1. aoshima, y. industrial relations

Upload: filipe-fagundes

Post on 06-Jul-2018

218 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    1/24

    Transfer of System Knowledge Across

    Generations in New Product Development:Empirical Observations from Japanese

    Automobile Development

    YAICHI AOSHIMA*

    The objective of this article is to explore effective ways of retaining knowledge

    involved in new product development. Although human-based mechanisms,

    such as the direct transfer of project members, more effectively convey

    knowledge required in design integration at the higher level of product systems,

    the article shows that standardized mechanisms are more appropriate for

    retention of lower-component knowledge. However, it is also argued that

    product architecture affects these relationships through its impact on a locus of 

    design change in a product system. Such argument is partially tested in the

    context of the Japanese automobile industry.

    Introduction

    Ever since effective new product development became crucial for competi-

    tiveness in many industries, numerous researchers have explored factors

    affecting new product development performance. Many of these studies focus

    on development processes within a single development project at a specific

    point in time. However, the transfer of technologies and other product-

    related knowledge across different generations of projects has received littleattention. A new product, by definition, should involve ‘‘newness’’—be

    different from past products. Thus, many researchers have addressed the issue

    of how to break automatic knowledge flows from preceding projects, and

    knowledge retention tends to be regarded as an obstacle to innovation or

    *Institute of Innovation Research, Hitotsubashi University, Tokyo, Japan.

    Industrial  Relations, Vol. 41, No. 4 (October 2002).    2002 Regents of the University of CaliforniaPublished by Blackwell Publishing, Inc. 350 Main Street, Malden, MA 02148, USA, and 108 Cowley

    Road, Oxford, OX4 1JF, UK.

    605

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    2/24

    something to be avoided (Allen and Marquis 1964; Leonard-Barton 1992;

    Henderson and Clark 1990; Anderson and Tushman 1990; Dougherty 1992).

    However, in most cases, new products are not ‘‘completely’’ new, neither

    in terms of technology nor market concepts. A technology developed forone product subsequently may be used in a range of products (Cusumano

    1991; Meyer and Utterback 1993; Uzumeri and Sanderson 1995; Cusumano

    and Nobeoka 1998). Additionally, knowledge about existing customers can

    be a useful basis for interpreting current customer needs (Christensen and

    Rosenbloom 1995). Therefore, successful new product development at least

    partially depends on the ability to understand technical and market

    knowledge embodied in existing products, and the adaptation of this

    knowledge to support new product development (Iansiti 1997; Iansiti and

    Clark 1994).Drawing on examples from the Japanese automobile industry, this study

    explores effective ways of retaining knowledge across different generations

    of product development. In the subsequent sections, I discuss two broadly

    defined mechanisms for knowledge retention—human-based and standard-

    ized mechanisms—by illustrating some practices of Japanese automobile

    producers. I then identify two factors that affect the relative effectiveness of 

    these two mechanisms. Next, I examine my argument empirically by using

    data obtained from 223 engineers engaged in 25 new automobile develop-

    ment projects at seven Japanese automobile producers.

    Knowledge Retention Mechanisms

    The existing literature proposes several mechanisms for retaining

    knowledge (Walsh and Ungson 1991; Krippendorf 1975; Huber 1990,

    1991; Garud and Nayyar 1994), but for the sake of simplicity, this article

    will focus on just two of these mechanisms. The first is a human-based

    mechanism. Knowledge obtained through past development activities may

    be partially stored in individuals. Therefore, bringing persons who haveappropriate experience in past development into current projects is one way

    to transfer knowledge across generations (Roberts 1979). In an extreme

    case, the same group of people successively takes charge of multiple

    generations of product development. There are, however, obvious limita-

    tions in completely depending on a particular individual or group: when

    people leave, knowledge disappears. To overcome this problem, companies

    may partially overlap project members across different generations and

    create a chain of overlapped common experience among their people. For

    example, Nakayama (1997) investigated the history of Japanese aircraft

    606 / Yaichi Aoshima

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    3/24

    development from the Zero fighter in the 1930s to the FSX support fighter

    in the 1990s and found that ‘‘development technology’’ has long been

    retained either through continuity of key persons or by overlapping project

    members across different generations of projects. In particular, he foundthat overlapped membership in the medium-scale intermediate projects,

    which bridge the gap between large-scale next-generation projects, played

    an important role in retaining development technology.

    This study of the Japanese automobile industry reveals the same kind

    of overlapped membership across generations of projects. For example,

    Figure 1 shows the project manager groups on the Celica/Carina/Corona

    projects in Toyota.

    The figure shows that most project managers for these projects were

    promoted from their former positions as subproject managers within thesame series development. At Toyota, project mangers and their staff 

    formerly worked in the Product Planning Office.1 Typically, engineers with

    seven to 10 years’ experience in body design or chassis design move to the

    Product Planning Office at around age 30. They then become responsible for

    particular vehicle development projects, working as staff members of project

    managers. New entrants to the Product Planning Office start from the Shu-

    tantoin (a lower-ranked subproject manager). They are then promoted to

    the Shusa-tsuki (a subproject manager) and Shusa (a project manager),

    most often within the same product line.2 They are trained as candidates for

    future project managers and learn how to coordinate and integrate theentire product development process. Once they become Shusa (now Chief 

    Engineer (CE)), they tend to be responsible for two successive product

    generations. As a reason for this continuity of project managers,

    Mr. Nakagawa, CE for the fifth generation of the Celica project, made

    the following comment:

    Even if the present Celica uses a different technological concept from the past, the

    characteristics of users have some commonalty. To understand the characteristics

    of users needs long experience [therefore, a project manager tends to stay in the

    same project for a long time]. For example, although the current Celica sharesbasically the same 2.2 liter engine, 5SAF, with the Camry, we did not use the

    balance shaft for the Celica engine, because I knew that Celica users require more

    power at the cost of noise. Also, when we decided to carry over a part of the under-

    floor panels from the previous model, I learned from the past project several issues

    1Toyota reorganized its product development organization in 1992 and divided it into three vehicle

    centers and one advanced engineering center. The Product Planning Office was disbanded and absorbed

    into each of three vehicle centers.2In the late 1980s, Toyota changed these titles. Since then, what was formerly a Shusa is called CE

    (Chief Engineer); the former Shusa-Tsuki is called Shusa.

    Transfer of System Knowledge Across Generations   /   607

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    4/24

    such as what kinds of problems occurred in previous projects both in technical and

    user characteristics.3

    This comment seems to indicate that continuity of project members may

    be required to deepen the understanding of linkages between user needs andrequired design features.

    Another mechanism for knowledge retention is a more standardized one.

    Project members may document knowledge obtained from past development

    activities in written or electronic form. A drawing or blueprint used in design

    activities can contain important knowledge. Companies also can make use of 

    various others kinds of documents and reports for retaining past experiences.

    Such documents and reports vary in terms of their formality. For example,

    FIGURE 1

    Project Management Groups of the Celica/Carina/Corona Projects

    3Interview with Mr. Nakagawa, Chief Engineer, Toyota Motor Corporation, May 31, 1994.

    608 / Yaichi Aoshima

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    5/24

    documents and reports used at Japanese automobile producers can be

    broadly classified into the following categories (although boundaries

    between these categories differ somewhat across companies):

    1. Companywide technical standards (e.g., Toyota Technical Standard).

    2. Department-level design standards, standardized design (test) proce-

    dures, and design (testing) tools (e.g., Layout Check List at Honda).

    3. Formal reports and documents storing nonstandardized knowledge,

    such as process and design know-how documents, testing reports, and

    user-claim reports (e.g., Know-How Document at Mitsubishi).

    4. Informal documents and memos to keep track of past problematic and

    successful cases (e.g., Problem-Information Memo at Mitsubishi).

    The first two documents describe standards and rules that engineers mustfollow. The last two documents are more flexible and describe design

    know-how, testing results, and other problematic and successful cases found

    in previous development activities. Japanese automobile producers use such

    documentation extensively as a means to store knowledge about past

    practices. Although engineers are not obliged to record information, it

    seems to be a common practice for them to write down information

    obtained from their development activities.

    In addition to documents, an advanced computer-aided design (CAD)

    system helps project members store and retain knowledge. CAD systems

    serve not only as tools to help design and test, but also as facilities for

    retaining prior knowledge obtained through design and testing experiences.

    Data stored in the CAD system can be directly reused in future design work.

    For example, since an automobile body consists of several separate body

    panels (e.g., front, center, and rear floor panels), each of which has a

    particular drawing stored in CAD, designers often reuse and edit designs

    from existing body panels and recombine them with newly designed panels

    to develop the entire body design. Similarly, computer-aided engineering

    (CAD) models reflect knowledge obtained from the prior prototype testing.

    Whether companies develop their own CAE models or use commercialpackages, it is critical to incorporate their experience of actual prototype

    testing into CAE models to make it usable.

    Factors Affecting Effectiveness of Transfer Mechanisms

    When should companies depend on a human-based mechanism for

    retaining knowledge in product development? When should they use a more

    Transfer of System Knowledge Across Generations   /   609

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    6/24

    standardized mechanism? What determines the relative effectiveness of these

    two methods? The following sections discuss two factors that have the

    potential to affect effectiveness: the nature of the knowledge and the

    frequency of design changes needed for a particular product architecture.

    Nature of Knowledge Retained.   First, some knowledge is inherently more

    suitable for a human-based knowledge retention mechanism than for an

    archival-based one, and vice versa. The inherent nature of knowledge

    directly influences the effectiveness of retention mechanisms. This article

    pays particular attention to the systemic nature of knowledge and the

    associated context-specificity (Zander and Kogut 1995; Garud and Nayyar

    1994; Badaracco 1991; Iansiti 1998).

    As many researchers have pointed out, the design of new ‘‘systems’’products (i.e., products that contain numerous components and technol-

    ogies that must work together) invariably depends on a complex interaction

    among potentially fragmented individual knowledge bases (Clark and

    Fujimoto 1991; Henderson and Clark 1990; Iansiti 1997; von Hippel 1994).

    On the one hand, knowledge required for new product development comes

    from various domain-specific and specialized disciplinary areas. Such

    knowledge tends to be generalizable and independent of a specific context.

    On the other hand, new product development demands the integration of 

    specialized knowledge with other information and its application to specific

    contexts. This latter type of knowledge will be called ‘‘system knowledge.’’4

    Architectural knowledge, defined by Henderson and Clark (1990) as

    knowledge that is about the interactions between physically distinctive

    components, is one special case of system knowledge.

    Development of a printer, for example, calls for an understanding

    of precision machinery engineering, mechanical engineering, optical elec-

    tronics, material sciences, imaging science, and so on. However, a mere

    collection of such specialized knowledge does not necessarily lead to the

    development of a good commercial printer. For a printer to work well as a

    coherent system, engineers must understand complex interactions and subtlebalances among such specialized technologies and knowledge, which must

    be managed so as to adapt the end result to a particular use environment.

    The system/domain-specific dimension of knowledge is closely connected

    to the tacit/explicit dimension, which many management studies have

    discussed (Nonaka 1994; Zander and Kogut 1995; Garud and Nayyar

    1994). System knowledge tends to be tacit. It is difficult, if not impossible, to

    articulate system knowledge because of its embeddedness in a specific

    4Iansiti (1997) uses the same conceptualization of system knowledge.

    610 / Yaichi Aoshima

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    7/24

    context. This embeddedness is one element of ‘‘tacit knowing,’’ a concept

    introduced by Polanyi (1966). According to him, we may say that we know

    more than we can tell when we comprehend the entity (a whole) without

    being able to specify its particulars (parts). He describes ‘‘achieving anintegration of particulars to a coherent entity to which we are attending.

    Since we were not attending to the particulars in themselves, we could not

    identify them (p. 18).’’ For example, we can identify physiognomy without

    specifying its features.

    The same concept can be illustrated by the development of products

    consisting of various components and technologies. Throughout product

    development processes, project managers try to find an appropriate way to

    integrate components, technologies, and other elements so that all will

    function together as a coherent product system that will be valuable totargeted customers. The appropriateness of this integration is understand-

    able only in light of the context of the final product, which is mostly

    provided by users of the product. The relationship between this context and

    the product’s elements tends to remain tacit.

    For example, in the case of automobile development, vehicle styling

    solely for the purpose of maximizing aerodynamic performance can be

    theoretically determined and generalizable. However, linkages between the

    styling, the body structure, the engine shape, and the suspension type will

    be different among vehicles, which will be of different sizes, for different

    purposes, and have different customer bases. Without understanding sur-rounding contexts of the product, one cannot know how and why the

    components are integrated in a particular way. In other words, we cannot

    communicate what we know about the integration of components without

    sharing the contexts.

    Regarding the context-specific nature of system knowledge, von Hippel

    and Tyre (1995) found that problems in the introduction of new process

    technology were often discovered through learning by doing by field

    engineers after process introduction. Since such problems are invariably

    caused by subtle and specific interactions among particular attributes of machine design and user environments, they are difficult to predict. Since

    field problems in production machines depend on context-specific factors,

    scientific investigation to discover universal cause-effect relationships may

    not be very effective.

    For context specific-knowledge to be usable in other settings, people

    need to retain and transfer information about numerous surrounding

    contingency factors behind easily observable facts and results. Therefore, it

    can be quite costly, if not impossible, to articulate context-specific system

    knowledge.

    Transfer of System Knowledge Across Generations   /   611

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    8/24

    The most direct way to transfer and retain less-articulable and context-

    specific system knowledge may be to transfer or retain individuals who have

    first-hand experience with the context. In some cases, system knowledge

    may be held by a group of people who share the context. In other cases,system knowledge might be embedded in specific relationships between

    people. This latter situation might make it necessary to retain the same

    group of people for a long period of time. Wilson and Hlavacek (1984)

    found that firms that benefited from technologies created in past projects

    kept knowledge alive by the active presence of a core group of people.

    Accordingly, I hypothesize that the more knowledge is systemic, the more

    effective a human-based retention mechanism is, such as the direct transfer

    of individuals or a group of people with shared experience. On the other

    hand, the more knowledge is domain-specific and specialized, the best wayof retaining it is through the use of archival mechanisms, such as

    documentation, standardization, and computerized systems.

    Frequency of Design Change.   The second factor affecting the relative

    effectiveness of retention mechanisms is the frequency of design changes.

    Retaining knowledge through standardized mechanisms often takes more

    time than through human-based mechanisms because the former includes

    articulation processes, abstraction from experiences, and knowledge

    transformation between different media. If market conditions change

    rapidly and fast product development cycles are critical to competition,retaining and quickly utilizing knowledge across generations of projects

    may become particularly important. If companies devote too much time to

    articulating past experience as, for example, propositions and causal

    relationships, the experience may become outdated before it can even be

    used.

    As noted above, articulating knowledge is also costly. Design standard-

    ization efforts, for example, often start by collecting the direct experiences

    of first-ranked engineers. To promote standardization, companies may have

    to ask such engineers to temporarily leave their current tasks, incurringhuge opportunity cost. If articulated and systematized knowledge is

    applicable for a relatively long period of time, such costs can be justified.

    If projects can use standardized designs for many generations, standard-

    ization becomes an efficient way to retain knowledge. However, when

    product change takes place frequently due to fast-changing environments,

    the cost of maintaining and renewing articulated knowledge increases

    considerably. Thus, the more frequent changes a product design receives,

    the more efficient human-based mechanisms of knowledge retention become

    compared to standardized mechanisms.

    612 / Yaichi Aoshima

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    9/24

    System knowledge can be retained through standardized mechanisms if 

    projects have very few design changes at the system level. However, project

    development may have to depend on human-based mechanisms of 

    knowledge retention if there is considerable and rapid design fluidity.The relative frequency of change between the system level and the

    component or subsystem level is influenced by the design strategy of 

    product architecture, whose role is discussed below.

    Product Architecture and Locus of Change.   Product architecture is one

    characteristic of a product system that involves interdependency between

    the constituent components. This interdependency is, in turn, determined by

    the pattern of allocating product functions to constituent components and

    the interfaces between these constituents (Ulrich 1995; Ulrich and Eppinger1994). Figure 2 classifies product architectures along these two dimensions.

    FIGURE 2

    Types of Product Architecture

    Transfer of System Knowledge Across Generations   /   613

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    10/24

    Under a modular architecture, located in the upper-right cell, a product

    system is divided into several physical chunks called modules, each of which

    has a simple correspondence to a particular product function (Ulrich 1995).

    Simple rule-based relationships among modules lower interdependencybetween constituents and maintain high design flexibility at the module

    level. For example, a stereo system consists of a CD player as an input

    device, an amplifier as an audio signal processor, and a speaker as an output

    device. Each device or module has its own exclusive function and there are

    standardized interfaces between them. As a result, users can improve

    performance of a stereo system by independently upgrading each module

    (Langlois and Robertson 1992).

    The PC is also a relatively modularized product. Each product function is

    mostly exclusively allocated: a short-term memory function is allocated toRAM; a long-term memory to a hard disk; a data-processing function to the

    CPU; an input function to a keyboard; an output function to a monitor,

    and so on. As long as designers keep interface rules, they can freely change

    module designs without considering other module designs (Clark and

    Baldwin 1997). Most of the recent dramatic performance improvement in

    the PC comes from technological advancement at the component level, such

    as the density improvement of the DRAM, the CPU, and the hard disk. In

    other words, a design strategy of modular architecture is a decision to shift

    sources of product improvement and design changes into lower levels of 

    product systems, such as modules and components. Changes of environ-ments are absorbed independently at the module level. Thus, a product with

    modular architecture tends to show more design changes at the component

    level than at the system level.

    The lower-left cell of Figure 2 illustrates integrated architecture, in which

    a product’s functions and physical constituents have a complex relationship,

    with no simple rules specifying interfaces of constituents (Ulrich 1995). An

    automobile has a relatively integrated architecture (Fujimoto 1998). For

    example, the upper body of an automobile fulfills various functions, such as

    shielding from bad weather, reducing air resistance, maintaining bodystrength, providing insulation from vibration, and so on. However, the

    function of reducing noise and vibration is also spread over various other

    components, such as the drive-shaft, the engine, the chassis, and so on. In

    this way, the relationship between product functions and components, or

    subsystems, in automobiles is complex and intertwined. This accounts for a

    relatively high interdependence among automobile components. For

    example, a change in suspension design will influence the designs for the

    braking system, chassis, and body. Because of such high interdependence,

    component designs are subject to more restrictions compared to modular

    614 / Yaichi Aoshima

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    11/24

    architecture designs. Since the interfaces between components are not

    specified by explicit rules, there are opportunities for improving the

    relationships between components. The entire product performance may be

    continuously improved by finding better linkages among components.Thus, I conjecture that, compared to products with modular architecture,

    products with integrated architecture show more frequent design changes at

    the system level and less frequent design changes at the component level. As

    an illustration, the average product life cycle of Japanese automobiles is

    four years, that of engines and suspensions, which have a more integrated

    architecture, often exceeds eight years.

    Figure 3 summarizes the above discussions.

    The more systemic and context-specific knowledge is, the more effective

    are human-based retention mechanisms. Since system knowledge integratesvarious technological elements, such knowledge may be more critical in

    design at the higher levels of product systems: more useful in subsystem

    designs than in detailed parts designs, more useful in entire product designs

    than in subsystem designs.

    Although I conjectured that human-based mechanisms are more suitable

    for system knowledge in the integration of design activities at higher levels

    of product systems, their suitability also depends upon the locus of design

    changes determined by product architecture. While a modular architecture

    facilitates more frequent design changes at the lower level of product

    systems, integrated architecture is expected to show a relatively highfrequency of design changes at the higher level. Thus, the question of 

    whether human-based mechanisms are suitable for retaining knowledge at

    the higher levels of product systems or at the lower levels rests on both the

    FIGURE 3

    Determinants of Relative Eectiveness of Knowledge Retention Mechanisms

    Transfer of System Knowledge Across Generations   /   615

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    12/24

    inherent systemic nature of knowledge and product architecture as a design

    strategy.

    Empirical Observations from Japanese Automobile DevelopmentProjects

    This section conducts a partial examination of the above discussion; it is

    ‘‘partial’’ because it deals with only with one type of product, an automobile.

    Assuming that an automobile has a relatively integrated architecture, I

    conjecture that knowledge for integration at the higher levels of a product

    system will tend to be retained through a human-based mechanism, whereas

    retention of knowledge in designing constituents at the lower levels of theproduct system will tend to rely on standardized mechanisms.

    Continuity of Core Project Members.  Automobile development projects

    involve numerous people with different engineering roles and functional

    backgrounds. Different people invariably require different types of knowl-

    edge to accomplish their tasks. For example, for project members primarily

    responsible for integration and coordination of dispersed functional

    activities, system knowledge should be more critical. On the other hand,

    specialized and functional knowledge should be more important for people

    whose primary task is to design particular component systems, such assuspension systems, braking systems, under-floor panels, and engine

    components.

    The discussion thus far has implied that transfer of system knowledge will

    be associated with direct personnel transfers. I expect to observe that project

    members primarily responsible for integration among different functional

    and technical areas are more likely to continue in their positions for

    successive generations of projects than are those responsible for functional

    activities. In the case of automobile development, the following three types

    of project members are defined as integrators: project managers, vehiclelayout engineers, and vehicle test engineers.

    The role of project managers as system integrators is discussed in other

    studies (Clark and Fujimoto 1991; Imai et al. 1985). Although project

    managers’ responsibilities vary from project to project, my research on

    Japanese automobile companies revealed that all project managers were at

    least responsible for concept making, the basic vehicle plan, and coordi-

    nation of development processes.

    Vehicle layout design is an activity that involves complex adjustments

    between conflicting engineering areas. The vehicle layout, basically, is the

    616 / Yaichi Aoshima

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    13/24

    design for how the mechanical components, body structures, luggage, and

    passengers will be distributed throughout the vehicle. It involves key

    decision making for platform designs, basic component configuration, and

    major physical dimensions (e.g., wheel base, tread, hip point). The vehiclelayout determines the basic architecture of an automobile, which expresses

    total vehicle concepts in physical terms (Clark and Fujimoto 1991).

    Accordingly, vehicle layout significantly influences interfaces among differ-

    ent component systems, for example, between suspension systems and

    under-floor panels, between engines and engine mount. Therefore, layout

    engineers thus require broad-based experience in different engineering areas.

    Vehicle test engineers, as opposed to component test engineers, also

    engage in extensive coordination activities cutting across different compo-

    nent development areas. Since the overall performance of automobiles, asindicated by, for example, NVH (Noise-Vibration-Harshness), safety, body

    strength, and driving stability, is derived from many complex interactions

    among individual components, a vehicle test engineer has to interact with

    related component engineers to achieve targeted performance levels.

    This leads to the following hypothesis:

    Hypothesis 1:   The probability that project managers, vehicle layout engineers,

    and vehicle test engineers are transferred from the previous generation of the

     project is higher than that for the other project members.

    Methods.  To test this hypothesis, we obtained data on project members’

    experiences in the previous generation of projects by a questionnaire survey.

    The sample contains 223 project members representing 25 new product

    development projects at seven Japanese automobile producers. The survey

    was conducted between 1987 and 1995. A questionnaire was distributed to

    the following 10 project members for each project, who represented their

    own activity areas for that project. These members will hereafter be referred

    to as ‘‘project core members.’’

    1. Project manager.

    2. Representative of vehicle layout designers/planners.

    3. Representative of vehicle test engineers.

    4. Representative of chassis design engineers.

    5. Representative of body design engineers.

    6. Representative of exterior/interior designers.

    7. Representative of engine design engineers.

    8. Representative of electronics component design engineers.

    Transfer of System Knowledge Across Generations   /   617

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    14/24

    9. Representative of production engineers.

    10. Representative of marketing/product planners.

    All 10 responses were obtained from 17 projects, but there is some

    missing data for the remaining eight projects. As a result, the sample

    comprises 229 core members. Out of these, 223 responses included usable

    answers with respect to experience in the previous projects.

    I identified whether respondents had experience in the previous project by

    asking them to list the names of past new product development projects on

    which they had spent more than an average of 30 percent of their time for at

    least six months. When the name of the direct predecessor model was on this

    list, that respondent was treated as a project member in the previous

    generation of projects. Respondents were not counted, even if they had

    participated in the previous model development, unless they had devotedmore than 30 percent of their time for at least six months.5

    Table 1 shows the number and the percentage of project core members

    transferred from the previous project by each category of person.

    Out of the 223 respondents, 83 project core members had experience in

    the previous generation of projects. Out of 10 different types of core

    members, layout engineers were more likely to be transferred from previous

    projects (15 out of 20 layout engineers), followed by project managers (12

    out of 24),6 and vehicle test engineers (11 out of 22). Being a layout

    engineer, vehicle test engineer, or a project manager, as opposed to the otherfunctional engineers, makes a significant difference as to whether one will be

    transferred from the previous generation of projects. This appears to

    support my hypothesis that integrators are more likely to be in charge of the

    same model development in successive generations.

    To examine the validity of the above finding more precisely, a logistic

    regression analysis was conducted. Three explanatory variables were

    considered as predictors for whether project core members had experience

    in previous project generations:

    5This way of defining project members needs to be treated with care. Since some components, such as

    electronics components, are often designed for multiple projects, engineers might not spend more than 30

    percent of their time even on their principal project. We therefore considered these people as not having

    worked for the preceding project unless they explicitly indicated otherwise in the questionnaire.6Some may wonder that only half the project managers are retained from previous projects despite

    the importance of accumulated project management experience. One reason may be that a project

    manager tends to be promoted to a higher position, especially when his project turns out to be successful.

    Also, companies that do not have enough competent project managers rotate them to other key product

    development projects. Additionally, if a product needs a significant change in terms of a market concept,

    retaining system knowledge may have a negative impact on project performance. Aoshima (1997) reports

    this effect.

    618 / Yaichi Aoshima

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    15/24

    •   System Integrator: A dummy variable indicating system integrators

    (1 if respondents are project managers, vehicle layout engineers, or

    vehicle test engineers; 0 otherwise).

    •  Task Newness: Percent of design change from the previous model.Available Project: The number of available projects at the time of the

    start of the focal project.

    Ten project core members were divided into two categories: System

    Integrators or functional engineers. As already defined, System Integrators

    include project managers, vehicle test engineers, and vehicle layout

    engineers. System Integrator was set equal to 1 if respondents were system

    integrators, and 0 otherwise.

    In addition to system integrator, two control variables were included. The

    first control variable is a percentage of new design, Task Newness, whichindicates the degree of change in engineering tasks from the previous

    project. The previous project members might be assigned to the current

    project because the required engineering work is similar between two

    successive projects. For example, it could be expected that project members

    will continue in successive generations if the current project activity involves

    only minor modification of the previous design.

    Project managers assessed the degree of novelty of the design for their

    new products. Answers encompassed 10 different component systems,

    including exterior/upper body, interior/trim, steering, floor panels, braking

    TABLE 1

    Transfer of Project Members from Previous Generation of Project

    Project members in the previous

    generation of project?

    Yes No Total

    Project managers** 12 (50.0%) 12 (50.0%) 24

    Layout engineers**** 15 (75.0%) 5 (25.0%) 20

    Vehicle test engineers** 11 (50.0%) 11 (50.0%) 22

    Exterior/interior designers 5 (20.8%) 19 (79.2%) 24

    Chassis engineers 4 (17.4%) 19 (82.6%) 23

    Body engineers 10 (43.5%) 13 (56.5%) 23

    Engine design engineers 9 (37.5%) 15 (62.5%) 24

    Electronics component engineers 4 (19.0%) 17 (81.0%) 21

    Production engineers 9 (47.4%) 10 (52.6%) 19Marketing/product planners 4 (17.4%) 19 (82.6%) 23

    Total 83 (37.2%) 140 (62.8%) 223

    Asterisks indicate results of the chi-square tests of independence between being each of three integrators, as opposed tothe other functional/component engineers, and being the previous project members (** p  <  0:05, *** p  

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    16/24

    systems, suspension systems, transmissions, engine, engine control, and

    instrument panels. Newness of each design was measured on a four-point

    scale: 1  ¼  minor modification of less than 20 percent of the existing design,

    2 ¼  major modification involving 20–80 percent of the new design, 3 ¼  newdesign involving more than 80 percent of the new design, and 4 ¼  entirelynew technology not applied to any model before. This scale was conceived

    to capture the theoretical distinction between component system designs

    that are purely ‘‘carry-over,’’ involved substantial redesign, or were entirely

    new designs.

    The novelty involved in activities of component engineers was calculated

    by averaging newness of appropriate component designs in terms of the

    number of new parts involved. We measured newness of System Integrator

    activities by the newness of the platform design as indicated by newness of suspension systems and under-floor panels (Nobeoka 1993). The extent of 

    novelty in the design activities of production engineers was measured by the

    percentage of new production equipment. Marketing/product planners were

    excluded from the analysis since indicators could not be obtained for their

    Task Newness that were comparable to those for the other core members.

    As a result, values of Task Newness were assigned to 192 project members.

    The second control variable, Available Project, is the number of available

    projects at the time of the current project’s start. It is expected that the greater

    the number of simultaneous ongoing projects in a company, the lower the

    probability of project members being transferred within the same model line.Tables 2 to 4 summarize descriptive statistics of the dependent variable,

    the explanatory variables, and two control variables.

    Results.  Table 5 shows the results of a fitted logistic regression analysis.

    Model II in Table 5 involves only control variables. It shows that the

    number of available projects has a negative effect on the transfer of core

    project members, as expected. However, Task Newness does not have a

    significant relationship with transfer of project members, although it does

    have the expected sign for the coefficient.Model III, in the third row of Table 5, involves System Integrator as a

    predictor. The parameter estimate of System Integrator is 1.10, which

    TABLE 2

    Descriptive Statistics for Experience with Previous Project Generation

    Frequency

    Cumulative

    frequency

    Cumulative

    percent

    0  ¼  not members of previous generation of project 107 107 58.5%1  ¼  members of the previous generation of projects 76 183 100.0%

    620 / Yaichi Aoshima

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    17/24

    indicates a positive relationship between System Integrator and previous

    experience, as expected. The chi-square goodness of fit statistic was reduced

    for 10.98 from Model II  ð p  

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    18/24

    more retention of such knowledge will benefit from the use of standardized

    mechanisms, such as documentation, standardization, and computerized

    systems. This implies that component engineers may use documents,

    reports, design standards, and computerized media to retain past knowledgemore frequently than do integrators, leading to the following hypothesis.

    Hypothesis 2:  Component engineers refer to documents and reports, and use

    design standards and standard design procedures, more frequently than do

    integrators.

    To partially examine this, engineers were asked to rate, on a five-point

    Likert scale (from 1 ¼   not refer at all   to 5 ¼   refer very frequently), thefrequency with which they referred to documents and reports that described

    design know-how and design solutions and problems identified in the pastdevelopment activities. Similarly, respondents rated how important were the

    roles design standards and standard design procedures played in their

    project activities, on a five-point Likert scale, from 1 ¼  not important at all to 5 ¼  played a very important role.

    Based on these ratings, a series of OLS regression analyses was

    conducted. Dummy variables were introduced into the analyses, each of 

    which indicates a particular type of engineer. The dummies were effect

    coded so that the coefficients are deviations from the overall mean. In

    addition, two control variables were taken into consideration: tenure of 

    project members and task newness. Tenure indicates the length of thesubject’s service in the current company. We used the same measure for task

    newness as used in the previous section.

    Next, individual dummies were substituted for a single dummy indicating

    integrators versus component engineers to explore the effect of being

    integrator. Integrators include layout engineers and test engineers.7 Table 6

    shows summary statistics of some of these variables.

    Table 7 shows the results. This table shows that layout engineers tend to

    rate significantly lower the frequency of reference to documents and reports

    in their development activities. Engine and body engineers reported morefrequent references to documents and reports for the purpose of learning

    from the past design activities.

    As for importance of design standards and standard design procedures,

    layout engineers also rated these significantly lower. Combined with the

    above result, this seems to suggest that knowledge regarding vehicle layout

    design is less likely to be transferred through documentation and

    7Project managers were excluded from this analysis because there are no standards available for

    project managers that are equivalent to design standards or test codes.

    622 / Yaichi Aoshima

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    19/24

    standardization. Considering the findings of the previous sections that

    layout engineers tend to continue their positions in successive generations of 

    projects, it appears that knowledge to design the total vehicle layout may be

    less codifiable and its retention may tend to depend on personal experience.For example, one engineer responsible for vehicle layout design in the four

    successive Familia (323) projects at Mazda, explained why it is important

    for layout engineers to continue their positions in successive generations:

    Since ‘‘Kikaku-sekkei’’ [basic design, which means the layout design in Mazda] is

    related to the all engineering areas, it’s important to gain broad experiences [within

    the same product line] to be able to capture a whole [complex relationships between

    different engineering domains].  . . . Thus layout engineers tend to stay in the same

    product line for long time.8

    However, contrary to the second hypothesis, Table 7 shows that both

    documentation and standardization seem to be important means of 

    capturing prior practices in vehicle test engineering. In particular, test

    engineers rated significantly higher the importance of standards (in their

    case, standards means test codes).

    This may be partially because documents and reports have particular

    meaning or usefulness to test engineers. While drawings are the primary

    outputs for design engineers, testing reports are the primary outputs for test

    engineers. Test engineers summarize test results in various forms of reports,

    which they distribute to related engineering and design departments.

    Knowledge related to testing activities, such as that about target perfor-mance setting, design, and production prototype evaluation, are thought to

    be transferred through documentation. For example, Cusumano and Selby

    (1995) report that software testers at Microsoft rely heavily on various types

    of checklists and scripts.

    The result also suggests that existing testing methods and established test

    codes become a critical basis for testing new vehicles. Compared to design

    work, testing work may require more consistency and thoroughness than

    TABLE 6

    Descriptive Statistics of Variables

    N  ¼   Mean SD Min. Max.

    Frequency of reference to documents and reports 167 3.78 0.95 1.00 5.00

    Importance of standards (design standards and

    test codes)

    165 4.06 0.94 1.00 5.00

    Tenure 167 17.82 7.38 4.00 42.00

    Task newness 167 0.55 0.28 0.05 1.00

    8Interview with Mr. Morioka, Mazda Motor Corp., May 19, 1994.

    Transfer of System Knowledge Across Generations   /   623

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    20/24

    creativity. Thus, standards and documents may become important.

    Although it was assumed that vehicle test engineering involves complex

    integrative efforts across many different component systems development,

    the testing function itself may be a well-established discipline.

    Because of test engineers’ reliance on standardized mechanisms, Columns 2and 4 in Table 7 show no significant difference between system integrators and

    component engineers in terms of use of documents and standards. The second

    hypothesis is not supported. Although the findings for layout engineers are

    consistent to with the hypothesis, from the above analysis, this cannot be

    attributed to the fact that they are system integrators. The tasks of test

    engineers may have distinctive characteristics that call for the extensive use of 

    documents and standards, which might obscure the effect of being system

    integrators. However, this cannot be confirmed from the available data.

    Summary

    The main objective of this article was to explore effective ways of 

    retaining knowledge involved in new product development. I discussed the

    idea that retention of knowledge about the interactions among fragmented

    functional domains may demand direct transfer of individual experience

    bases, but that standardized mechanisms may be more beneficial for

    retaining specialized and functional knowledge. This further implied that

    TABLE 7

    Regression Analyses for the Use of Documents and the Importance of

    Standards

    Frequency of reference

    to documents and reports

    Importance of standards

    (design standards and test code)

    Tenure   0:01 0.04   0:07   0:00Task newness   0:11   0:12   0:11   0:16**Chassis engineer   0:05   0:04

    Body engineer 0.16* 0.08

    Engine engineer 0.19 0.04

    Electronics comp. engineer 0.04 0.05

    Manufacturing engineer 0.11 0.19*

    Layout engineer   0:23**   0:27***Test engineer 0.11 0.17*

    System integrators   0:08   0:17df of residuals 157 163 155 161

    Adjusted R2 0.05* 0.02 0.07* 0.01

    * p  <  0:1.** p  <  0:05.*** p  

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    21/24

    although human-based mechanisms more effectively convey knowledge

    required in design integration at the higher levels of product systems,

    standardized mechanisms are appropriate for retention of lower-component

    knowledge. However, it was also argued that product architecture affectsthese relationships through its impact on a locus of design change in a

    product system. Modular architecture, for example, focuses design change

    at the module and component levels. The resulting high frequency of design

    changes at module and component levels may make human-based mech-

    anisms of retaining component knowledge more suitable than archival-

    based mechanisms despite its less systemic nature.

    Two empirical analyses of data obtained from Japanese automobile

    development projects were conducted to test these ideas. In these analyses,

    first, it was found that project members responsible for integration activitiescutting across different functional domains tend to continue in their positions

    for successive generations of projects. This implies that knowledge for

    integration may tend to be associated with the transfer of people. On the other

    hand, component engineers were less likely to be responsible for successive

    generations of the same product. Second, it was found that layout engineers

    tend to rely less on documents, reports, and standards to learn from past

    design practices than do component engineers. These results may imply that

    standardized mechanisms alone are not enough to retain knowledge

    regarding interaction and integration among different design domains.

    Limitations and Further Research

    Limitations of this study are mostly due to considering only one industry

    in one country. Because of this, the empirical part of this article has weak

    connections to the theoretical discussions.

    First, this study dealt with the automobile that, it is assumed, has integrated

    product architecture, I did not examine knowledge retention practices in the

    development of modularized products. There could be quite different results

    in development projects of modularized products. For example, in the case of 

    PC development, there may be more dependence on human-based retention

    mechanisms in the development of components such as hard disks, monitors,

    CPUs and other semiconductor devices, and less on integration of those

    components into final goods. Investigation of products that have been rapidly

    modularized recently, such as the bicycle, may provide important additional

    insights into the management of knowledge retention processes.

    Focusing only on the automobile industry made it impossible to examine

    the effect of the frequency of design change on retention mechanisms. The

    Transfer of System Knowledge Across Generations   /   625

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    22/24

    theoretical discussions in this article suggest that industries facing higher

    rates of design changes tend to rely more on human-based retention

    mechanisms than on standardized mechanisms. To examine this, this study

    needs to be extended to include a variety of industries in terms of speed of change, ranging from a rapidly growing industry to a mature one.

    Second, this article focused on the benefit side of retention mechanisms.

    However, there is a cost in using them, which may differ across different

    institutional settings. For example, the cost of retaining a specific individual

    across multiple generations of projects may be very high in countries that

    have high labor mobility. Since this study includes only a Japanese sample,

    such institutional influences could not be examined.

    In our sample, only one project member out of 223 had working

    experience in other automobile companies: almost all members had workedfor only one company. Further examination is needed to determine whether

    this low mobility is common in Japan or specific to the Japanese auto

    industry. The low labor mobility in our sample probably reduces the cost of 

    retaining a project member, which may generate some biases in favor of 

    human-based retention mechanisms. On the other hand, where labor

    mobility is quite high, companies cannot help depending on standardized

    mechanisms to retain knowledge even if it is systemic. Although the cost of 

    articulating system knowledge may be high, retaining people may be even

    more costly. In that case, companies either invest in standardized retention

    mechanisms or modularize and standardize a product itself to decrease theneed for context-specific system knowledge.

    The third limitation is lack of project level analyses. If our interest resides

    in performance of new product development, knowledge retention at the

    project level needs to be considered. In particular, since system knowledge

    can be stored in multiple people, knowledge retrieval may be triggered when

    members with common experiences get together. Such a group-level

    phenomenon was not considered in this paper.9 Once data is analyzed at

    the project level, we can examine, for example, a relationship between the

    generation linkage and the cross-functional integration that is been knownto exist in Japanese companies.10

    9On project-level phenomenon, see Aoshima (1997).10By examining the same database at the project level, Aoshima (1996) examines the relationship

    between the cross-functional integration and the cross-generation linkage, and finds a high correlation

    between two. There, the cross-generation linkage is measured by the percentage of core project members

    having experiences in the previous generation of projects, and the degree of common past project

    experience among members. From this result it can be implied that the cross-generation linkage may be a

    foundation for the cross-functional integration.

    626 / Yaichi Aoshima

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    23/24

    To overcome the above limitations, future research should make

    comparisons across different industries in different countries, dealing with

    data at the project level.

    References

    Aoshima, Y. 1996. ‘‘Knowledge Transfer Across Generations: The Impact on Product Development

    Performance in the Automobile Industry.’’ Ph.D. dissertation. MIT.

     ———. 1997.  Knowledge Retention: The Impact on New Product Development Performance. Working

    Paper #97-08. Institute of Business Research, Hitotsubashi University.

    Allen, T., and D. G. Marquis. 1964. ‘‘Positive and Negative Biasing Sets: The Effects of Prior Experience

    on Research Performance.’’  IEEE Transaction on Engineering Management  11:158–62.

    Anderson, P., and M. Tushman. 1990. ‘‘Technological Discontinuities and Dominant Designs: A

    Cyclical Model of Economic Change.’’  Administrative Science Quarterly  35(4):604–633.Badaracco, J. L. 1991.   The Knowledge Link: How Firms Compete Through Strategic Alliances . Boston,

    MA: HBS Press.

    Christensen, C. M., and R. S. Rosenbloom. 1995. ‘‘Explaining the Attacker’s Advantage: Technological

    Paradigms, Organizational Dynamics, and the Value Network.’’  Research Policy  24:223–57.

    Clark, K. B., and C. Y. Baldwin. 1997. ‘‘Managing in an Age of Modularity.’’  Harvard Business Review

    September–October: 84–93.

     ——— and T. Fujimoto. 1991.  Product Development Performance: Strategy, Organization, and Man-

    agement in the World Auto Industry. Boston, MA: Harvard Business School Press.

    Cusumano, M. A. 1991.   Japan’s Software Factories: A Challenge to U.S. Management. New York:

    Oxford University Press.

     ——— and K. Nobeoka. 1998.  Thinking Beyond Lean. New York: Free Press.

     ——— and R. W. Selby. 1995.   Microsoft Secret: How the World Most Powerful Software Company

    Create Technology, Shapes Markets, and Manage People. New York: Free Press.

    Dougherty, D. 1992. ‘‘Interpretive Barriers to Successful Product Innovation in Large Firm.’’  Organi-

    zation Science  3(2):179–202.

    Fujimoto, T. 1998. ‘‘Japanese Automobile Product Development in 1990s—Capability-Building Com-

    petition by Front-Loading (in Japanese).’’  Business Review  46(1):22–45.

    Garud, R., and P. R. Nayyar. 1994. ‘‘Transformative Capacity: Continual Structuring by Intertemporal

    Technology Transfer.’’  Strategic Management Journal  15:365–85.

    Henderson, R. M., and K. B. Clark. 1990. ‘‘Architectural Innovation: The Reconfiguration of Existing

    Product Technologies and the Failure of Established Firms.’’   Administrative Science Quarterly

    35:9–30.

    Huber, G. P. 1990. ‘‘Organizational Learning: The Contributing Processes and the Literatures.’’

    Organizational Science  2(1):88–111.

    Huber, G. P. 1991. ‘‘Organizational Learning: The Contributing Processes and The Literatures.’’

    Organizational Science  2(1):88–111.

    Imai, K., I. Nonaka and H. Takeuchi. 1995. Managing the New Product Development Process: How

    Japanese Learn and Unlearn. In K. B. Clark, R. H. Hayes and C. Lorenz, (Eds.). The Uneasy

    Alliance: Managing the Productivity-Technology Dilemma. Boston, MA: Harvard University

    Press.

    Iansiti, M. 1997. Technological Integration: Making Critical Choices in a Turbulent World. Boston,

    MA: Harvard Business School Press.

    Iansiti, M. 1998.  Technology Integration. Boston, MA: HBS Press.

     ——— and K. B. Clark. 1994. ‘‘Integration and Dynamic Capability: Evidence from Product

    Development in Automobiles and Mainframe Computers.’’   Industrial and Corporate Change

    3(3):557–605.

    Transfer of System Knowledge Across Generations   /   627

  • 8/17/2019 1. Aoshima, y. Industrial Relations

    24/24

    Krippendorf, K. 1975. ‘‘Some Principle of Information Storage and Retrieval in Society.’’   General 

    Systems:15–35.

    Langlois, R. N., and P. L. Robertson. 1992. ‘‘Networks and Innovation in a Modular System: Lessons

    from the Microcomputer and Stereo Component Industries.’’  Research Policy  21:297–313.

    Leonard-Barton, D. 1992. ‘‘Core Capabilities and Core Rigidities: A Paradox in Managing New ProductDevelopment.’’  Strategic Management Journal  13(Summer):111–26.

    Meyer, M. H., and J. M. Utterback. 1993. ‘‘The Product Family and the Dynamics of Core Growth.’’

    Sloan Management Review  29(4):29–47.

    Nakayama, T. 1997. ‘‘The Keisho of Development Technology: The Case of the Japanese Aircraft

    Industry.’’  Journal of Product Innovation Management  14(5):393–405.

    Nystrom, P. C., and W. H. Starbuck. 1984. ‘‘To Avoid Organizational Crises, Unlearn.’’ Organizational 

    Dynamics  12:53–65.

    Polanyi, M. 1966.  The Tacit Dimension. Gloucester, MA: Peter Smith.

    Ulrich, K. 1995. ‘‘The Role of Product Architecture in the Manufacturing Firm.’’   Research Policy

    24:419–40.

     ——— and S. Eppinger. 1994.   Product Design and Development. New York: McGraw-Hill.

    Uzumeri, M., and S. Sanderson. 1995. ‘‘A Framework for Model and Product Family Competition.’’Research Policy  24:583–607.

    von Hippel, E. 1994. ‘‘‘‘Sticky Information’’ and the Locus of Problem Solving: Implications for

    Innovation.’’ Management Science  40(4):429–39.

     ——— and M. J. Tyre. 1995. ‘‘How Learning by Doing is Done: Problem Identification in Novel Process

    Equipment.’’ Research Policy  24:1–12.

    Walsh, J. P., and G. R. Ungson. 1991. ‘‘Organizational Memory.’’  Academy of Management Review

    16(1):57–91.

    Wilson, T. L., and J. D. Hlavacek. 1984. ‘‘Don’t Let Good Idea Sit on the Shelf.’’  Research Management

    27(3):27–34.

    Zander, U., and B. Kogut. 1995. ‘‘Knowledge and the Speed of the Transfer and Imitation of Orga-

    nizational Capabilities: An Empirical Test.’’  Organization Science  6(1):76–92.

    628 / Yaichi Aoshima