• Like
  • Save
New Product Development as a Complex Adaptive System of Decisions
Upcoming SlideShare
Loading in...5
×
 

New Product Development as a Complex Adaptive System of Decisions

on

  • 2,261 views

Early research on new product development (NPD) has produced descriptive frameworks and models that view the process as a linear system with sequential and discrete stages. More recently, recursive ...

Early research on new product development (NPD) has produced descriptive frameworks and models that view the process as a linear system with sequential and discrete stages. More recently, recursive and chaotic frameworks of NPD have been developed, both of which acknowledge that NPD progresses through a series of stages, but with overlaps, feedback loops, and resulting behaviors that resist reductionism
and linear analysis. This article extends the linear, recursive, and chaotic frameworks by viewing NPD as a complex adaptive system (CAS) governed by three levels of decision making — in-stage, review, and strategic—and the accompanying decision rules. The research develops and presents propositions that predict how the configuration and organization of NPD decision-making agents will influence
the potential for three mutually dependent CAS phenomena: nonlinearity, selforganization, and emergence. Together these phenomena underpin the potential for NPD process adaptability and congruence. To support and to verify the propositions, this study uses comparative case studies, which show that NPD process adaptability occurs and that it is dependent on the number and variety of agents, their corresponding connections and interactions, and the ordering or disordering effect of the decision levels and rules. Thus, the CAS framework developed within this article maintains a fit among descriptive stance, system behavior, and innovation type, as it considers individual NPD processes to be capable of switching or toggling between different behaviors — linear to chaotic — to produce corresponding innovation outputs that range from incremental to radical in accord with market expectations.

Statistics

Views

Total Views
2,261
Views on SlideShare
2,259
Embed Views
2

Actions

Likes
1
Downloads
55
Comments
0

2 Embeds 2

http://www.slashdocs.com 1
http://searchutil01 1

Accessibility

Categories

Upload Details

Uploaded via as Adobe PDF

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Processing…
Post Comment
Edit your comment

    New Product Development as a Complex Adaptive System of Decisions New Product Development as a Complex Adaptive System of Decisions Document Transcript

    • J PROD INNOV MANAG 2006;23:437–456r 2006 Product Development & Management AssociationNew Product Development as a Complex Adaptive System ofDecisionsIan P. McCarthy, Christos Tsinopoulos, Peter Allen, and Christen Rose-Anderssen Early research on new product development (NPD) has produced descriptive frameworks and models that view the process as a linear system with sequential and discrete stages. More recently, recursive and chaotic frameworks of NPD have been developed, both of which acknowledge that NPD progresses through a series of stages, but with overlaps, feedback loops, and resulting behaviors that resist reduc- tionism and linear analysis. This article extends the linear, recursive, and chaotic frameworks by viewing NPD as a complex adaptive system (CAS) governed by three levels of decision making—in-stage, review, and strategic—and the accom- panying decision rules. The research develops and presents propositions that predict how the configuration and organization of NPD decision-making agents will influ- ence the potential for three mutually dependent CAS phenomena: nonlinearity, self- organization, and emergence. Together these phenomena underpin the potential for NPD process adaptability and congruence. To support and to verify the proposi- tions, this study uses comparative case studies, which show that NPD process adaptability occurs and that it is dependent on the number and variety of agents, their corresponding connections and interactions, and the ordering or disordering effect of the decision levels and rules. Thus, the CAS framework developed within this article maintains a fit among descriptive stance, system behavior, and innova- tion type, as it considers individual NPD processes to be capable of switching or toggling between different behaviors—linear to chaotic—to produce corresponding innovation outputs that range from incremental to radical in accord with market expectations.Introduction sectors. Given the importance and value of NPD to firm performance, researchers have developed de-I t is widely recognized that effective new product scriptive frameworks based on linear, recursive, and development (NPD) processes are causally chaotic system perspectives, which provide different important in generating long-term firm suc- insights and descriptive theories about NPD processcess (Cooper, 1993; Ulrich and Eppinger, 1995; structure and behavior. These are then often the basisWheelwright and Clark, 1995). They can lead to a for normative research, which seeks to predict andcore competence that either differentiates a firm from prescribe causality in NPD processes (Griffin, 1997b;its competitors (Prahalad and Hamel, 1990) or pro- Page, 1993).vides a threshold competency that is necessary just to Linear frameworks help explain how the organiza-survive in fast-changing and innovative industry tion and management of NPD processes relate to NPD performance, specifically to lead times and due Address correspondence to: Christos Tsinopoulos, University of dates. Yet this overriding focus on process structure,Durham, Durham Business School, Durham, Mill Hill Lane, DH13LB, UK. reliability, and control has tended to ignore the
    • 438 J PROD INNOV MANAG I. P. MCCARTHY ET AL. 2006;23:437–456factors that govern the ability to innovate (Badaracco, and presents a complex adaptive system (CAS) frame-1991; Moenaert et al., 2000). This myopia occurs be- work of NPD. It combines the system dimensionscause linear frameworks represent the NPD process as presented by Schoderbek, Schoderbek, and Kefalasan ordered, sequential, and relatively predictable sys- (1985) with differences in NPD decision-making levelstem of activities (Bonner, Ruekert, and Walker, 2002). and rules to interpret how the CAS phenomena ofThis leads to a mechanistic interpretation and focus nonlinearity, self-organization, and emergence occuron process efficiency, which is inclined to ignore how in NPD processes. These constructs provide a basisprocess factors such as flexibility, informality, feed- for understanding the balance between the processback, and autonomy might influence innovation order and control emphasized in the linear view and(Clark and Fujimoto, 1991; Dougherty, 1992; the process instability and creativity emphasized inGriffin, 1997a). Consequently, researchers have the recursive and chaotic views. Thus, the CAS frame-responded by developing recursive and chaotic-based work asserts that NPD process behaviors and config-frameworks to understand better how these factors urations are not necessarily fixed. An individualand resulting process behaviors are associated with process has the ability to adapt and to produce adifferent types of innovation. range of process behaviors—linear to chaotic—with To complement and to build on the linear, recur- corresponding types of innovation output—incremen-sive, and chaotic frameworks, this article develops tal to radical. The foundation of the CAS framework is the rec- BIOGRAPHICAL SKETCHES ognition that NPD processes are systems whose ele- Dr. Ian McCarthy is Canada Research Chair in Management of ments, known as agents, are partially connected and Technology in the Faculty of Business Administration at Simon have the capacity for autonomous decision making Fraser University. His research interests focus on the diversity and and social action, known as agency. The decision design of industrial organizations, encompassing the areas of oper- ations management, organization science, technology management, rules, interactions, and outcomes of agents create and evolutionary theory. He is a chartered engineer, a college mem- three mutually dependent phenomena that define ber of the UK Engineering Physical Sciences Research Council and characterize CASs: nonlinearity, self-organi- College, and a director of the Complexity Society. Previously he was a tenured faculty member at the University of Warwick and the zation, and emergence. Thus, studies of CASs are University of Sheffield and held management positions at Philips not concerned with the relative complicatedness or Electronics, British Alcan, and Footprint Tools. complexity of systems and their elements (Anderson, Dr. Christos Tsinopoulos is lecturer of operations and strategy in 1999; McCarthy, 2004; Morel and Ramanujam, the Durham Business School at Durham University. His research 1999). They are interested in how certain systems interests focus on understanding innovation in organizations using are able to learn and to create new rules, structures, systems methods that stem from evolutionary and complex systems thinking. Previously he was a research fellow in new product de- and behaviors at several interrelated levels. These velopment at the University of Warwick. He holds a Ph.D. from the characteristics are necessary and important for sys- University of Warwick and an M.Sc. from the University of Shef- tems concerned with exploration and innovation, but field. He has held management positions in the copper and alumi- num industries in the quality and maintenance departments. they are also difficult to represent and to analyze using static, linear models. Dr. Peter Allen is professor and head of the Complex Systems Management Centre in the School of Management at Cranfield The decision aspect of the framework acknowledg- University. He is also coordinator of NEXSUS, the ESRC Priority es that the principal role of NPD agents is to make Network in Complex and Dynamic Processes. His research is di- judgments and choices that bridge the gap between an rected toward the application of the new ideas concerning evolu- tionary complex systems to real-world problems. He has a Ph.D. in idea and reality. This process agency is also a central theoretical physics, was a Royal Society European Research Fellow component of CAS theory (Choi, Dooley, and Rung- from 1970 to 1971, and was a senior research fellow at the Univer- tusanatham, 2001; Eisenhardt and Bhatia, 2002; ´ site Libre de Bruxelles from 1972 to 1987, where he worked on the McCarthy, 2004) that helps to examine the black- theory of complex systems with Nobel laureate Ilya Prigogine. box activities of NPD (Krishnan and Ulrich, 2001) Christen Rose-Anderssen worked as a naval architect and manager regardless of functional perspective (Whetten, 1989). in the ship-building industry in Norway, Northern Europe and Asia for many years. He has worked with strategy, marketing, human Thus, the present study assumes that NPD project resource management, and organizational change in offshore engi- teams deal with uncertainty through information neering. His main interests are in organizational change through processing (Galbraith, 1977) and that decision levels collective work and the management of knowledge and learning for innovation through effective communication development across and accompanying decision rules affect the mode and cultural boundaries in complex activity networks. viability of NPD operation, contingent upon a fit be- tween the NPD process and its environment.
    • NEW PRODUCT DEVELOPMENT AS A COMPLEX ADAPTIVE SYSTEM OF DECISIONS J PROD INNOV MANAG 439 2006;23:437–456 With this introduction, the contribution and the on the cooperation, coordination, and communica-structure of this article are as follows. First, it pro- tion of those involved in the NPD process (Coopervides a review of the linear, recursive, and chaotic and Kleinschmidt, 1986). By focusing on the structureNPD frameworks to understand their relative inter- of process and the interstage connections, linearpretations, benefits, and limitations. It then introduc- frameworks attempt to explain how process behaviores and defines a CAS and proposes three NPD affects product quality, execution of key tasks, prod-decision levels as constructs for the CAS framework. uct development costs, product reliability, productThe aim of these two review sections is to position the variety, and managerial complexity (Muffatto andcontribution of this article within the context of NPD Roveda, 2000; Shepherd and Ahmed, 2000).literature and to show how a CAS framework pro- Probably the best-known framework based on thevides a complementary perspective that both com- linear view is the stage-gate method (Cooper, 1990).bines and extends the linear, recursive, and chaotic This framework represents NPD as a sequential anddescriptive interpretations. A CAS framework views ordered process of four to five stages of activities andindividual NPD processes as adaptable or malleable decisions. These typically include concept develop-systems capable of producing a range of behaviors to ment, product design, testing and validation, andsuit different innovation expectations, levels of mar- product launch and ramp-up. The inputs into theket uncertainty, and rates of change. This interpreta- system are new ideas and market needs, the systemtion forms the basis for four propositions that predict elements are resources such as engineers and market-a fit among NPD process adaptability, the market ers, and the outputs are new products (Clark andexpectations imposed on the process, and the effect Wheelwright, 1993). Between the stages are gates ordecision levels have on the order and disorder in the checkpoints, where the progress and outputs of theNPD process. To support and provisionally test these previous NPD process stages are evaluated. This log-propositions the second half of this article uses three ical and systematic representation of NPD helps tocase studies to describe and to characterize causes of organize and to comprehend the complexity of theCAS phenomena in relatively regulated and formally process (Clift and Vandenbosch, 1999) and to con-managed NPD processes. The case-study observa- ceptualize the potential for parallel tasks and activitiestions reveal how the configuration of NPD decision (Winner et al., 1988).makers and corresponding decision levels and deci- Although linear frameworks can reveal how inap-sion rules might affect process adaptability and the propriate structure and poor control can result inpotential to produce innovative outputs congruent planning and coordination problems, they tend to ig-with the needs of its environment. Finally, the article nore the behaviors and system features that governconcludes with a discussion of the implications that a the innovative capacity of an NPD process (Bonner,CAS framework of NPD might have for managerial Ruekert, and Walker, 2002; Clift and Vandenbosch,practice and suggests a future research agenda. 1999; Johne and Snelson, 1988; Olin and Wickenberg, 2001; Quinn, 1985). This is in part because linear frameworks focus on and represent the structural log- ic of an NPD process, which then facilitates and pro-Literature Review motes understanding of the factors that governNPD Frameworks process lead time and consistency. The result is a descriptive compromise between NPD control andTo date, frameworks for interpreting and describing autonomy, each respectively influencing processNPD reality are based primarily on three schools of reliability and process creativity.thought that reflect three different systems views: lin- During this period, researchers also challenged theear, recursive, and chaotic. assumption that NPD activities were exclusively an Linear NPD frameworks stem from traditional and ordered and sequential system of discrete stageslogical project management methods that seek to de- (Kline and Rosenberg, 1986; Leonard-Barton, 1988;liver appropriate outputs on time and within cost. Schroeder et al., 1989). A seminal article by RothwellThey interpret the process of innovation as a series of (1992) argued that the study of NPD as an automatic,events and activities, which are sequential and discrete dependable, and routine decision-making process didin nature (Zaltman, Duncan, and Holbek, 1973). not explain how radical innovations emerge (DewarAlso, process control and efficiency are dependent and Dutton, 1986; Leifer et al., 2000; McDermott and
    • 440 J PROD INNOV MANAG I. P. MCCARTHY ET AL. 2006;23:437–456O’Connor, 2002; Utterback, 1996). As radical inno- From a descriptive stance, the recursive perspectivevations or really new products can significantly alter assumes that the connections and boundaries betweenand redefine markets (Cooper and Kleinschmidt, NPD events are less clear and rigid (Cheng and Van1993; Schmidt and Calantone, 1998; Song and Mon- de Ven, 1996), because innovation is a process that istoya-Weiss, 1998), it is important to distinguish be- always dynamic, mostly nonlinear, and often untidytween NPD process characteristics that typically and muddled. For example, Repenning’s (2001) studygenerate incremental product innovations and those of firefighting—solving unplanned problems andcapable of producing breakthrough products. To challenges—in formal and controlled NPD projectsavoid contradictory or paradoxical representations, concluded that firefighting is a self-reinforcingsuch differentiation requires the development and use phenomenon more likely to occur in large complexof frameworks that are capable of interpreting the var- projects. As the level of concurrent and cross-iations in process activities and structure, in accord with functional activity in NPD projects has grown, a cor-the mix of environmental and organizational factors responding increase has occurred in the dynamicthat govern NPD behavior (Van de Ven et al., 1999; complexity of NPD projects (Smith and Eppinger,Crawford and Di Benedetto, 2000). Cooper, Edgett, 1997; Wetherbe, 1995). With these developments,and Kleinschmidt (1999, 2002), who were the main Ford and Sterman (1998) argued that the frameworksexponents of the stage-gate method, recognized that for understanding NPD have simply not been keepingdescriptive linear frameworks encourage linear pace with this increased complexity. This results in in-management practices, which tend to produce only in- adequate descriptions and limited explanations (Paichcremental innovations. To emphasize the significance of and Sterman, 1993; Rechtin, 1991; Sterman, 1994).achieving a balanced product portfolio—that is, an as- The third NPD framework, the chaotic view, is ansortment of both incremental and radical innovations— extension of the recursive framework (Cheng and VanCooper and colleagues also advocated the use of stra- de Ven, 1996; Koput, 1997). It assumes that chaotictegic buckets to define how management should allocate behavior can occur in NPD processes and depictsNPD resources in terms of innovation, market and highly innovative NPD processes as systems with ran-product type. With these developments, a consensus dom-like and nonlinear behavior that generate irreg-had formed that innovation is a temporal process made ular or disordered actions. Such NPD processes areof identifiable events but that the linear perspective fails relatively unpredictable, as small changes in one partto fully represent the connective, dynamic, turbulent, of the process can rapidly develop and take the systemand fuzzy aspects of the radical NPD process. Thus, along new trajectories, referred to as sensitivity to in-recursive and chaotic NPD frameworks were developed itial conditions. To study innovation from this per-to advance interpretation and understanding of the spective Cheng and Van de Ven (1996) used a chaos-types of activities that underlie the development of rad- theory algorithm from physics to examine the effectsical innovations. of feedback loops in NPD. The result was a model Kline and Rosenberg (1986) offered one of the first suggesting that the process of innovation starts cha-alternatives to the linear framework. They presented a otically and finishes in stability and that the latterchain-linked model with feedback loops to describe stages of product innovation are better suited to linearthe relationships and iterations among research, in- frameworks. The work by Koput (1997) also used avention, innovation, and production. Leonard-Barton chaotic framework but was concerned with the dy-(1988) made similar observations in a study that de- namics of searching for innovation. The main conclu-scribed NPD as a series of small and large recursive sions were that the innovation activities of search,cycles that represent project setbacks and restarts. screening, and implementation are inextricably linkedThese recursive frameworks of NPD seek to represent to each other and that understanding how process‘‘events in which activity is multiple, concurrent, and feedback loops influence the properties of thesedivergent, in which the process includes feedback activities is an important and significant researchand feed-forward loops’’ (Adams, 2003, p. 232). challenge.Thus, recursive frameworks challenge the idea of Together the three NPD frameworks provide aorderly sequences and assert, particularly for radical complementary hierarchy or ladder of abstractioninnovations, that NPD stages overlap, creating for interpreting and describing different typesfuzziness and disorder in the process (Adams, 2003; of NPD reality (see Table 1). Linear frameworksConstant, 2000; Schroeder et al., 1989; West, 1990). provide a simple and logical overview of the process
    • NEW PRODUCT DEVELOPMENT AS A COMPLEX ADAPTIVE SYSTEM OF DECISIONS J PROD INNOV MANAG 441 2006;23:437–456 structure and flows. They are suitable for NPD (i.e., does not consider process adaptability). Does not consider the dynamic behaviors projects that are significantly close to the customer and relationships associated with agency, the structural and behavioral instabilities and for innovations that are incremental or sustaining whole process and does not represent overall process configuration is fixed Assumes similar behavior across the freedom, and resulting innovations. the stages and presupposes that the in nature. Recursive frameworks emphasize the feed- Focuses on differences between back connections and nonlinearity in NPD, especially Limitations if greater levels of product newness or innovation are expected. Chaotic frameworks on the other hand as- sert that the initial stages of NPD tend to be chaotic and that the final stages are relatively ordered. of the process. Individually, each framework provides valuable in- sights and understanding about the behavior and structure of NPD processes. However, collectively they are more than just rival frameworks. As a group, they provide rich and holistic interpretations of the reliable market push or strong market pull forces. exploration aspects of very radical innovations or NPD processes and facilitate a contingency theory Recognizes different system behaviors across the process. Suited to more radical innovations with Represents the dynamic and fluid nature of the cumulative causation. Suited to the search and process and acknowledges the effects of highly incremental innovation activity with relatively Provides a simple and effective representation approach (Galbraith, 1977; Lawrence and Lorsch, of the structural logic and flows. Suited to 1967; Scott, 1987). This is the basis of the theoretical view and contribution in this article. It proposes that a push–pull market force combinations. CAS framework is capable of facilitating interpreta- tions of NPD reality that maintain a fit among de- Benefits scriptive stance, system behavior, and innovation type. Thus, the congruence of an individual NPD process is dependent on the connections and interac- really new products. tions between process agents. This system configura- Table 1. Linear, Recursive, and Chaotic Frameworks of New Product Development tion will govern the process ability to switch or toggle between behaviors that range from linear to chaotic to produce corresponding innovations that range from incremental to radical. of feedback across the process. The initial stages exhibit A process with concurrent and multiple feedback loops be random and unpredictable, whereas the latter stages during the initial stages, resulting in different degrees Complex Adaptive Systems A process where the linkages and flows are greater and outcomes that are more difficult to predict. chaotic dynamics and outcomes that appear to between stages that generate iterative behavior The concept and study of CASs originates in the life and physical sciences (Gell-Mann, 1994; Kauffman, Descriptive Interpretation and sequential stages. The connections, A process with relatively fixed, discrete flows, and outcomes of the process are 1993, 1995; Prigogine and Stengers, 1984); has been developed and used by the engineering sciences (Friz- are relatively stable and certain. elle and Suhov, 2001; Holland, 1995; Krothapalli and comparatively deterministic. Deshmukh, 1999); and has been discussed significant- ly by the social sciences in areas such as strategic or- ganizational design (Anderson, 1999; Brown and Eisenhardt, 1998; Dooley and Van de Ven, 1999; Eisenhardt and Bhatia, 2002; McKelvey, 1999), sup- ply chain management (Choi, Dooley, and Rung- tusanatham, 2001) and innovation management (Buijs, 2003; Chiva-Gomez, 2004; Cunha and Comes, 2003). To develop and to apply this perspective to the NPD Framework study of NPD, this section of the article introduces and defines a CAS and then shows how the key Recursive concepts of nonlinearity, self-organization, and Chaotic Linear emergence apply to NPD processes. This discussion provides the basis for the first two research
    • 442 J PROD INNOV MANAG I. P. MCCARTHY ET AL. 2006;23:437–456Figure 1. The Configuration of a Complex Adaptive Systempropositions that reflect the exploratory and descrip- high levels of predictability and efficiency but withtive nature of this research. They are concerned with low levels of adaptability (Eisenhardt and Bhatia,framing CAS concepts in NPD processes, as opposed 2002). Chaotic systems, on the other hand, are rela-to hypotheses that require measures (Eisenhardt and tively unstructured and loosely coupled, resulting inBhatia, 2002; Whetten, 1989). As part of the intro- outcomes that appear so random and disorganizedduction to CASs the limitations of the CAS perspec- that it is not possible for the system to adapt. A CAStive are considered, and it is argued that multiple is somewhere between a linear and a chaotic system,methods are required to examine the multilevel and with partially connected agents whose decision mak-nested nature of CASs. This provides a rationale for ing and interactions produce behavior and outcomesusing case studies to corroborate and to provisionally that are neither fully controlled nor arbitrary. Browntest the propositions. and Eisenhardt (1997) called this level of connectivity Complex adaptive systems are complicated systems semistructures. It produces system behavior that lieswith the ability to develop new system configurations between order—no change or periodic change—andand corresponding levels of order or disorder chaos—irregular change—and leads to the zone of(Dooley, 1997; Gell-Mann, 1994; Holland, 1995; system adaptability known as the edge of chaosSchoderbek, Schoderbek, and Kefalas, 1985). Appen- (Kauffman, 1993).dix A provides an account of the differences between a Complex adaptive systems consist of a nested andcomplicated system and a CAS using four systems di- scaleable system of agents; that is, the level of systemmensions. These are (1) the number of elements that abstraction could be an individual, a group, or an or-make up the system; (2) the attributes of the elements; ganization. To illustrate this, Figure 1 shows two lev-(3) the number and type of interactions among the els of abstraction. With the first level, the aggregateelements; and (4) the degree of organization inherent view, there are three agents (1, 2, 3), each is assumedin the system (Schoderbek Schoderbek, and Kefalas, to be a recognizable group of people. The agents par-1985). These dimensions illustrate how CASs differ tially connect to and interact with other agents. Theirfrom linear systems and chaotic systems, due to the actions and outputs— chosen options—are controlledstructure and connectivity between the system ele- by organizational rules and criteria (i.e., formal proc-ments, known as agents. Linear systems are highly ess procedures, checks). Their behavior, however,structured and tightly coupled, leading to relatively is also partly autonomous and affected by the
    • NEW PRODUCT DEVELOPMENT AS A COMPLEX ADAPTIVE SYSTEM OF DECISIONS J PROD INNOV MANAG 443 2006;23:437–456self-directed decisions and actions of other agents uncertainty about the use of research methods (Morelthroughout the CAS. The second level shown in and Ramanujam, 1999), and misconceptions that sys-Figure 1 provides a disaggregated view. It shows tem outcomes are purely random and therefore thatthat agent 2 is composed of three agents (2-1, 2-2, nothing can be predicted (Baumol and Benhabib,and 2-3), each assumed to be the individual members 1989; Radzicki, 1990). These problems have becomeof the group. The disaggregated view also shows more exacerbated as the interest in and expectations ofdetail, with a basic illustration of agent rules and CASs have grown. For example, there has been adecision option space. The option space includes both general anticipation that CAS research would pro-relatively known options (i.e., previous solutions to duce universal models and metatheories that apply tosimilar problems) and relatively unknown options all types of systems. However, as argued by Levy(i.e., novel solutions). As agents experiment with (1994) this is very unlikely, as the causes of adapta-and select and reject both rules and option space, bility in physical, life, and social systems each havethey produce a system agency and collective dynamic very different starting conditions. To acknowledgethat results in nonlinearity, self-organization, and these different system contexts, Bradach (1997),emergence. Together these CAS phenomena are the Brown and Eisenhardt (1997), and Eisenhardt andbasis of adaptability in complex adaptive systems Bhatia (2002) argued that the framing of CAS behav-(Anderson, 1999; Choi, Dooley, and Rungtusanat- ior in social systems will require qualitative researchham, 2001; Dooley and Van de Ven, 1999; McCarthy, approaches to complement the computational meth-2004; Morel and Ramanujam, 1999; Stacey, 1995). ods developed by the physical sciences (e.g., agent-The resulting behaviors and outcomes resist reduc- based models, cellular automata models). They sug-tionism, the practice of analyzing and describing phe- gest that case studies are well suited to capturing thenomena in terms of their simple or fundamental rich and qualitative features of social CASs, and theconstituents. This is because the behavior of a CAS highly iterative and descriptive nature of case studiesas a whole is different from the sum of the behaviors permits theory building and initial theory testing.of the individual agents that comprise the CAS. A A CAS framework of NPD extends the linear, re-CAS is therefore capable of producing behaviors and cursive, and chaotic perspectives by recognizing thatsystem changes that are sometimes inconsistent with nonlinearity and feedback can occur at multiple levelsany rules imposed on the system as a whole. between individual agents and between groups of If the notion of CASs is applied to NPD, then agents. This in turn drives self-organization and emer-agents represent people and groups of people—de- gence, which together enable the process to be malle-pending on the level of analysis—within the process. able and to exhibit different modes of behavior—They have the ability to receive and process informa- linear to chaotic. Existing NPD frameworks ignoretion, such as marketing or engineering, and to re- these CAS characteristics or use a reductionist ap-spond in accordance with their current internal rules proach to model and to aggregate behavior. The aimand their connections with other agents. They are able of the next section is to elaborate on how nonlinearity,to follow rules and be self-directed—that is, they are self-organization, and emergence appear in NPD andable to sense changes in project expectations and cir- to conclude with propositions concerning the poten-cumstance and to respond in an independent, timely, tial for complex adaptive behavior in NPD processes.and goal-directed fashion. Their current rules reflecttheir accumulated experience, but as they act and ex- Nonlinearity. NPD processes are composed ofperiment within these rules the outcomes—good or feedback loops due to their connections. The feed-bad—result in learning and potential modifications of back loops produce system sensitivity and dispropor-their internal rules and evaluation criteria. These com- tional outcomes known as nonlinearity. Whatbined attributes and interactions of the individual happens locally in one part of the process often doesagents govern the behavior and performance of the not necessarily apply to other parts of the same proc-NPD process as a whole. ess (Sterman, 2002). For example, if an NPD process The limitations and criticisms of the CAS perspec- is allocated extra resources, such as more designers,tive revolve around how the accompanying theory, the result is unlikely to be a proportional increase inmethods, and models have been developed and adopt- the number of ideas or products produced. The causesed. There is confusion and misuse of the terms of nonlinearity in NPD include inadequate or latecomplex and complexity (Eisenhardt and Bhatia, 2002), information—feedback problems—because the product
    • 444 J PROD INNOV MANAG I. P. MCCARTHY ET AL. 2006;23:437–456design space can be infinite. In addition, development Emergence is thus the product of self-organization.activities are typically a series of reiterations culminat- It is the manifestation of new process characteristicsing in a market launch event (Hart and Baker, 1994; due to the collective behavior of the agents, as op-Jin, 2000). The potential outcome from this nonlinear- posed to the individual behavior of each agentity is a chaotic dynamic similar to that identified by (Anderson, 1999; Holland, 1995; Kauffman, 1995;Flake (1999) and Sterman (2002) in their system dy- Waldrop, 1992). In the NPD context, it is the mate-namic models. Although the phenomenon of nonline- rialization of different process configurations, which,arity in NPD is central to the recursive and chaotic like the innovations they seek to produce, are oftenframeworks, they do not provide interpretations that not predetermined and sometimes are unwelcome andexplain and predict how scaled behavior occurs across unwanted. Emergence occurs because the process al-different process stages and decision levels. lows experimentation, rule breaking, and exploratory actions, all of which can generate novel behaviors and Self-organization and emergence. The second and corresponding levels of product innovation. Thisthird phenomena of a CAS are both facets of adapt- emergent capability is dependent on the NPD proc-ability and are the natural result of nonlinearity. Self- ess having a variety or surplus that is congruent withorganization is the rate of increase of order or regu- the array of changes an environment may impose onlarity in an NPD process (Von Foerster, 1960), and it the process. Thus, if a market requires high levels ofarises due to the autonomous behaviors of the various innovation, the NPD process should have corre-agents that constitute the process not because of ex- sponding levels of internal variety to facilitate explor-ternal or central control (Maturana and Varela, 1980). atory behavior. This condition is known as Ashby’sThis is not because there is an absence of formal con- law of requisite variety (Ashby, 1956) or as excess di-trol rules; rather, it is because the process as a whole versity (Allen, 2001) or as organization slack (Bour-independently adapts and develops new confi- geois, 1981; Nohria and Gulati, 1996).gurations. As discussed earlier, this self-organized In summary, the number and variety of agents in anorder arises when process agents are partially connect- NPD process, along with the resulting interactions,ed. If they were tightly coupled and highly structured determine the potential for nonlinearity, self-organi-then the process would be relatively rigid. On the zation, and emergence. These CAS phenomena areother hand, an unstructured and loosely coupled set both causes and characteristics of adaptability, pro-of agents would result in feedback and instabilities viding an individual NPD process with the ability tothat inhibit the emergence of regularity. That is, the generate a range of process behaviors and corre-process would be so recursive that it would struggle sponding innovation outputs in accordance with theto make progress toward new configurations and degree of order or chaos imposed on the process.behaviors. Formally stated these principles are the basis of Prop- Anderson (1999) stated that for a system to self-or- ositions 1 and 2:ganize it must maintain a level of internal energy thatis proportional to the level of disorder imposed on the Proposition 1: The rate of NPD process adaptability issystem by its environment. In terms of NPD, this determined by the rates of change and levels of stabilitymeans that the process should possess appropriate or disorder imposed on the process, which leads to con-amounts of passion, enthusiasm, and curiosity to cope gruence between the innovative output of the firm’swith the levels of stress, confusion, and uncertainty NPD processes—incremental versus radical—and thetransmitted to the process. The ability to generate needs of its environment.sufficient energy will depend on the number and Proposition 2: NPD process adaptability is determinedvariety of agents, along with the decision rules that by the number and variety of agents, their correspond-govern process coupling, structure, priorities, timing, ing connections and interactions, and the ordering orand permissions within which agents act (Eisenhardt disordering effect of the process rules and organization.and Sull, 2001; Chiva-Gomez, 2004). These factorsare consistent with Schoderbek, Schoderbek, andKefalas’s (1985) first and second system dimensions New Product Development Decisionsand will govern system autopoiesis, the ability to pre-serve a coherent process form while evolving new ones The principal role of NPD agents is to make decisions(Maturana and Varela, 1980). that bridge the gap between an idea and reality.
    • NEW PRODUCT DEVELOPMENT AS A COMPLEX ADAPTIVE SYSTEM OF DECISIONS J PROD INNOV MANAG 445 2006;23:437–456To examine how different NPD decisions influence deal with the cost of prototyping, the method ofprocess adaptability and resulting innovation outputs, prototyping to be used, and how the experimentsthis study uses the widely recognized NPD stages of should be designed (Thomke, Von Hippel, andconcept development, product design, testing and val- Franke, 1998). Finally, during the product launchidation, and launch and ramp-up and three levels of stage, the output of these decisions might concernNPD decisions: strategic, review, and in stage. the need for test marketing and the sequence in which At the highest level, NPD decisions relate to the products are introduced to the market. In the presentmarket and product strategy (Cooper and Kleinsch- article, these are called in-stage decisions.midt, 1986) and the funding and management of In addition to variations in NPD decision levels,project portfolios (Cooper, Edgett, and Kleinschmidt, different decision rules govern how the NPD process1999, 2002). Typically, these decisions set the initial is supposed to operate. Decision rules refer to the wayaims and objectives of an NPD project and then over- strategic, review, and in-stage NPD decisions aresee the process from a strategic level. They are con- made and shape the process in terms of agent at-cerned with the target market for a new product, its tributes and the degree of organization. These aredesired competitive position or fit with the organiza- consistent with Schoderbek, Schoderbek, and Kef-tion’s current product offerings, and the technology alas’s (1985) second and fourth system dimensions, asplatform to be employed (Bonner, Ruekert, and different decision rules will generate different types ofWalker, 2002). In the present article, these are called NPD behavior (e.g., rules that promote order versusstrategic decisions. rules that promote disorder) with varying degrees of The second level of decisions occurs between stag- uncertainty. Together, decision levels and decisiones. They are process milestone points that usually fol- rules create NPD progress paths, or trajectories,low the completion of development stages such as with outputs that range from incremental to radical.concept development, product design, and testing The CAS view reflects the NPD process in real(Clark and Fujimoto, 1991; Thomas, 1993). Such de- time, as open, changing, and potentially adaptive. Itcisions include whether the project should proceed or implies that process agency can yield different NPDbe terminated or whether it should be reviewed and behaviors with trajectories and outcomes that may orhow it should proceed (Cooper and Kleinschmidt, may not be predetermined by the inputs. To some ex-1991; Thomas, 1993). Decisions at this level are gen- tent, the nature of the trajectories will be consistenterally made by middle to senior managers, and with the linear, recursive, and chaotic NPD frame-the outcomes significantly affect NPD lead time work perspectives. For example, a trajectory where(Murmann, 1994). In the present article, these are the effect of a decision on the project would be the onecalled review decisions. anticipated—such as the improvement of a certain Finally, the third level of decisions relates to those product feature following customer feedback—is con-taken within and at the operational level of each sistent with the linear framework. In contrast, a cy-process stage (Krishnan and Ulrich, 2001). Creative clical path is consistent with the recursive framework,and exploratory behavior primarily occurs at this lev- as it will feed back and reuse similar and familiar de-el, in accord with the degree of agency and freedom cisions from previous and comparable projects. Fi-allowed or encouraged by the strategic and review nally, chaotic behavior in NPD is associated with thedecisions. At this level, the customer requirements are capacity to produce novelty or surprises, as any deci-assessed using a range of direct and derived impor- sion could create a potential change in trajectory di-tance measures (Griffin and Hauser, 1993), and the rection. This is because decision outcomes can becharacteristics of the product are decided. The out- amplified, producing paths that lead to good or badputs of these decisions vary for each stage of the de- process performance. The probability and degree ofvelopment process. So for instance, at the concept trajectory change will depend on which stage of thedevelopment stage the outputs might address the as- NPD process the decision takes place at and the hie-sessment of customer requirements—that is, product rarchal level of the decision making. It is important tospecification decisions. At the design stage, such de- note that chaotic trajectories are different from stoc-cisions could result in geometric models of assemblies hastic ones, which have such a rate of new and un-and components, a bill of materials, and control doc- wanted process inputs that the next process state orumentation for production (Krishnan and Ulrich, configuration cannot be deduced from the present one2001), whereas the testing and validation stage might (Stewart, 1989). In contrast, a chaotic trajectory is so
    • 446 J PROD INNOV MANAG I. P. MCCARTHY ET AL. 2006;23:437–456Table 2. Complex Adaptive System Framework of New Product DevelopmentNPD Framework Descriptive Interpretation Benefits LimitationsCAS A process with partially Assumes that overall Semantic confusion connected agents whose process configurations concerning the terms interactions cross and behaviors are malleable. complex and complexity. stages and decision They can be internally changed Challenges in framing levels. Collectively they are to match push or pull market and measuring the process able to produce a forces and innovation constructs coupled with process dynamic between expectations that range the misconception that order and chaos, from incremental to very radical. process outcomes are random which results in process and therefore unpredictable. adaptability and the potential to generate different behaviors and innovation outcomes.sensitive to system changes that it appears stochastic permit assessment and validation of the frameworkbut, in fact, exhibits patterns over time. and its accompanying propositions. This combined With this review of NPD decisions and building on top-down and bottom-up approach is an importantP1 and P2, it is possible to deduce that different de- and customary aspect of the research process. Itcision levels and decision rules will promote or con- facilitates theory building and the identification ofstrain agent coupling and interactions, which in turn possible correlations, which become the basis forwill determine the potential for adaptive behavior in quantitative descriptive models and normative theo-an NPD process. These factors combine with the CAS ry building (Brown and Eisenhardt, 1997; Eisenhardt,phenomena of nonlinearity, self-organization, and 1989; Feldman, 2004; Whetten, 1989).emergence and the system dimensions concerned The rationale for using case studies was to observewith the number and variety of agents, their coupling, and to describe the NPD process elements, interac-and degree of organization. Together they provide the tions, and behaviors within their real-life settings. Astheoretical constructs for a CAS framework of NPD a research method, it is well suited to relatively newsummarized in Table 2 and followed by Propositions research topics, especially when the phenomena are3 and 4. poorly understood and characterized (Eisenhardt, 1989). This is certainly the case with CAS interpreta- Proposition 3: Strategic and review decisions and their tions of NPD processes. The case-study approach also accompanying decision rules generate order and disor- provides a number of research contributions, includ- der at the in-stage decision level of NPD, which results ing description (Harris and Sutton, 1986), theory in corresponding potentials for process adaptability and building (Gersick, 1988), and initial theory testing matching innovations. (Pinfield, 1986). The following sections report how the case studies were selected and conducted. Proposition 4: In-stage decisions and their accompany- ing decision rules are able to produce enough internal energy to explore and to produce self-determined proc- ess adaptability. Company Selection Three companies were selected that would helpThe Study conceptualize and describe CAS behavior in NPD processes. This multiple case-study approach isThe process this research followed was highly itera- appropriate for examining phenomena such as CAStive. The review of existing NPD frameworks and ac- behaviors in more than one natural organizationalcompanying theories were primarily deductive in setting. The names of the case-study companies arenature and generated constructs and propositions not disclosed and will be referred to as Company A,concerning NPD as a CAS. The case studies present- Company B, and Company C.ed in this section provide observations and descrip- The companies were selected using criteria similartions of CAS phenomena in NPD processes that to those adopted by Swink, Sandvig, and Mabert
    • NEW PRODUCT DEVELOPMENT AS A COMPLEX ADAPTIVE SYSTEM OF DECISIONS J PROD INNOV MANAG 447 2006;23:437–456(1996). This required companies to (1) have substan- istence of a formal and documented NPD processtial experience in NPD; (2) be developing relatively within the organizations. To address this issue, thediscrete products; (3) be classified as producing rela- selected case-study companies provided process mapstively incremental and predictive innovations for ex- or workbooks that showed the NPD process used,isting customers and markets; and (4) collectively along with identifiable activities, stages, and resourc-represent a diversity of internal process characteris- es. The companies were also required to have formaltics in terms of size, rules, structure, and organization. and distinct operating budgets for their NPD proc-It was also important to exclude NPD processes char- esses. A similar criterion was also employed byacterized as unstructured, free thinking, and relatively McDermott and O’Connor (2002). To study theconstraint free, as the aim was to identify the existence fourth dimension, the selected companies needed toand potential for CAS behavior in a nonobvious NPD exhibit different NPD management rules, structures,environment. Therefore, the focus was on processes procedures, and technologies. For example, Companytypified as linear, organized, and controlled and with A operated a fixed multinational process based ona tendency to develop products in consultation with expertise—that is, the NPD team members were lo-their customers. In addition, the three case-study cated and organized in terms of geography and func-processes were required to be different according to tionality regardless of the project—whereas CompanySchoderbek, Schoderbek, and Kefalas’s (1985) four B, also a multinational business, organized its NPDsystem dimensions: number of elements, attributes of members according to a multidisciplinary matrixspecified elements, number of interactions among structure which changed according to the needs ofthese elements, and degree of inherent organization. the project. The process organization and structureThus, the selection constrained any potential case for Company C were fixed and informal, with thevariation due to extraneous process context—that is, same team working on all projects.government laboratories and other blue-sky researchfacilities were not considered—while focusing thestudy on the constructs central to the propositions. Data Collection and Analysis The first and third dimensions—number of ele-ments and number of interactions among those ele- Data were collected using questionnaires, structuredments, respectively—relate to the number of agents, interviews, and observation. The questionnaires werethe level of decision-making authority, and the types used to gather detailed information on the companies’and number of interactions that take place between NPD process. They included both closed and open-each agent. These dimensions are central to P1 and ended questions on the configuration of the process inP2; therefore, the case-study companies needed to ex- terms of stages; the number and professional types ofhibit variations in the size of their organizational and the people involved in the process; and informationNPD process settings. Consequently, Companies A about the decision levels, decision rules, and any ac-and B were large manufacturing organizations, and companying decision tools and techniques employed.their NPD teams consisted of more than 60 designers The senior managers—chief executives, vice presi-and engineers located in different parts of the world. dents, and directors—responsible for NPD in theCompany C was a relatively small organization, and case-study companies completed the questionnaires.its NPD team consisted of 12 engineers and designers, The questionnaire data were then used to design semi-all located in the same facility. In all three cases, the structured interviews for each of the case-study or-total number of people involved in the process varied ganizations. The interviews had two aims. The firstaccording to NPD stage, with the initial stages utiliz- was to overcome the limitations of questionnaire sur-ing fewer people than the middle and final stages. veys such as item nonresponse bias—when some ques- The second and fourth system dimensions—at- tions have not been answered—and nonconsistency intributes of the elements and the degree of organiza- understanding of questions (Hussey and Hussey,tion, respectively—relate to P2 and P4 and concern 1997). The second aim was to achieve a deeper un-the decision rules and management practices that gov- derstanding about the type of process organizationern NPD agents and the process. To examine the sec- and decision rules employed by the NPD teams. Toond dimension and to ensure that the case-study achieve this, 14 people in all were interviewed whoprocesses exhibited some control and organization collectively were involved at each stage and decisionover the agents, it was important to identify the ex- level of the NPD process. As reported by McDermott
    • 448 J PROD INNOV MANAG I. P. MCCARTHY ET AL. 2006;23:437–456and O’Connor (2002) and Eisenhardt (1989) this use example of nonlinearity in Company A and Compa-of multiple interviewees reduces the risk of undue in- ny B was what this present study calls exponential ex-fluence that an individual interviewee may have on the pectation change. This is when information such asstudy and brings a rounded and balanced insight into project due dates increase or decrease exponentiallyeach case. when passed from agent to agent. This occurred be- The final phase of data collection involved four sets cause agents at different stages and levels would seekof three-day visits at each company to observe current to cushion or to expedite process progress and out-NPD activities and to attend NPD meetings. The aim comes by altering this type of information. It is im-of this phase was to observe, in real time, the actions portant to note that these observations are notand behaviors of people and to triangulate and to necessarily novel. Their purpose and value is to con-validate the data collected from the previous two firm that such behaviors exist and then, using a CASstages. During these visits, secondary data such as lens, to better understand how simple rules of inter-Gantt charts, progress reports, and project notes were action, coupled with variations in agent-decision level,collected. Annual reports, company brochures, and can generate different aggregated outcomes that affectcorporate websites were also analyzed to gather back- process adaptability and innovation output. For ex-ground information about the case-study companies. ample, with Company A, the senior managers based at their parent company took the strategic NPD de- cisions. These decisions focused on selecting the port-Framing New Product Development as a folio of products that would satisfy the marketComplex Adaptive System requirements in terms of functionality, performance, and cost. The senior managers would also decide theIn this section, the cases are described to generate in- level of funding for and the timing and allocation ofsights and to conceptualize NPD as a CAS. This in- NPD projects for the company. Once these strategicvolves a summary of the cross-case analysis, where the decisions were made, the review and in-stage decision-documentation for each case was reviewed to identify making agents had relative autonomy to organizeand to categorize any common factors that constrain themselves and to develop new products within a rec-or promote CAS behavior in terms of nonlinearity, ognized framework and set of governing procedures.self-organization, and emergence. The relationship Thus, Company A’s strategic decisions had limitedamong these phenomena, the decision levels—in influence on any prospective nonlinear behavior in thestage, review, and strategic—and decision rules were NPD process.evaluated to further refine and to build insights that The opposite situation existed in Company C, thewould assess and provisionally test the propositions. smallest of the three case-study companies. The stra- tegic decisions in this NPD process were the respon- sibility of the chief executive, who would constantlyNonlinearity in New Product Development develop and change agreements with distributors and key customers, which in turn governed strategic deci-Despite the market-focused, project-controlled and sions concerning the type of product architecture, therelatively sequential nature of each NPD process stud- number of product variants, the supply chain, and theied, it was possible to identify nonlinear behavior in level of funding. The result was a process whereby theall three companies, especially at the in-stage decision strategic decisions and in-stage decisions were closelylevel. For example, when Company A and Company coupled, and any changes in the strategic decisionsB added new team members in an effort to advance significantly influenced in-stage behavior, generatingand to improve progress, the outcome was not a NPD outputs that could be nonlinear—amplifying orcorresponding improvement but was a temporary reducing—depending on the stability of the marketdecrease in progress as the system learned to and the judgment of the chief executive.accommodate, to train, and to integrate these new The review-decision level provides checks and con-agents during the first four weeks. This observation is trols for any major NPD milestones. The case studiesconsistent with existing research on the effects of provided examples of how the frequency and durationadding NPD resources and changing team character- of decisions at this level can affect the propensity foristics to increase NPD performance (Cooper and nonlinear behavior. For example, with Company A,Kleinschmidt, 1991; Gomes et al., 2001). Another the NPD director and a project-planning team
    • NEW PRODUCT DEVELOPMENT AS A COMPLEX ADAPTIVE SYSTEM OF DECISIONS J PROD INNOV MANAG 449 2006;23:437–456conducted review decisions on a bimonthly basis tive product development platforms, and computer-according to project progress, funding availability, aided design. The procedures and rules involved aand a reevaluation of market and product priority formal design change and documentation process; aobjectives. The relative regularity and complexity of make versus buy assessment for each designed com-these decisions exerted a dampening control that ponent; and an approval and sign-off procedure forsought to ensure a correlation between causes and an extensive range of in-stage product decisions foreffects. The situation was similar for Company B, but issues such as product color, material types, productrepresentatives from the NPD teams and the customer shape, prototyping plans, vendor selection, and test-companies undertook the review decisions jointly. ing schedules. Feedback, changes, and adjustments inWith Company C, the review decisions created posi- the process were also relatively formal to ensure trace-tive and negative feedbacks that affected the type and ability, accountability, and archiving of informationlevel of nonlinearity. This occurred when the chief for future NPD projects. By comparison, the in-stageexecutive, who regularly participated at this decision decision makers in Company C were all located in thelevel, made weekly decisions that reflected his approv- same office room, and the process of communicational or disapproval of the project progress. The scale throughout all stages of the NPD process was rela-and regularity of this decision feedback rapidly esca- tively informal, open, and unmonitored.lated and determined the direction of the project. In summary, the case studies revealed links be-These negative and positive feedback loops do not tween agency and nonlinearity across all three deci-respectively correspond to criticism or praise. Rather, sion levels, with the degree and type of nonlinearitya negative feedback loop keeps the system at a desired either constrained or propelled by the type andstate; it negates changes in the system; in contrast, a number of connections, interactions, and feedbackpositive feedback loop has an amplifying or self-rein- loops with agents at the various decision levels. Theforcing effect on the system that eventually makes the associated dynamics and outputs from these relation-system unstable and open to change. Thus, disap- ships underpin the tendency for process adaptabilityproval, or positive feedback, from the chief executive in terms of self-organization and emergence, whichwould quickly destabilize the project, counteract any are discussed in the next section.progress made, and lead to a reorganization of activ-ities, whereas approval, or negative feedback, wouldstrengthen and maintain the current NPD configura- Self-Organization and Emergence in New Producttion and trajectory. The frequency and intensity of the Developmentfeedback loops in Company C regularly created be-haviors such as system overconfidence, oscillation, As described earlier in this article, self-organizationgrowth, and overshoot as described by Sterman and emergence are simply two aspects of the same(1994) in his account of the fundamental modes of process capability: adaptability. New product devel-dynamic behavior. opment processes are able to adjust and to adopt new At the in-stage decision level, agents deal with mul- ways of working when their agents are free to connect,tiple decisions such as how the product components to communicate, and to cooperate in a way that pro-should be joined, if a certain material type would duces sufficient levels of self-organization. Adaptabil-work, what color it should be, and what shape the ity is a process behavior that cannot be attributed toproduct should have. Collectively these decisions in- any one agent, because, as observed in the case stud-volve producing and processing a rich diversity of ies, it is the result of communal and mutual decision-rules and criteria that are the basis of creative activ- making that occurs primarily at the in-stage level.ities and ideas in the NPD process. Thus, at the in- Thus, even though NPD agents may follow simplestage decision level the number of interactions and the rules or targets, it is possible for the process to gen-type of organization—formal control versus autono- erate behavior, which has the potential to be distinctlymous—imposed on the decision makers influenced different and possibly more effective than that origi-NPD nonlinearity. For example, Companies A and B nally designed and planned by the organization.used formal procedures, rules, and technologies to As with nonlinearity, all three case-study compa-permit communication among engineers located in nies exhibited self-organizing and emergent behavior.several countries. The technologies included group- Companies A and B used collaborative product de-ware meetings, virtual e-rooms, intranet, collabora- velopment software with process procedures and rules
    • 450 J PROD INNOV MANAG I. P. MCCARTHY ET AL. 2006;23:437–456to facilitate and to record interactions among all decisions because of significant customer involvementdecision levels and across all NPD stages. Yet despite and regulation requirements for its aerospace prod-this relatively high level of control and monitoring, ucts. Yet an occasion arose when process controlwhen decisions occurred outside the boundaries set by procedures were ignored, creating an unplanned in-formal practices and rules, instances of self-organizing novation. It occurred when a prototype componentbehavior were observed. For example, with Company failed an important functionality test. Instead of doc-A the concept for an entirely new product technology umenting and reporting the failure, it was concealeddeveloped during and after a game of golf that took from senior management to avoid an assessment ofplace during a personnel development course that the project progress and viability. Motivated by theseveral members of the NPD team happened to be fear of project failure and termination, the NPD teamattending. The outcome was an unplanned and tem- members secretly revaluated the test results and ex-porary social unit, with some of the NPD members plored substitute applications. The outcome was anmeeting in person for the first time. The result was a unplanned and radical innovation with a level of com-new process configuration with novel system interac- mercial success much greater than the original NPDtions and organization. As explained by one of the project application. This was a classic example of self-manufacturing engineers interviewed, ‘‘This innova- organization by rule breaking, whereby the decision-tion was the result of meeting and communicating in a making agents exhibit unpredictable, or loose-can-non-work place environment. We were able to coop- non, behavior. Thus, even though the most controllederate and problem solve without the normal bureau- of NPD processes may appear to operate according tocratic constraints that make up the day-to-day grind imposed organizational rules or accepted practices,of our jobs.’’ Again, the identification of this type of they have the potential for self-organizing and emer-brainstorming event is not unique and should not be gent behavior. Olin and Wickenberg (2001) arguedcredited to the use of a CAS framework. However, a that rule breaking is a positive attribute, as their re-CAS framework promotes and facilitates novel in- search shows that NPD processes that regularly breaksights based on questions about how this event rules tend to be more creative. A possible reason foremerged with no one individual planning, organizing, this is that rule breakers tend to be more entrepre-and controlling it; about what sort of parallel inter- neurial and exploratory in nature, whereas membersactions took place between the agents; and about of rule controlled NPD processes often feel con-what environmental conditions are required to facil- strained in their ability to discover new ideas, toitate the emergence of new working practices and take risks, or to develop solutions to problems.outcomes. These issues are relevant to management With Company C, its in-stage decision-making en-practice because in an attempt to replicate the success vironment was relatively unceremonious, dynamic,of the ad hoc and informal brainstorming event, Com- and disordered, which reflected the fluid and chang-pany A went on to organize a biannual three-day re- ing nature of its strategic and review decisions. Thetreat known as Exploration Days. These events were result was a NPD process that was highly adaptabledesigned to allow NPD members to work, to socialize, with informal and new behaviors created regularly inand to learn in a rule-free environment. Yet partici- response to the confusing and changing rules and un-pants from the original event who had also attended known constraints and product specifications. Thisthe Exploration Days stated that the two events were process of self-organization was a forceful and disor-very different. With Exploration Days, the environ- ganized combination of top-down and bottom-up in-ment may have fewer regulations and restrictions teractions. The top-down interactions were governedcompared to the normal workplace, but there are still by the hierarchal structure and authority of the sys-expectations and rules not present during the golf tem in terms of decision levels such as strategic deci-event. For example, attendance at Exploration Days sions, whereas the bottom-up interactions were due tois compulsory for all NPD members, and the purpose the agency at the lower level of the hierarchical struc-is clear: to produce ideas. ture such as the in-stage decisions. Also, for Company Another instance where self-organizing behavior C’s relatively small size, the NPD process exhibitedresulted in new emergent NPD practices was when the greatest number and variety of system connectionsNPD team members in Company B decided to disre- and interactions. This helped produce an emergentgard the process rules and procedures. Company B dynamic that was consistent with the notion of beingsought to formalize and to record all in-stage far from equilibrium (Prigogine and Stengers, 1984)
    • NEW PRODUCT DEVELOPMENT AS A COMPLEX ADAPTIVE SYSTEM OF DECISIONS J PROD INNOV MANAG 451 2006;23:437–456or at the edge of chaos (Kauffman, 1993), whereby the greater levels of process stability. Their environmentNPD process rarely maintained an ordered pattern of had historically favored incremental innovations;behavior. This state occurs when there are enough accordingly, the processes were relatively formal andconnections to be dynamic, nonlinear, and adaptable, linear. With Company C, the process was exposed tobut not so many as to make the system unstable or to an environment that rapidly and frequently switchedcause it to disintegrate. To maintain a position at the between a relatively certain and stable environment toedge of chaos required a continual and appropriate one that was highly turbulent and unpredictable.flow of energy importation—strategic decisions— Consequently, it was observed that the process wasbalanced with an appropriate level of operational also able to quickly adapt from a controlled and rel-agency—in-stage decisions. atively static process to a relatively fluid and respon- A final observation from all three case studies was sive process. Although the aim of this study was not tothe tendency for nonlinearity, self-organization, and measure and to relate the performance of the NPDemergence to vary according to the NPD stage. This process to the type of configuration, other studies ofwas particularly the case with Companies A and B theory and practice do show a correlation. With allwhere the size of the teams and the duration of the three case-study companies, the observations did notprojects provided an appropriate disconnect between refute or support the notion that variations in processconcept development and product launch. Relative to stability or disorder would translate directly and re-the decision makers in the later stages, those at the spectively into incremental or radical innovations.concept development stage had more autonomy to This is simply because of the time lag between con-develop their own ideas. As explained by the NPD ducting the case studies and the eventual launch anddirector of Company A, ‘‘. . . with concept develop- market adoption, or rejection, of the associated inno-ment we need to have some control, but if we try to vation outputs. Thus, it was not possible to categorizecontrol this stage finitely and take all the decisions for adequately any resulting innovations as either suc-them, then that limits their influence and creativity.’’ cessful or unsuccessful or as incremental or radical. However, a relationship is known to exist between system behavior and innovation output, and thisDiscussion study did find that NPD process adaptability occurred to maintain congruence between process behavior andThe notion of viewing NPD processes as a CAS is the demands of the environment. Thus, it is proposedtheoretically attractive. However, how does this sys- that this element of P1 is still valid and worthy oftems perspective differ from existing NPD frame- future verification.works, and how do CAS concepts help to improve The case studies were limited in number and fo-understanding of NPD? These questions were the mo- cused on NPD processes characterized by a customertivation behind this study, which sought to develop pull for engineered products combined with relativelyand to describe the basis of a CAS framework using a low levels of market uncertainty and complexity. Asystematic review of the literature and by applying the consequence of P1 is that NPD process diversity andframework to case-study companies. In this section of the potential for fit and misfit will be much greaterthe article, the resulting case observations and analy- than the cases presented in this article. Yet even withsis are discussed to assess the propositions and to de- this limited range of environmental variety, the casetermine the potential for and consequences of CAS studies yielded different process behaviors with cor-behavior in NPD. responding potentials for process adaptability. Pre- The first proposition predicted that NPD processes dicting the factors that govern this processconform to contingency theory and that their ability adaptability is the basis of P2, which asserts that theto adapt is governed by the rates of change and the number and variety of agents, their correspondinglevels of stability or disorder imposed on the process. connections and interactions, and the effect of theThe result is potential congruence between the inno- process rules and organization will determine the po-vative output of the NPD process—incremental ver- tential for process adaptability. With each of the casesus radical—and the needs of its environment. With studies, there were significant variations in how thethe context variations among Companies A, B, and C, agent tasks and roles differed (e.g., marketing, design,it was clear that the relatively ordered and controlled test) along with variations in the cognitive norms thatenvironment of Companies A and B resulted in accompanied each task and role. This diversity
    • 452 J PROD INNOV MANAG I. P. MCCARTHY ET AL. 2006;23:437–456increases when considering the different process can go beyond existing preconceptions of how NPDdecision levels—strategic, review, and in stage—that processes operate. In addition, the process of applyingagents occupy and the types of connections between the the framework to the case-study companies does helpagents. The absolute number and variety of agents and to validate or to refute the propositions, which in turnresulting connections were greater in Companies A and can help translate constructs into measures and prop-B than in Company C. However, relative to the number ositions into hypotheses. Weaknesses of the case-of agents, Company C had a greater degree of agent study method include the potential for resulting the-variety, connectivity, and interactivity. The case-study ories that are overly complex as they too closely reflectfindings also showed how the connections between the reality they have just studied. There is also theagents were based on interactions that could be phys- possibility that any resulting theoretical contributionical, social, or electronic in nature, with different ca- could be too narrow due to the specifics of the datapacities for promoting or restricting CAS behavior. and the nature of the case studies. As this study was P3 and P4 predict how decision levels and rules exploratory in nature and was motivated and guidedpromote or constrain process adaptability. The deci- by two established but separate research areas—de-sion levels proposed and studied by this research pro- scriptive interpretations of NPD and the study ofvide architectural constructs for understanding the CASs—it is unlikely that these issues will adverselycapacity for autonomous decision making and social affect the findings and contribution.action in the process. The case-study findings showedthat the degree of coupling between decision levelsand the frequency and type of feedback between Conclusionsagents at each level have the potential to produce orto limit CAS characteristics in NPD. This is because The theories and methods for defining and modelingthe relationships between agents across decision levels CASs are broad, complicated, and often misunder-are influenced by their co-evolving behavior, their re- stood as being an area of research concerned withspective contiguity, and their appreciation of CAS solving the problem of complexity. Instead, they pro-behaviors across the process. The strategic and review vide a perspective for theorizing and modeling sys-decision rules exist to direct and to control the NPD tems based on the supposition that it is not possible toprocess and, as a result, play a significant role in de- understand properly overall system behaviors by sim-termining the intensities and kinds of interactions at ply breaking the system down into constituent partsthe in-stage decision level. The agents at strategic and and then aggregating any observations to developreview levels are able to grant the necessary freedom convenient and linear rules. This is because CASsor control and to respectively engender or to stifle are composed of agents, whose decision-making abil-nonlinearity and the resulting self-organization and ities result in self-determination and exploration atemergence at the in-stage level. In effect, the strategic multiple levels within the system.and review decisions create a top-down management Overall, this research proposes and supports theenergy that equates to order or disorder at the in-stage notion that a CAS framework of NPD has descriptivedecision level, though behaviors associated with fear, value in terms of studying, classifying, and definingcuriosity, obsession, naughtiness, enthusiasm, and the attributes and relationships that govern adaptiveanxiety can at the in-stage level create a level of in- behavior and outcomes in NPD processes. A CASternal energy that results in a bottom-up generated framework provides insights that complement the lin-process adaptability. ear, recursive, and chaotic frameworks by inherently Finally, this study’s process of building theory and maintaining a fit among descriptive stance, systemprovisionally testing propositions using case-study re- behavior, and innovation type. This is because a CASsearch was consistent with Eisenhardt’s (1989, p. 546) framework views NPD processes as systems of par-account that case-study research bounces between de- tially connected agents, operating within strategic andduction and induction, as ‘‘. . . an investigator may review-decision rules that promote or constrain themove from cross-case comparison, back to redefini- potential for self-organization and emergence at thetion of the research question, and out to the field to in-stage level. The result is adaptability, whereby in-gather evidence . . .’’ This type of approach is appro- dividual NPD processes are able to switch betweenpriate for descriptive research, as it involves resolving different process behaviors and the assumptions of thecase-study data to develop new interpretations, which corresponding NPD framework.
    • NEW PRODUCT DEVELOPMENT AS A COMPLEX ADAPTIVE SYSTEM OF DECISIONS J PROD INNOV MANAG 453 2006;23:437–456 Although this contribution is descriptive and aca- resulting range of system states. Such research coulddemic in orientation, related implications can be help quantitatively address questions concerning howfound concerning how managers might develop and many agents are necessary for self-organization, whatapply normative models to create and to operate NPD is a corresponding and appropriate concentration ofprocesses. Managers will need to consider how their connections, and what type and frequency of feed-traditional methods and tools will affect future proc- back is required to stabilize or to disorder the systemess congruence and performance, bearing in mind that according to its environmental characteristics and de-in terms of their size, structure, and behaviors, indi- velopment stage. This would involve simulation andvidual NPD processes are not eternally fixed. As with regression models that predict the process variablethe environments they serve, they are able to adapt, values needed to switch between behaviors and thewhich means that an individual NPD process has the impact of top-down directed process adaptation ver-potential for more than one system behavior. Thus, sus bottom-up self-regulated adaptation. As this in-market expectations, innovation levels, process be- volves modeling the structural and temporalhaviors, and management approaches will all be de- dimensions of NPD processes, the use of graph the-pendent on each other in terms of system control and ory techniques such as NK models (Kauffman, 1993)reliability versus system freedom and autonomy. or Q-analysis (Atkin, 1978; Rakotobe-Joel, McCar- To design and to manage processes accordingly will thy, and Tranfield, 2002) would be appropriate. Theserequire comprehension of how decision rules, agent methods provide graphical and mathematical meth-variety, and agent connectivity influence the pace of ods for eliciting the relationships between two or moreprocess self-organization and emergence. This is be- agents or sets of agents; therefore, it would be possiblecause the agents within a NPD process are multiat- to represent process connections, interactions, and thetribute switches whose links and outputs define the presiding decision rules.architecture, flows, and performance of the process.Such management insights would help avoid systemlock-in or competency traps, where process configu-rations become contradictory or inconsistent with the Referencesmix of environmental and organizational factors thatgovern NPD goals and success. For example, if the Adams, R. (2003). Perceptions of Innovations: Exploring and Develop- ing Innovation Classification. Ph.D. diss., Cranfield University,rules and procedures from past NPD projects are Cranfield, UK.simply applied to future projects that have new inno- Allen, P.M. (2001). A Complex Systems Approach to Learning invation and market expectations, then this contingency Adaptive Networks. International Journal of Innovation Manage- ment 5(2):149–80.ignorance is likely to lead to NPD system behavior Anderson, P. (1999). Complexity Theory and Organization Science.and innovation outcomes that are inappropriate in Organization Science 10(3):216–32.terms of cost, time, and level of novelty. The proba- Ashby, R. (1956). Introduction to Cybernetics. New York: John Wiley.bility that this incongruence will occur is exacerbated Atkin, R.H. (1978). Q-Analysis: A Hard Language for the Soft Sci- ences. Futures December, 492–9.if the organization that operates the NPD process also Badaracco, J.L. Jr. (1991). The Knowledge Link. Boston: Harvardhas company-wide practices and performance models Business School Press.that motivate and reward behavior inconsistent with Baumol, W.J. and Benhabib, J. (1989). Chaos: Significance, Mecha-the desired NPD market, expectations, innovation nism, and Economic Applications. Journal of Economic Perspec- tives 3(1):77–105.levels, and process behaviors. Awareness of these sim- Bonner, J.M., Ruekert, R.W. and Walker, O.C. (2002). Upper Man-ple but high-level design rules will help managers agement Control of New Product Development Projects andshape and direct an individual NPD process to pro- Project Performance. Journal of Product Innovation Management 19(3):233–45.duce different innovations—incremental to radical— Bourgeois, L.J. III (1981). On the Measurement of Organizationalfor different environments. This approach to NPD Slack. Academy of Management Review 6(1):29–39.design would also encourage a strategic bucket ap- Bradach, J.L. (1997). Using the Plural Form in the Management ofproach to resource allocation within and across NPD Restaurant Chains. Administrative Science Quarterly 42:276–303. Brown, S.L. and Eisenhardt, K.M. (1997). The Art of Continuousprojects. Change: Linking Complexity Theory and Time-Paced Evolution in Finally, to extend and to test the research presented Relentlessly Shifting Organizations. Administrative Science Quar-in this article will require empirical models that go terly 42(1):1–35. Brown, S.L. and Eisenhardt, K.M. (1998). Competing on the Edge:beyond the case-study insights and measure the rela- Strategy as Structured Chaos. Boston: Harvard Business Schooltionships between the framework constructs and the Press.
    • 454 J PROD INNOV MANAG I. P. MCCARTHY ET AL. 2006;23:437–456Buijs, J. (2003). Modelling Product Innovation Processes, from Linear Flake, G.W. (1999). The Computational Beauty of Nature. Cambridge Logic to Circular Chaos. Creativity & Innovation Management MA: MIT Press. 12(2):76–93. Ford, D. and Sterman, J. (1998). Dynamic Modeling of Product De-Cheng, Y.T. and Van De Ven, A. (1996). Learning the Innovation velopment Processes. System Dynamics Review 14(1):31–68. Journey: Order Out of Chaos? Organization Science 7(6):593–614. Frizelle, G. and Suhov, Y.M. (2001). An Entropic Measurement ofChiva-Gomez, R. (2004). Repercussions of Complex Adaptive Systems Queueing Behaviour in a Class of Manufacturing Operations. Pro- on Product Design Management. Technovation 24(9):707–11. ceedings of the Royal Society of London Series A-MathematicalChoi, T.Y., Dooley, K.J. and Rungtusanatham, M. (2001). Supply Physical and Engineering Sciences 457(2011):1579–1601. Networks and CAS: Control versus Emergence. Journal of Opera- Galbraith, J.R. (1977). Organization Design. Reading, MA: Addison- tions Management 19(3):351–66. Wesley.Clark, K.B. and Fujimoto, T. (1991). Product Development Perfor- Gell-Mann, M. (1994). The Quark and the Jaguar. New York: Freeman mance: Strategy, Organization and Management in the World Auto & Co. Industry. Boston: Harvard Business School Press. Gersick, C. (1988). Time and Transition in Work Teams: Toward aClark, K.B. and Wheelwright, S.C. (1993). Managing New Product and New Model of Group Development. Academy of Management Process Development: Text and Cases. New York: Free Press. Journal 31(1):9–41.Clift, T.B. and Vandenbosch, M.B. (1999). Project Development and Gomes, J., De Weerd-Nederhof, P., Pearson, A. and Fisscher, O. Efforts to Reduce Product Development Cycle Time. Journal of (2001). Senior Management Support in the New Product Develop- Business Research 45(2):187–98. ment Process. Creativity and Innovation Management 10(4):234–42.Constant, E. (2000). Recursive Practice and the Evolution of Techno- Griffin, A. (1997a). The Effect of Project and Process Characteristics logical Knowledge. In: Technological Innovation as an Evolutionary on Product Development Cycle Time. Journal of Marketing Re- Process. J. Ziman (ed.). Cambridge, UK: Cambridge University search 34(1):24–35. Press, 219–33. Griffin, A. (1997b). PDMA Research on New Product DevelopmentCooper, R.G. (1990). Stage Gate Systems: A New Tool for Managing Practices: Updating Trends and Benchmarking Best Practices. New Products. Business Horizons 33(3):44–53. Journal of Product Innovation Management 14(6):429–58.Cooper, R.G. (1993). Winning at New Products. Reading, MA: Addi- Griffin, A. and Hauser, J.R. (1993). The Voice of the Customer. Mar- son Wesley. keting Science 12(1):1–27.Cooper, R.G. and Kleinschmidt, E.J. (1986). An Investigation into the Harris, S. and Sutton, R. (1986). Functions of Parting Ceremonies in New Product Development Process: Steps, Deficiencies, and Im- Dying Organizations. Academy of Management Journal 29(1):5–30. pact. Journal of Product Innovation Management 3(2):71–85. Hart, S.J. and Baker, M.J. (1994). The Multiple Convergent ProcessingCooper, R.G. and Kleinschmidt, E.J. (1991). The Impact of Product Model of New Product Development. International Marketing Re- Innovativeness on Performance. Journal of Product Innovation view 11(1):77–92. Management 8(4):240–52. Holland, J.H. (1995). Hidden Order: How Adaptation Builds Complex-Cooper, R.G. and Kleinschmidt, E.J. (1993). Major New Products: ity. Reading, MA: Addison-Wesley. What Distinguishes the Winners in the Chemical Industry? Journal Hussey, J. and Hussey, R. (1997). Business Research. Basingstoke, UK: of Product Innovation Management 10(2):90–111. Macmillan Press.Cooper, R.G., Edgett, S.J. and Kleinschmidt, E.J. (1999). New Product Jin, Z. (2000). How Product Newness Influences ‘‘Learning and Prob- Portfolio Management: Practices and Performance. Journal of ing’’ and the Linearity of Its Development Process. Creativity and Product Innovation Management 16(4):333–51. Innovation Management 9(1):21–45.Cooper, R.G., Edgett, S.J. and Kleinschmidt, E.J. (2002). Portfolio Johne, F.A. and Snelson, P. (1988). Success Factors in Product Inno- Management for New Products, 2nd ed., Reading, MA: Perseus vation: A Selective Review of the Literature. Journal of Product Books. Innovation Management 5(2):114–28.Crawford, C.M. and Di Benedetto, C.A. (2000). New Products Man- Kauffman, S.A. (1993). The Origins of Order: Self-Organization and agement, 6th ed., New York: McGraw-Hill. Selection in Evolution. New York: Oxford University Press.Cunha, M.P. and Comes, J.E.S. (2003). Order and Disorder in Product Kauffman, S.A. (1995). At Home in the Universe: The Search for the Innovation Models. Creativity & Innovation Management Laws of Self-Organization and Complexity. New York: Oxford 12(3):174–87. University Press.Dewar, R.D. and Dutton, J.E. (1986). The Adoption of Radical and Kline, J. and Rosenberg, N. (1986). An Overview of Innovation. In: Incremental Innovations: An Empirical Analysis. Management Sci- The Positive Sum Strategy: Harnessing Technology for Economic ence 32(11):1422–33. Growth. R. Landau and N. Rosenberg (eds.). Washington, DC:Dooley, K. (1997). A Complex Adaptive Systems of Model of Organ- National Academy Press, 275–305. ization Change. Nonlinear Dynamics, Psychology & Life Science 1(1):69–97. Koput, K. (1997). A Chaotic Model of Innovative Search: Some An- swers, Many Questions. Organization Science 8(5):528–42.Dooley, K. and Van de Ven, A. (1999). Explaining Complex Organ- izational Dynamics. Organization Science 10(3):358–72. Krishnan, V. and Ulrich, K.T. (2001). Product Development Deci- sions: A Review of the Literature. Management Science 47(1):Dougherty, D. (1992). Interpretive Barriers to Successful Product In- 1–21. novation in Large Firms. Organization Science 3(2):179–202. Krothapalli, N. and Deshmukh, A. (1999). Design of NegotiationEisenhardt, K.M. (1989). Building Theories from Case Study Re- Protocols for Multi-agent Manufacturing Systems. International search. Academy of Management Review 14(4):532–50. Journal of Production Research 37(7):1601–24.Eisenhardt, K.M. and Bhatia, M.M. (2002). Organizational Complex- Lawrence, P.R. and Lorsch, J.W. (1967). Organization and Environ- ity and Computation. In: Companion to Organizations. J.A.C. ment: Managing Differentiation and Integration. Boston: Harvard Baum (ed.). Oxford: Blackwell, 442–66. University.Eisenhardt, K.M. and Sull, D.N. (2001). Strategy as Simple Rules. Leifer, R., McDermott, C.M., O’Connor, G.C., Peters, L.S., Rice, Harvard Business Review 79(1):107–16 (January). M.P., Veryzer, R.W. and Rice, M. (2000). Radical Innovation: HowFeldman, M.S. (2004). Resources in Emerging Structures and Proc- Mature Companies Can Outsmart Upstarts. Boston: Harvard Busi- esses of Change. Organization Science 15(3):295–309. ness School Press.
    • NEW PRODUCT DEVELOPMENT AS A COMPLEX ADAPTIVE SYSTEM OF DECISIONS J PROD INNOV MANAG 455 2006;23:437–456Leonard-Barton, D. (1988). Implementation as Mutual Adaptation of Schoderbek, P.P., Schoderbek, C.G. and Kefalas, A.G. (1985). Man- Technology and Organization. Research Policy 17(5):251–67. agement Systems, Conceptual Considerations. Plano, TX: BusinessLevy, D. (1994). Chaos Theory and Strategy: Theory Application and Publications Inc. Managerial Implications. Strategic Management Journal 15:167–78 Schroeder, R.G., Van De Ven, A.H., Scudder, G.D. and Polley, D. (Summer). (1989). The Development of Innovation Ideas. In: Research on theMaturana, H. and Varela, F. (1980). Autopoiesis and Cognition: The Management of Innovation: The Minnesota Studies. A.H. Van de Realization of the Living Boston Studies. In: Philosophy of Science. Ven, H.L. Angle and M. Poole (eds.). New York: Harper & Row, R.S. Cohen and W.W. Marx (eds.). Dordecht: D. Reidel Publish- 107–33. ing, 42. Scott, R.W. (1987). Organizations: Rational, Natural and Open Sys-McCarthy, I.P. (2004). Manufacturing Strategy—Understanding the tems. Englewood Cliffs, NJ: Prentice Hall. Fitness Landscape. International Journal of Operations and Produc- Shepherd, C. and Ahmed, P.K. (2000). NPD Frameworks: A Holistic tion Management 24(2):124–50. Examination. European Journal of Innovation ManagementMcDermott, C.M. and O’Connor, G.C. (2002). Managing Radical In- 3(3):160–73. novation: An Overview of Emergent Strategy Issues. Journal of Smith, R.P. and Eppinger, S.D. (1997). A Predictive Model of Sequen- Product Innovation Management 19(6):424–38. tial Iteration in Engineering Design. Management ScienceMcKelvey, B. (1999). Self-Organization, Complexity, Catastrophe, and 43(8):1104–20. Microstate Models at the Edge of Chaos. In: Variations in Organi- Song, X.M. and Montoya-Weiss, M.M. (1998). Critical Development zation Science—in Honor of Donald T. Campbell. J.A.C. Baum and B. Activities for Really New versus Incremental Products. Journal of McKelvey (eds.). Thousand Oaks, CA: Sage Publications, 279–307. Product Innovation Management 15(2):124–35.Moenaert, R.K., Caeldries, F., Lievens, A. and Wauters, E. (2000). Stacey, R.D. (1995). The Science of Complexity: An Alternative Per- Communication Flows to International Product Innovation spective for Strategic Change. Strategic Management Journal Teams. Journal of Product Innovation Management 17(5):360–77. 16(6):477–95.Morel, B. and Ramanujam, R. (1999). Through the Looking Glass of Sterman, J.D. (1994). Learning in and about Complex Systems. System Complexity: The Dynamics of Organizations as Adaptive and Dynamics Review 10(2–3):291–330. Evolving Systems Complexity. Organization Science 10(3):278–93. Sterman, J.D. (2002). All Models Are Wrong: Reflections on BecomingMuffatto, M. and Roveda, M. (2000). Developing Product Platforms: a System Scientist. System Dynamics Review 18(4):501–31. Analysis of the Development Process. Technovation 20(11):617–30. Stewart, I. (1989). Does God Play Dice? The Mathematics of Chaos.Murmann, P.A. (1994). Expected Development Time Reductions in Cambridge, MA: Blackwell. German Mechanical Engineering Industry. Journal of Product In- Swink, M.L., Sandvig, J.C. and Mabert, V.A. (1996). Customizing novation Management 11(3):236–52. Concurrent Engineering Processes: Five Case Studies. Journal ofNohria, N. and Gulati, R. (1996). Is Slack Good or Bad for Innova- Product Innovation Management 13(3):229–44. tion? Academy of Management Journal 39(5):1245–64. Thomas, J.R. (1993). New Product Development—Managing and Fore-Olin, T. and Wickenberg, J. (2001). Rule Breaking in New Product casting for Strategic Success. New York: John Wiley & Sons. Development—Crime or Necessity? Creativity and Innovation Man- Thomke, S.E., Von Hippel, E. and Franke, R. (1998). Modes of Ex- agement 10(1):15–25. perimentation: An Innovation Process—and Competitive—Varia-Page, A.L. (1993). Assessing New Product Development Practices and ble. Research Policy 27(3):315–32. Performance: Establishing Crucial Norms. Journal of Product In- Ulrich, K. and Eppinger, S. (1995). Product Design and Development. novation Management 10(4):273–90. New York: McGraw-Hill Inc.Paich, M. and Sterman, J.D. (1993). Boom, Bust, and Failures to Learn Utterback, J.M. (1996). Mastering the Dynamics of Innovation. Boston: in Experimental Markets. Management Science 39(12):1439–58. Harvard Business School Press.Pinfield, L.T. (1986). A Field Evaluation of Perspectives on Organiza- Van de Ven, A.H., Polley, D.E., Garud, R. and Venkataraman, S. tional Decision Making. Administrative Science Quarterly 31(3): (1999). The Innovation Journey. New York: Oxford University 365–88. Press.Prahalad, C.K. and Hamel, G. (1990). The Core Competences of the Von Foerster, H. (1960). On Self-Organising Systems and Their Envi- Corporation. Harvard Business Review 68(3):79–91. ronments. In: Self-Organising Systems. M.C. Yovits and S. Came-Prigogine, I. and Stengers, I. (1984). Order out of Chaos. New York: ron (eds.). London: Pergamon Press, 30–50. Bantam Books. Waldrop, M.M. (1992). Complexity: The Emerging Science at the EdgeQuinn, J.B. (1985). Managing Innovation: Controlled Chaos. Harvard of Order and Chaos. New York: Simon & Schuster. Business Review 63(3):73–84. West, M.A. (1990). The Social Psychology of Innovation in Groups.Radzicki, M.J. (1990). Institutional Dynamics, Deterministic Chaos, and In: Innovation and Creativity at Work: Psychological and Organi- Self-Organizing Systems. Journal of Economic Issues 15(1):57–102. zational Strategies. M.A. West and J.L. Farr (eds.). Chichester,Rakotobe-Joel, T., McCarthy, I.P. and Tranfield, D. (2002). A Struc- UK: John Wiley & Sons, 309–33. tural and Evolutionary Approach to Change Management. Jour- Wetherbe, J.C. (1995). Principles of Cycle Time Reduction: You Can nal—Computational & Mathematical Organization Theory 8(4):337– Have Your Cake and Eat It Too. Cycle Time Research 1(1):1–24 64. (Memphis: Fedex Center for Cycle Time Research).Rechtin, E. (1991). Systems Architecting: Creating And Building Com- Wheelwright, S. and Clark, K. (1995). Leading Product Development. plex Systems. Englewood Cliffs, NJ: Prentice-Hall. New York: Free Press.Repenning, N. (2001). Understanding Fire Fighting in New Product Whetten, D.A. (1989). What Constitutes a Theoretical Contribution. Development. Journal of Product Innovation Management 18(5): Academy of Management Journal 14(4):490–5. 285–300. Winner, R.I., Pennell, J.P., Bertrand, H.E. and Slusarczuk,Rothwell, R. (1992). Successful Industrial Innovation: Critical Factors M.G. (1988). The Role of Concurrent Engineering in Weapons for the 1990s. R&D Management 22(3):221–39. Systems Acquisition. IDA Report R-338, Institute for DefenseSchmidt, J.B. and Calantone, R.J. (1998). Are Really New Product Analysis, Alexandria, VA. Development Projects harder to Shut Down? Journal of Product Zaltman, G., Duncan, R. and Holbek, J. (1973). Innovations and Or- Innovation Management 15(2):111–23. ganizations. New York: John Wiley & Sons, Inc.
    • 456 J PROD INNOV MANAG I. P. MCCARTHY ET AL. 2006;23:437–456 Appendix AUsing Schoderbek, Schoderbek, and Kefalas’s (1985) system properties framework it is possible to distinguishbetween complicated systems and CASs. The framework defines a system in general, as a set of elements withattributes that are connected by relationships to each other and to their environment to form a whole. It alsoproposes four dimensions that govern the control and behavior of a system and consequently distinguish CASsfrom complicated systems. These are (1) the number of elements that make up the system; (2) the attributes ofthe elements; (3) the number and type of interactions among the elements; and (4) the degree of organizationinherent in the system. From the given definition, the notion of a set refers to a collection of elements that are obviously meant to bewithin that set—for example designers within design departments, or a group of engineering and marketingpersonnel within NPD teams. The system elements not only constitute the system, but they are also the system’sfunction. Each element will have attributes that reflect their properties and characteristics and thus determine thevariety of elements in the system. For NPD processes, this refers to decision-related attributes as determined byjob roles, skills, experience, and authority. System relationships are the interactions that connect the elements.They can be symbiotic, synergistic, or redundant, and for every relationship in a system there will be some formof interaction and organization. The system’s environment will include any other system or element whosechanges in attributes have an effect on the system. For NPD processes, the environment would include cus-tomers, competitors, and competing resources in the same organization. The final feature of the system def-inition, wholeness, indicates that a system is a meaningful family of elements, relationships, and attributes. Thereis natural purpose and a degree of organization governing the system’s existence. With this framework, a complicated system such as mechanical clock will have a somewhat high number ofelements with relationships, attributes, and interactions that are relatively fixed and unchanging. This results in acomparatively low level of system organization with corresponding order, stability, and predictability. Suchsystem features make it possible to understand, to model, and to reproduce complicated systems by dismantlingthe system to its constituent elements, known as reductionism. With a CAS, the system is still relatively com-plicated, but the system elements have the ability to change their individual attributes and interactions to pro-duce new system configurations and behaviors. It is this ability to adapt that distinguishes a CAS from acomplicated system.