From this context, we proposed to focus in three major issues:
= What dements of the organizational decision process can be address with the experimental tools?
= What types of knowledge can be better addressed with the use of the proposed experimental tools?
= How the experimental tools can be used as memorization instruments for the knowledge issue from the
organizational leaming process?
2. Conceptual framework
The research was based on three major conceptual definitions:
= Decision process in the organizations
= Typesof knowledge articulated by decision makers in the organizations
= Organizational Leaming concept
The decision process is build upon two dimensions: a problem solving axis and a consensus-building axis (see Figure
1: Axis of the decision process in organizations):
Satistycing
Problem
Solving
Not
Satisfycing
Not Sufficient T Sufficient
Consensus Building
Figure 1: Axis of the decision process in organizations
One axis corresponds to the problem solving heuristic (POMEROL 1997). The problem solving process is focus on
finding a satisfycing solution to the objective decision problem (ROY 1992). However, if the selected solution is not
accepted by the organization it will not be implemented at an organizational leva. Therefore, a second axis exists that
chives the problem solving heuristic to a consensus building process looking to build ownership on the solution.
According to the position of the outcome in the decision mairix, the decision process could be satisfycing but not
organizational, recognized by the organization but not effective or effective and organizational enough to be
implementable as an organizational action (ROSAS FLUNGER 1999 pp. 106-114).
The second major conceptual definition is the classification of the knowledge articulated in a Management Situation’.
Thee types of knowledge can be identified (HATCHUEL & WEIL 1992):
= Know-How (Savoir faire): Is the knowledge that expresses a transformation procedure through known,
actions. Can be represented through a collection of procedural instructions that can be easily codified in a
computer language
= Understanding (Savoircomprendre); Is the knowledge set articulated to understand why a situation
perceived doesn’t comespond with a desire situation. I typical example can be the maintenance engineering
context when a dysfunctional situation is perceived and its causes identified.
= Combination knowledge (Savoircombiner): Is the set of knowledge articulated to combine resources and
objectives in a project approach. Is knowledge of the entrepreneur that is able to organize its resources and
objectives to achieve a business system.
Finally, we defined organizational leaming as the durable modification of the elements of the decision process of
stakeholders in a management situation oriented to the implementation of organizational actions with a problem-
solving objective. The concept is based in a concrete set of assumptions (ROSAS FLUNGER 1999 PP. 138-141):
Leamingis aprocess
‘Leaming is based on a modification the objectives, representations and rational of stakeholders
Modifications should be saved in the “ organizational memory” in order to become durable
Modifications happens at individual level of stakeholders
Stakeholders are evaluated by the outcome of their actions (management situation)
‘Leaming is expressed through an organizational action
The process started is perceived problem
The leaming process has the intention to solve aproblen
Experiences
3.L Laboratory experiences
3.11 Experience desaiption
The Laboratory experiences are a modified version of the “ Beer Distribution Game” (STERMAN 1994,
1982). As part of operations research courses in two engineering schools in Venezuela a supply chain
system with 5 echelons was simulated with more than 100 students (4 experiences with more than 25
students each). The students were grouped in teams acting as the retailers, distributors and wholesalers. The
moderators (professors) served as the clients and the plant of the system. Each team was evaluating
comparetively against the group. Grades based on team position and chain position against other chains was
measured. In consequence, the students were stimulated to compete between them but at the same time they
had the responsibility of improving the overall chain performance. Major differences with the original Beer
Gameare:
= Thedemand cuveisnotastep overa constant order volume but a cyclical demand
= The students were allowed to talk. At the beginning of the experiences the slences paradigm
prevailed as the experiences were made in a formal evaluation context. Once the students realized it
‘was possible to talk, the important issue was the effectiveness of the communication and the team
dynamics allowing an efficient resolution of the game challenge
3.12. Decision framework
The Beer Game demands from the student's three main decision processes:
313,
Organization: The students have to organize themselves in the teams to be able to meet the time
pressure of the experience. The teams have very short time to place the orders and make the
deliveries. In consequence, an intemal organization is imperative to be efficient performing the
game activities. Two level of organization are required in the game: team and game organization.
First the teams have to allocate intemally the task of the game. Secondly, once the teams have
discovered they can communicate through the chain they have to define a communication model
to solve the information delay issue. For the intemal team organization three main options are
available: functional, mix and no organization. At the chain level also three models are available:
non communication, representative and integral models depending on the level and stakeholders
involved in the communication processes
Stock policy: The teams have to define a stock policy that drives the order process. The students
were allowed to use all their resources (books, etc) to define the policy. They were formally asked
inthe experience protocol to make explicit their stock policy.
Order placement: At each game step the teams have to define the quantities in the orders and the
quantities in the deliveries,
Knowledge assessment
Each identified decision articulated a different type of knowledge.
Organization: Understand the task and objectives of the game as well as resources and constraint
are the major issues. The task can only be achieved articulating the combination knowledge
(savoi-combiner). All the teams were not able to define or to implement a clear organizational
model.
Stock policy: To understand the feedback processes in information and physical flows of the game
is the major issue for this decision. Understanding type of knowledge (savorcomprencte) is
articulated for this decision. Build the causal loops and identify the critical parameters are the
inherent task and knowledge required. A clear evolution of the knowledge involved was perceived
through the different moments of the game. At the beginning a first policy was set up without
considering any communication with the chain. Once it was realized communication was allowed,
a second stock policy was defined. Finally, once stabilized the orders the teams moved from a
cooperative environment to a more competitive one to try to get better grades than the others. (See
Figure 2 : Evolution of deviation average of ordas from réailas, wholesalers and distibutors with polynomial
approximation (95%)
Order placement: Defining the quantities be ordered and delivered is a procedural task based on
knowing how to apply the defined stock policy. Correspond to a typical know how (savoir-faire)
context.
Figure 2 : Evolution of deviation average of orders from retailers, wholesalers and distributors with
polynomial approximation (95% )
3.2. Real world organization
32.1 Experience desaiption
A family owned pharmaceutical industy in Venezuela was defining its strategic intent After several
decades of success as a niche player, good levels of profitability and financial stability the next generation of
owners was questioning the sustainability of its business system. An organizational Leaming initiative was
designed and implemented targeting to:
= Createacommon vision in arecently hired new management team
= Develop an adapted business system sustainable in the new market conditions of V enezuda
The intervention was designed as a set of workshop were high leva causal models were defined and
MCDA tools were deployed to support the required decisions.
3.22. Decision framework
Duning the initiatives 4 decision were identified as been the most ctitical:
= Product dassification: A management decision on product portfolio was required in the laboratory.
A significant number of product, especially the two most important were classified and sold as
pharmaceutical products when it was possible through simple initiatives to transform them in over
the counter (OTC) products. The change would also impact all commercialization chain moving it
to the common mass market. It was not a simple classification exercise but a management decision
to betaken.
= Performance assessment: A full set of performance indicators was defined to be able to evaluate
the performance of the laboratory and an assessment process was launched requiring a
measurement effort in all areas of the company
= Mission and vision definition: In workshops with the top management of the firm and the owners
the new vision and mission of the company was defined.
= Objectives definition: The mission and vision were articulated in detailed performance targets with
Clear measures
3.23. Knowledge assessment
The identified decisions were based on specific knowledge:
= Product dassification: A ticulated the know-how of rules and laws that defines the current product
portfolio according to industry standards. The MCDA tools were used to support the criteria
definition and assessment to identify current portfolio posttioning. The type of knowledge use was
know-how
= Performance assessment: Articulated a set of knowledge targeting the understanding of the current
performance of the company. The type of knowledge as understanding (savoir. comprendre). The
main issue is to understand why the current performance is at the perceived levels
= Mission and vision definition: using all knowledge developed about the firm, a clear understanding
of available resources (coming form the performance evaluation) the new company goals were
defined. This was the result of combining available competences with feasible objectives. The type
of knowledge atticulated was combination (savoir. combiner)
= Obgctives definition: Based on the vision and mission the detailed resources and detailed
objectives should be defined and articulated to achieve the desired goals. The type of knowledge in
useare combination knowledge (savoir combiner)
4, Condusions
The beer game allowed to understand the proposed organizational leaming concepts but not the effective use of the
experimental tools. In the experience, it was clearly enough what types of knowledge were articulated on each
decision. However, the time pressure and dynamics of the beer game does not allow to use simulation models neither
MCDA procedures.
The conclusions driven from the experience were:
Organizational models with clear allocation of responsibilities increased the performance of the teams
Open but organized communications were ciitical to achieve high performance levels
The use of explicit stock policies increased the performance levels
The performance level were proportionally related to the leaming velocity of the teams for the understanding
knowledge related to the definition of the stock policy and also to the velocity on developing the know-how
(order placement)
The pharmaceutical company experience allowed to use the experimental tools in a real world environment. The
conclusions from the experience were:
= The clear allocation of responshilities as well as the understanding of the allocated task increase the
organizations performance
= The openness generated by the joint workshops, were the top management interacted, making the identified
decisions through the support of the experimental tools increase the organizations performance
= The MCDA allowed to improve the transparency and faimess of the performance evaluation process
improving the consensus building axis of the decision process
= The utilization of simulation models and causal diagrams improved the identification of ctitical issues
improving the development of understanding knowledge (savoircomprendre) and increasing the
effectiveness of the problem solving axis of the decision process
Further development must be made as the sample of the research is restraint to a single example (pharmaceutical
laboratory) but gives important insights in the implementation of the 2 suggested techniques.
References
ARGY RIS, Chris, SCHON, Donald, A., (1996), Organizational leaming II: Theory, method, and practice,
Addison-Wesley Series on Organizational Development, 1996.
GIRIN, Jaques, (1990), "Analyse empirique des situations de gestion : déments de théorie et de méthode", in
Epistémologies et science de gestion, Economica coll. Gestion, Paris, pp. 141-182.
HATCHUEL, Amand, WEIL, Bemard, (1992), L'Expert etle systme, Economica, Paris, 1992.
HUBER, George P., (1996), "Organizational leaming: The contributing processes and the literatures", in COHEN,
Michael D., SPROULL, Lee S. (Eds.), Organizational learning, Organization Science Series, Sage Publications,
1996, pp. 124-162.
MARCH, James G., (1994), A primer on decision making: How decisions happen, The Free Press, New Y ork,
1994.
MIDLER, Christophe, (1992), “Evolution des régles de gestion et processus d’ apprentissage”, in L’ Apprentissage
Organisationnel, Cahier N29, Centre de Recherche en Gestion (CRG), Ecole Polytechnique, Paris, décembre 1992.
POMEROL, Jean-Charles, (1997), "Artificial intelligence and human decision making", European Jounal of
Operation Research, 99 (1997), pp. 3-25
POMEROL, Jean-Charles et BARBA-ROMERO, Sergio, (1993), Choix multicritare dans I’ entreprise, Hermes,
Paris, 1993.
ROSAS FLUNGER, Rudolf. (1999). L approche de la dynamique des systames et! ‘aide a la décision multicritare
comme outils d ‘apprentissage organisationnel, Thése de doctorat, Université Paris IX -Dauphine, Paris.
ROY, Bemand, (1992), Science de la décision ou science d'aide a la décision, Cahier du LAMSADE NE? 97,
LAMSADE, Université Paris-IX Dauphine, Patis, février 1992.
SIMON, Herbert A., (1997), Administrative behavior: A study of decision-making processes in administrative
organizations, Fourth Edition, The Free Press, New Y ork, 1997.
STERMAN, John D., (1992), The Beer Distribution Game: An annotated bibliography covering its history and use
ineducation and research, System Dynamics Group, Sloan School of Management, MT, Cambridge, July 1992.
STERMAN, John D., (1984), Instruction for rumning the Beer Distribution Game, System Dynamics Group, Sloan
School of Management, M.LT, Cambridge, October 1984.
WOLSTENHOLME, Etic F,, (1990), System enquiry: A system dynamics approach, John Wiley & Sons Ltd,
Chichester, UK, 1990.
' For a further discussion on the relevance of the concept decision sciences and decision aid sciences ROY 1992
* For further details on the steps of the decision process SIMON 1997.
° For a definition of management situation GIRIN 1990
