ArticlePDF Available

Adapting Real Options to New Product Development by Modeling the Second Toyota Paradox

Abstract and Figures

Uncertainty in product development projects creates significant challenges for managers who are under intense competitive pressures to increase product quality, while reducing development time and costs. Traditional wisdom dictates the early selection of a single design in order to freeze interfaces between product subsystems so that team members can work effectively in parallel, resulting in more productive product development efforts. Prior research, however, uncovered a paradoxical case. Toyota Motor Corporation achieves the fastest development times in its industry by intentionally delaying alternative selection, a strategy termed set-based development. The current work adapts real options concepts to product development management to partially explain this paradox. A formal simulation model is used to show that converging too quickly or too slowly degrades project value. Furthermore, the model demonstrates that the wisdom of set-based strategies can be explained by the application of a real options approach to product development management. Implications for managers and directions for future work are discussed.
Content may be subject to copyright.
IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, VOL. 52, NO. 2, MAY 2005 175
Adapting Real Options to New Product Development
by Modeling the Second Toyota Paradox
David N. Ford, Member, IEEE, and Durward K. Sobek, II
Abstract—Uncertainty in product development projects cre-
ates significant challenges for managers who are under intense
competitive pressures to increase product quality, while reducing
development time and costs. Traditional wisdom dictates the early
selection of a single design in order to freeze interfaces between
product subsystems so that team members can work effectively
in parallel, resulting in more productive product development
efforts. Prior research, however, uncovered a paradoxical case.
Toyota Motor Corporation achieves the fastest development times
in its industry by intentionally delaying alternative selection, a
strategy termed set-based development. The current work adapts
real options concepts to product development management to
partially explain this paradox. A formal simulation model is used
to show that converging too quickly or too slowly degrades project
value. Furthermore, the model demonstrates that the wisdom of
set-based strategies can be explained by the application of a real
options approach to product development management. Implica-
tions for managers and directions for future work are discussed.
Index Terms—Product development, project management, real
options, set-based concurrent engineering, system dynamics.
I. INTRODUCTION
PRODUCT development organizations must recognize and
capture as much value as possible to be competitive in
today’s global market. Some value is easy to recognize and
relatively predictable, such as increased productivity from
training or potentially lower costs from shorter cycle times.
Such value can be recognized and realized using traditional
management methods and tools. However, significant value
may remain hidden and, therefore, unexploited in the uncertain
portions of projects, so-called latent project values [1]. Effec-
tive development strategies for recognizing and exploiting these
latent sources of value can increase overall project value. For
example, delaying the selection of a final design can add value
if the delay allows developers to select a better alternative.
Without effective strategies, product development managers
may not recognize latent project value and, therefore, miss
opportunities to increase the value of their projects to the firm.
However, exploiting the latent project value due to uncer-
tainty in product development (PD) is difficult for at least two
reasons. First, complex development processes and managerial
Manuscript received February 1, 2004; revised July 1, 2004. Review of this
manuscript was arranged by Department Editor J. K. Liker.
D. N. Ford is with the Department of Civil Engineering, Texas A&M Univer-
sity, College Station, TX 77843-3136 USA (e-mail: DavidFord@tamu.edu).
D. K. Sobek, II is with the Department of Mechanical and Industrial Engi-
neering, Montana State University, Bozeman, MT 59717-3800 USA (e-mail:
dsobek@ie.montana.edu).
Digital Object Identifier 10.1109/TEM.2005.844466
decisions interact to affect PD behavior, project performance,
and value in delayed and nonlinear ways which obfuscate the
impacts of uncertainty on PD project performance. Second, PD
programs are subject to multiple sources of uncertainty whose
effects are not typically cumulative, so strategies designed to
manage uncertainty based on one or two conditions alone can
be seriously flawed. Numerous process models have been pro-
posed in the new PD literature [2], but how specific strategies
designed to manage uncertainty affect development behavior,
performance, and value is not well understood.
Real options represent one approach to managing assets in
the face of uncertainty. A real option is a right without an obli-
gation to take specific future actions in managing a real asset de-
pending on how uncertain conditions evolve [3], [4]. The central
premise of real options is that, if future conditions are uncertain
and changing the strategy later incurs substantial costs, then in-
vesting in flexible strategies can increase overall project value.
In short, retaining flexibility in the form of options to change
course can have value in the face of uncertainty.
This paper describes key challenges in applying traditional
real options models to the management of new PD projects,
and introduces a methodological approach for applying real op-
tions concepts to new PD management. The primary purpose
is to provide managers with tangible insight into the impacts
and value of different flexible development plans. We propose
to do this by adapting options decision rules to PD, then in-
corporating these decision rules into a system dynamics sim-
ulation model of a PD process and using the model to under-
stand project performance as a key decision parameter varies.
We focus on a single but important PD practice, set-based de-
velopment, a unique approach to managing PD uncertainty that
has been associated with sustained competitive advantage and
industry-leading profits [5]. By using a model based on real
options concepts to show how set-based development can add
value, we demonstrate that real options concepts can at least
partially explain the Second Toyota Paradox [5].
This work contributes to PD by expanding set-based devel-
opment from a binary (set-based or not) to a continuous (how
set-based) description, formalizing a model of alternative se-
lection in set-based development, and testing real options con-
cepts as an explanatory framework of Toyota’s paradoxical ap-
proach to PD. The work contributes to the real options body of
knowledge by proposing and initially testing a methodological
approach to adapting real options concepts to PD management.
The following section introduces real options and reviews tra-
ditional models for their application to PD management. The
set-based development model is then described, followed by
a case example of set-based decision-making at Toyota Motor
0018-9391/$20.00 © 2005 IEEE
Authorized licensed use limited to: MONTANA STATE UNIV BOZEMAN. Downloaded on October 20, 2009 at 15:44 from IEEE Xplore. Restrictions apply.
176 IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, VOL. 52, NO. 2, MAY 2005
Corporation. Next, Sections IV and V present the formal mod-
eling approach used, rst by describing the adaptation of real op-
tions concepts to PD, then presenting a simulation model based
on the Toyota case. The results of a series of simulation runs
provide insights about how set-based approaches impact devel-
opment project behavior, performance, and value. A discussion
of the implications of the results for PD managers is followed
by conclusions and recommendations for future research.
II. REAL OPTIONS
Real options can be described along several dimensions, in-
cluding ownership, the source of value, complexity, and de-
gree to which the option is available. A common topology sep-
arates real options according to the type of managerial action
applied [6][8]. Categories include options that postpone (hold
and phasing options), change the amount of investment (growth,
scaling, or abandonment options), or alter the form of involve-
ment (switching options). The current work focuses on an option
to postpone the elimination of design alternatives.
Real options have been used to value strategies in many do-
mains, including specic aspects of PD [3], [7], [9][12]. Ex-
amples include valuation of options to hedge technology in-
vestment risk [8], and application to design modularity [13],
research and development resource allocation [14], and max-
imum price contracts for construction project options [15]. Real
options have also been promoted as an effective tool for im-
proving the insight of strategic planners and managers [3], [4],
[16][18]. Potential benets of real options applied to PD stem
from several sources, including: a broader range of strategies
considered, a focus on objectives instead of solutions, sensitivity
to multiple project futures, more frequent testing of plans, and
increased awareness of the value of exibility [19].
Thus, it seems a real options approach can improve planning
in PD by helping managers recognize, design, and use exible
alternatives to manage dynamic uncertainty. However, there has
been little research on how real options can be operationalized to
improve PD effectiveness. PD managers may underutilize real
options to capture the value of exibility because they lack tools
and means to build insight about the impacts of those strate-
gies. By modeling different decision strategies for coping with
uncertainty, real options can potentially reveal and quantify la-
tent project value. Further, specifying the managerial signals,
decision rules, exercise criteria, and action plans operationalizes
exible strategies for implementation because it gives managers
specic levers to use rather than general admonitions. These im-
proved models of PD project valuation can add a precision and
rigor to decision-making that engineers commonly apply to de-
sign but rarely to project planning.
Conventionally, researchers have estimated the value of real
options based on approaches used to value nancial options
[20][22]. Much of the formal modeling of real options has
focused on valuing (in economic terms) options for specic
asset characteristics (value, uncertainty, discount rate) and op-
tion designs (timing, exercise conditions, and exercise costs),
but researchers have identied common modeling assumptions
that do not apply to typical PD projects [23]. Specically, most
real options models assume that: 1) future asset behavior and
value conform to well-dened processes; 2) markets are com-
plete and arbitrage opportunities are available; 3) sources of un-
certainty are few and independent; 4) payouts or other costs of
delaying decisions are small; and 5) planners have one or few
options. None of these assumptions hold well for PD environ-
ments: asset value behaviors are not well-dened or market-
based, many sources of uncertainty exist and interact, delaying
decisions can be very costly, and planners usually have, practi-
cally speaking, unlimited options.
In addition, Garvin and Cheah [24] describe problems in as-
suming risk-neutrality and using overly simple models of uncer-
tainty as it affects performance and value, as oversimplication
can lead to erroneous conclusions. Further, Alessandri et al.
[25] describe problems posed by assuming that asset perfor-
mance and option holder activity are independent, when in fact
PD option holders (i.e., project managers) purposefully and sig-
nicantly manipulate the linkages between uncertainties and
project values. Finally, the relative costs of PD versus other real
options applications also call for models that closely reect PD
practice. The development costs incurred to maintain progress
are large compared to the analogous costs in stock options (div-
idends) that are often assumed small or zero; while the exercise
cost of an option in PD (such as abandoning a design alternative)
may be small compared with the exercise costs of other options.
The current work directly addresses these issues through
a model of PD processes that behaves similarly to actual PD
projects. The model explicitly allows for uncertain asset be-
haviors, includes multiple interdependent options, and enables
the option holder to affect option value. In addition, the current
work assumes relatively large development costs and small
exercise costs. The model is motivated by a provocative case
from the literature, described in Section III.
III. SECOND TOYOTA PARADOX
A key decision that PD managers face is how to converge
from an initial set of conceptual ideas to one idea that will
become the nal design. Prior research has identied two
contrasting PD convergence strategies. The rst is an early con-
vergence strategy termed point-based concurrent engineering
and typied by the popular early-design-freeze policy [26]. In
a point-based approach, design teams initially consider a range
of alternatives from membersindividual perspectives (e.g.,
styling, body engineering, stamping, etc.), but quickly select
the best alternative to reduce project complexity and constrain
development costs. In the spirit of concurrent engineering, the
design thought to be the best alternative is critiqued from mul-
tiple perspectives, changes made accordingly, and the iteration
process continues until all constraints are satised. But, in
large PD projects with hundreds of contributors, a point-based
strategy does not necessarily resolve into a satisfactory design
fast enough. This is because any change propagates changes
elsewhere, which in turn create more changes, and so on,
resulting in what some have called design churning [27]. En-
gineers have conded to us that frequently the process never
does resolve completely; it just stops because the project has
reached a deadline.
Authorized licensed use limited to: MONTANA STATE UNIV BOZEMAN. Downloaded on October 20, 2009 at 15:44 from IEEE Xplore. Restrictions apply.
FORD AND SOBEK: ADAPTING REAL OPTIONS TO NEW PRODUCT DEVELOPMENT 177
In contrast, Toyota applies a slow convergence strategy, a key
component of what Ward et al. [5] term set-based concurrent
engineering.Upon evaluating the initial range of alternatives,
rather than selecting the apparently best alternative, PD teams
develop the set of viable alternatives from multiple perspectives.
Alternatives are gradually eliminated as weaknesses in areas
of performance, reliability, cost, manufacturability, and systems
integration become apparent. Eventually, the process converges
to a single, nal alternative. This can occur as late as the ve-
hicle prototype stages for some vehicle components. In addi-
tion, Toyota often uses a set-based approach as it develops engi-
neering requirements, rst dening a range that is gradually re-
ned to a nal requirement as alternatives are explored [28]. On
the surface, a slow convergence strategy would seem to result
in slow and expensive development projects, but the astounding
fact is, Toyota has been the industry benchmark in automobile
development speed for many years, and does so with a leaner
engineering workforce than most competitors! Ward et al. [5]
called this apparent contradiction the Second Toyota Paradox.
Design theory and the PD literature strongly support gener-
ating multiple alternative solutions to design problems as an es-
sential component of an effective PD process [29], [30]. Toward
this goal, tools and strategies have been developed to aid design
teams in selecting the best alternative, such as decision trees and
alternative comparison matrices. If uncertainty is small enough,
these approaches allow useful analysis and effective alterna-
tive selection, but often the performance, costs, and impacts on
project duration of undeveloped alternatives cannot be predicted
accurately enough in early stages to identify the best alternative.
Take, for example, the challenge of selecting a body style
for a new automobile from among several alternatives. On one
hand, management would like to decide on a body style early in
the process so that all development resources can focus on the
chosen style with the aim of successfully bringing the product
to market on time and under budget. On the other hand, body
design involves a complex set of tradeoffs between aesthetics
and creative design, aerodynamics, vehicle packaging, and man-
ufacturability. Developing alternatives in parallel not only en-
hances the creative process by considering a broad range of
competing designs, but also enables the design team to gather in-
formation on the tradeoffs associated with the set of alternatives.
This information is invaluable in making the best alternative se-
lection and/or in creatively resolving conicts among the com-
peting criteria. In addition, stamping technology itself is fairly
complex and subtle, and it is not always obvious what is easily
manufacturable. So developing multiple design alternatives in
parallel and delaying design decisions can be advantageous.
A. Case Study of Set-Based Development at Toyota
To illustrate the set-based approach in practice and to pro-
vide a basic framework on which to base the model described
later, this section presents a case study of a typical development
process for a major automobile subsystem at Toyota. This case
is based on Sobeks [31] study of Toyotas vehicle development
system from dozens of hours of interviews with 23 Toyota engi-
neers and engineering managers across the major vehicle devel-
opment units in Japan, engineers at numerous Toyota suppliers,
and several American engineers working in Toyotas U.S. op-
erations. Interviews were semi-structured and data were cap-
tured from eld notes taken during the interviews combined
with notes transcribed from memory immediately following the
interview. Interview data were systematically analyzed using a
grounded theory methodology [32].
1) Concept Design: The vehicle development program be-
gins with the chief engineer developing a vehicle concepta
written document describing his vision for the product and the
overall vehicle specications. The functional engineering units
(body engineering, chassis engineering, etc.) develop compa-
rable concept documents in parallel. Vehicle stylists simulta-
neously develop dozens of two-dimensional renderings of nu-
merous artistic conceptions for the chief engineer so he can see
the impact of decisions regarding vehicle specications. Exam-
ples of questions concerning body development include What
is the aesthetic impact of a shorter wheel-base?or How does
a taller passenger compartment impact driver comfort?This
process ends with the approval of the chief engineers concept.
2) System-Level Design: After concept approval stylists
continue to explore ideas based on the chief engineers concept
document, and eventually 610 ideas are selected for 1/5 scale
clay models. Body engineering conducts engineering design
studies on these ideas, called kentouzu drawings. These studies
include all the planning required to realize the vehicle: typical
cross sections of structural parts, joint denition, preliminary
part layout, wire harness routing, crash analysis, and so on.
Taking input from engineering, manufacturing, and aesthetic
evaluation, 23 ideas are chosen for full-scale clay modeling.
The kentouzu studies continue on the remaining alternatives
and feedback is given to the chief engineer and styling team.
By the time styling approval occurs, body engineering has laid
out a fairly well-dened body design plan for each alternative
under consideration. Styling approval is a high-publicity event
within the corporation at which one style is chosen. That styles
surface geometry is then converted to digital CAD data and sent
to body engineering. Body engineering compiles the relevant
kentouzu drawings, updating or adding to them as necessary,
and creates a document called the body structures design plan,
or kozokeikaku (K4) in Japanese. The K4 is distributed for
approval and feedback from all relevant vehicle development
units.
3) Detail Design: From the K4, detailed design begins on all
the body panels and structural components. As they are com-
pleted, body engineering drawings are sent to die engineering
without tolerances, delaying the determination of nal dimen-
sions until after die tryout. Die engineers design die sets to create
parts that are as close to nominal dimensions as possible, then
have soft-tool dies produced that can stamp out prototype parts.
The prototype parts are assembled in a slow buildof the ve-
hicle. Fit and function are adjusted through modications to the
part or die design, whichever is most appropriate. Full vehicle
prototypes are built and tested. Starting with prototype testing
and continuing past completion, hard-tool dies are ordered and
tried out with a rst set of parts. These parts are assembled into
ascrew bodyand the nal adjustments are made to the body
or die design, whichever is most economical and achieves engi-
neering standards. From there, the vehicle enters a trial produc-
tion stage before ramping up to full production.
Authorized licensed use limited to: MONTANA STATE UNIV BOZEMAN. Downloaded on October 20, 2009 at 15:44 from IEEE Xplore. Restrictions apply.
178 IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, VOL. 52, NO. 2, MAY 2005
Fig. 1. Body design and development at Toyota.
Fig. 1 depicts this process graphically. The diagram shows
the gradual narrowing of alternatives over time. The process can
be divided into three phases as shown. The double-line arrows
represent drawings without tolerances, which imply that a range
(albeit small) of nal designs may be acceptable.
Although prior work has demonstrated that Toyota achieves
competitive advantage using set-based development, parallel
development also requires resources for alternatives that will
not be realized, potentially increasing overall development
costs. In fact, these development costs can be so large that
they dominate the creation of value by the project. Without
understanding the underlying mechanisms at work, applying
set-based development is fraught with risk. A better under-
standing of the underlying causal relationships within this
approach is needed for PD organizations to take maximum
advantage of the approach while minimizing risk. Because
Toyotas purposeful delaying of design alternative selection
decisions resembles the use of real options in other contexts,
the current work uses a real options approach as a rst step to
explain the utility of set-based approaches to PD.
IV. ADAPTING REAL OPTIONS TO PRODUCT DEVELOPMENT
To adapt and operationalize real options concepts into PD ap-
plications, using the Second Toyota Paradox as a test case, we
must rst formalize the convergence decisions from an options
perspective. In addition, to evaluate the effectiveness of alterna-
tive strategies, we must develop a method for valuing the exi-
bility afforded by the strategy. This section presents the formal-
ization of convergence decisions and our exibility valuation
approach.
A. Formalization of PD Convergence Decisions
In real options terms, a project that converges early (point-
based) does not retain options to abandon alternatives in the
event that the chosen alternative does not work out or encoun-
ters signicant difculties. In other words, delaying the start of
convergence retains an option to choose any design alternative
while additional information becomes available. This is consis-
tent with the classic real options message, that delaying deci-
sions while uncertainty resolves can add value.
One way to structure this form of alternative selection is with
a decision rule. An option holder decides whether to exercise
the option based on the assets performance when compared
to specied criteria. For an option to buy a common stock, the
stocks value at the decision time is the signal, and the
stocks minimum value to justify purchasing the stock (the strike
price, ) is the decision criterion. The decision rule is
THEN Purchase Stock
ELSE Do not Purchase Stock (1)
Adapting this basic decision rule for use in PD may require
multiple signals from the development project (S) and several
criteria (C) to decide whether or not to exercise an option. There-
fore the signal(s) about the assets performance and decision
criteria can be represented as vectors of parameter values. In-
creasing net project value to the option-holder is the traditional
justication for exercising an option, but the literature on the
strategic advantages of real options [3], [4], [16], [19] suggests
noneconomic benets that may also be the basis for deciding
whether to exercise an option. Generalizing (1) for an option to
change a PD strategy, the decision rule becomes
IF THEN Change to New Strategy
ELSE Retain Current Strategy (2)
where
vector of signaling parameter values used to describe
performance;
vector of parameter values describing conditions that must
be met to justify a change in PD strategy.
To illustrate (2), a project manager could decide to change
from only loading the workforce in regular workweeks (the
current strategy) to one using overtime (the new strategy) if the
forecasted completion date ( ) exceeds the project deadline
(). The conditional portion of the decision rule can be ex-
panded to include multiple, and potentially linked, criteria; for
Authorized licensed use limited to: MONTANA STATE UNIV BOZEMAN. Downloaded on October 20, 2009 at 15:44 from IEEE Xplore. Restrictions apply.
FORD AND SOBEK: ADAPTING REAL OPTIONS TO NEW PRODUCT DEVELOPMENT 179
example, that the cost of other contingencies must also be less
than the forecasted budget surplus. The comparison operator in
(2) is dependent on how signaling and criteria parameters are
chosen and specied.
Toyota holds an option to select from among all design alter-
natives and then narrows to a single design through a gradual
elimination process. We model this elimination process as a set
of options to abandon individual design alternatives. Therefore,
we specify (2) by dening two option exercise condition param-
eters: the convergence initiation time and the probation pe-
riod . The convergence initiation time is the time from the
start of the project, when no alternatives have been eliminated, to
the time when the project team begins eliminating alternatives.
During this time (when ), development proceeds on all
alternatives. Once reaches , elimination of inferior alterna-
tives can begin. However, managers are aware that temporarily
quality problems can be misleading and, therefore, they do not
eliminate an alternative as soon as it appears to be the worst.
Instead, they require that an alternative be the apparent worst
for a minimum period of time before eliminating itthis is the
probation period. Therefore, the decision rule for eliminating an
alternative is
IF AND THEN Eliminate
ELSE Retain (3)
where
time that alternative has been the apparent worst
alternative;
probation period required to abandon an alternative;
project time;
convergence initiation time.
Components of the convergence timing decision rule (3) are
similar to some in traditional real options models.1However,
the convergence initiation time and probation period reect im-
portant managerial policy decisions. Describing alternative se-
lection decisions made in practice with these parameters in a
form similar to an option can help managers capture the strategic
thinking benets of real options described above.
Generally, point-based development is described with early
and fast convergence to a single alternative (small and )
and set-based development with later and slower convergence
(larger and ). Posed this way, development convergence
speed is characterized on a continuum of more or less set-based,
rather than as a binary decision. It also suggests that, under some
circumstances, one can be too set-based,taking the idea to the
extreme and possibly decreasing expected project value.
Assuming unique alternatives, only one alternative can be the
apparent worst at any time and, therefore, only that alternative
is a candidate for elimination. The decision rule in (3) does not
guarantee that only one alternative will remain at the end of
the project because uncertainty can create repeated changes in
1The convergence initiation time is a condition for exercise and is anal-
ogous to the lifespan of the option to select from among all alternatives, the time
to exercise. The time that an alternative has been the worst is a signal of
asset performance and is analogous to a current stock price. The probation pe-
riod is a condition for abandonment and analogous to a strike price.
the apparently worst alternative in periods less than the proba-
tion period, preventing the elimination of alternatives. Project
managers feel pressure to end the project with only one alterna-
tive and reduce the probation period as the project approaches
its deadline. Doing so accelerates the elimination of apparently
suboptimal alternatives until only one alternative remains at the
end of the project. This simple dynamic feature of decision-
making in PD management is a specic example of the com-
plexity that makes traditional approaches to modeling PD man-
agement with real options difcult to apply.
B. Flexibility Valuation
To address some of the challenges in applying traditional
option valuation models to PD management described above,
we use a simple, direct approach to valuing the exibility in
set-based development. The expected value of the exibility is
assumed to be the difference between expected project value
using (exible) set-based development and expected project
value using a more point-based (and, therefore, more rigid)
development
(4)
where
expected value of exibility provided by set-based
development;
expected value of a project using set-based develop-
ment ;
expected value of project using more point-based
development .
Positive values of exibility suggest an advantage of
set-based development over more point-based development,
while negative values suggest the opposite and recommend not
adopting set-based development.2
We specify this approach to the Second Toyota Paradox
as follows. At Toyota, missing launch dates is simply not
permitted. Toyota developers may work in excess of 80 h/week
as phase deadlines approach in order to nish projects by the
launch date. Since a project must nish within a specied time
frame regardless of which alternative is selected, values of
PD projects ( or ) can be quantied based on quality
and cost alone. Quality is quantied as the percent of the nal
alternatives scope without quality problems (e.g., defects
or unsatised requirements) and monetized with an average
marginal prot value for a percent of quality above a minimum
level . Development costs are the product of the amount of
work performed during the project on all alternatives ( ) and
an average unit cost of performing a development activity ( ).3
Expanding the project values in (4) to reect the value added by
quality and lost due to development costs, the net value of the
2Such a recommendation for point-based development when the value of ex-
ibility is negative is consistent with assigning an option no value and the implied
recommendation to not use options with negative values for exibility that is in-
corporated into the Max (0, value of exibility) form of some traditional real-op-
tion valuation models (e.g., binomial models).
3In comparing our structuring of exibility in set-based development to tra-
ditional option valuation models the value of quality is analogous to the value
of an income stream and development costs are analogous to stock dividends.
Authorized licensed use limited to: MONTANA STATE UNIV BOZEMAN. Downloaded on October 20, 2009 at 15:44 from IEEE Xplore. Restrictions apply.
180 IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, VOL. 52, NO. 2, MAY 2005
exibility in set-based development compared with point-based
development is, therefore
(5)
where
nal alternative quality at project completion with set-
based or point-based strategy;
average marginal value of a percent of quality;
development work effort with set-based or point-
based strategy;
unit development cost.
Rearranging and consolidating to more directly reect differ-
ences between performance in quality and development effort
with the two development approaches
(6)
The quality ( ) and development effort ( ) of PD projects
managed with different design alternative selection approaches
are estimated as the mean values of many projects with dif-
fering evolutionary paths for the uncertain parameters as simu-
lated with a dynamic computer model, described next. Although
the model and (6) provide a means of estimating the economic
value of exibility, the conceptual and formal model potentially
are as, or more, valuable for their ability to describe project be-
havior and performance as well as economic value.
V. P RODUCT DEVELOPMENT SIMULATION MODEL
From the case study of vehicle development at Toyota
and the formalized decision rule for convergence just de-
scribed, a system dynamics model [33], [34] was built that
reects Toyotas development processes. System dynamics is
a methodology for studying and managing complex systems.
This approach focuses on how performance evolves in response
to the interactions of management decision-making and de-
velopment processes. System dynamics has been successfully
used to explain failures in fast track implementation [35], [36],
impacts of changes [37], and other project management issues,
and is considered appropriate for modeling PD processes.
A. Model Structure
The model has four sectors. In the work ows sector PD
work packages move through a rework cycle based on work by
Ford and Sterman [38], but the model was adapted for the cur-
rent work to reect multiple alternatives, the interactions among
phases within alternatives, and the path dependent uncertain-
ties that impact progress. The current work also developed a
resource sector that adjusts the quantity of development labor
based on remaining work and the impacts of schedule pressure,
and allocates resources among alternatives and development ac-
tivities within alternatives. The alternative selection sector de-
scribes Toyotas managerial decision-making based on (3) and
is unique to this model. A performance sector models project
duration, quality, and development costs.
In the work ows sector, work packages move through four
parallel development efforts, each representing a different
design alternative. If not eliminated, each alternative passes
through three sequential phases as depicted in Fig. 2. The
Fig. 2. Model structure of alternatives in Toyotas development process.
model simulates the ows and accumulations of development
work within and among phases, as impacted by the uncertainty
in predicting alternative quality and managerial behaviors.
Within each phase a rework cycle begins with a stock of de-
velopment work to be completed. As work is completed, there is
a probability that a change is needed in the work. Changes could
be defects or errors that must be corrected to meet minimum re-
quirements or could be improvement opportunities to increase
customer satisfaction or improve manufacturability. Completed
work is checked by quality assurance and either passes inspec-
tion and moves to the next phase, or does not pass and enters a
rework loop. In the rework loop, work is corrected or improved
with a probability of creating or exposing the need for another
change, and then returned to quality assurance. Rework not dis-
covered by the originating phase is released, reducing the overall
quality of the alternative.
The four development activities in each phase (initial com-
pletion, quality assurance, change, and coordination) require
resources. Consistent with Toyotas management practices, re-
sources (primarily labor) are dedicated to phases and allocated
to alternatives and development activities within phases in di-
rect proportion to labor demand (Fig. 3).
The willingness of Toyota developers to work in excess of 80
hours per week to meet deadlines is modeled with a set of strong
control loops in the resources sector that adjust labor quantity
in response to schedule pressure and force the project to nish
by the prescribed deadline. Product quality, measured with the
fraction of work that do not require changes, is controlled in
the model through feedback loops that shift resources toward
quality assurance and change activities when those backlogs are
relatively large. Therefore, when quality drops and testing and
rework backlogs grow, more developers are assigned to those
activities to improve quality.
B. Modeling Design Alternatives
The four alternatives are distinguished by their complexity
and tractability in each phase (24 descriptive variables, see
Table I). Complexity reects the novelty and difculty of
development and is modeled with the likelihood that ini-
tially completing a work package generates a need for a change
[P(generate change initially)]. Tractability reects the difculty
of discovering and subsequently changing work and is modeled
with the likelihood of discovering a change [P(discover change
need)]. In this way, alternatives with different characteris-
tics can be modeled. For example, an alternative with many
hiddendefects can be modeled by assigning a high change
generating probability and a low change discovery probability.
Likewise, an alternative with many easily resolved problems
Authorized licensed use limited to: MONTANA STATE UNIV BOZEMAN. Downloaded on October 20, 2009 at 15:44 from IEEE Xplore. Restrictions apply.
FORD AND SOBEK: ADAPTING REAL OPTIONS TO NEW PRODUCT DEVELOPMENT 181
Fig. 3. Resource allocation in model of Toyota development.
TABLE I
INITIAL AND MEAN VALUES OF UNCERTAINTY PARAMETERS
can be modeled by assigning a high change generation prob-
ability and a high change discovery probability. We chose
parameter values so that more complex alternatives are less
tractable and less complex alternatives are more tractable. The
descriptive parameters for the four alternatives are specied
according to Table I. Otherwise, all alternatives are identical.
Notice that Alternative 1 generates the fewest changes and
discovers the largest fraction of those change needs, whereas
Alternative 4 generates the most changes and discovers the
lowest fraction. The alternatives were described such that
they increase in complexity and decrease in tractability from
Alternative 1 to Alternative 4. This was done so that the alterna-
tives would unambiguously and monotonically progress from
best (Alternative 1) to worst (Alternative 4), thereby providing
a means of assessing the quality of alternative selection. Better
selection is reected in lower average nal alternative numbers.
C. Modeling Uncertainty
Uncertainty is modeled by varying the likelihood of gener-
ating a change [P(generate change initially)] and the likelihood
of discovering a need for a change [P(discover change need)]
within each phase and alternative (24 uncertain variables).
Their behaviors are path dependent to reect the evolutionary
nature of alternative development. In each time period, uncer-
tain values move a portion of the distance to their mean value
(Table I) and a random draw from a normal distribution, con-
strained to keep probabilities within the range {0, 1}. Similar
random mean-reverting behavior has been used to value other
real options (e.g., [39] and [40]). Uncertainty levels relative to
mean values are kept constant (10%) for all uncertain variables
by specifying the coefcient of variation ( standard devia-
tion/mean). Changing the initial seeds of the random number
generators allows the simulation of different scenarios of the
evolution of the uncertain variables.
Alternative elimination in the model implements (3). To
reect the PD managers perspective, only discovered imper-
fections are used in convergence decision-making. Therefore,
the model captures the impacts of change discovery timing, an
important PD management challenge. The performance sector
implements (6).
D. Model Calibration, Behavior, and Testing
Detailed model calibration data was not available. Therefore,
the model was calibrated to reect a generic single system of an
automobile development, such as body development, based on
information about Toyota practice. From its annual report [41],
Toyota spends approximately $6.7 billion on research and devel-
opment per year, supporting 1012 models per year. This cor-
responds to approximately $610 million in R&D expenses per
model on average. If body engineering accounts for about 1/5 of
the development costs (based on engineering head count), body
development would account for $121 million of a vehicles de-
velopment cost. The average simulated development cost for a
project with a convergence initiation time of 450 days (a reason-
able estimate of Toyota practice) was $123 million.
New model introductions are generally on a four-year
cycle. However, development projects at Toyota, at the time
the Paradox was documented, started about 3033 months
before product launch. Styling approval in recent programs has
occurred as late as just 15 months before launch. Thus, the
Authorized licensed use limited to: MONTANA STATE UNIV BOZEMAN. Downloaded on October 20, 2009 at 15:44 from IEEE Xplore. Restrictions apply.
182 IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, VOL. 52, NO. 2, MAY 2005
model uses convergence initiation times varying from zero (to
simulate very point-based convergence) to 900 days, well past
the 15 months, to capture the full range of convergence timing
possibilities.
Product quality is paramount in the automobile industry, and
quality is a cornerstone of Toyotas success. The model mea-
sures quality as a percent of work packages with imperfections.
Imperfections could be anything from a minor manufactura-
bility opportunity (worth tens of thousands of dollars over the
vehicles production life) to a major warranty or recall issue
that costs the company hundreds of millions of dollars. Given
the high importance of quality, we chose a fairly high marginal
value of quality ($10 million per percent quality).
According to Toyotas annual report [41], Toyotas net
income per vehicle sold in 2003 was $1000 on average, world-
wide. Ford Motor Corporation (not a set-based developer), had
net earnings of about $460 per vehicle sold in North America
[42]. Assuming 100 000 vehicles sold per year and four-year
production life, a typical Toyota model nets $216 million
more than a typical Ford model. Again assuming that the body
subsystem accounts for roughly 1/5 of this, we would expect
the model to produce value differences in the neighborhood of
$43 million. This value will be used later to validate the model
calibration.
The model was also tested for usefulness using established
system dynamics validation methods [34], [43]. The structural
consistency of the model with the target system is strengthened
by the use of previously developed and tested system dynamics
model structures that reect the in-depth knowledge of Toyota
development processes. Qualitatively, model behavior appears
consistent with development project behavior and Toyotasex-
perience. In addition, model behavior is reasonable under a wide
range of parameter values. For example, increasing available re-
sources signicantly decreases project durations, and increasing
the differences among alternatives make alternative selection
more consistent across many uncertainty scenario simulations.
The results of model calibration and testing support the use of
the model to test the operationalization of real options concepts
to the Second Toyota Paradox.
VI. SIMULATION RESULTS
Expected project performance is the average of 100
simulations for a specic condition. Point-based develop-
ment was simulated as the basis for performance
differences in (6). Toyota typically does not delay the initia-
tion of convergence until very near the end of their projects.
However, they do consider when to initiate convergence across
some time span. Therefore, we investigated performance and
value with convergence initiation times over essentially the
entire project length (0900 days).
A. Average Project Quality
As increases from 0 to 900, expected quality improves
from a minimum of 73% to a maximum of 92%, as shown in
Fig. 4. We surmise that quality improvements are a result of
better alternative selection because the average number of the
Fig. 4. Expected quality of point-based and set-based development projects.
Fig. 5. Expected values of exibility in set-based development relative to
point-based development.
nal alternative decreases with the convergence initiation time
in a pattern similar to Fig. 4. This is consistent with the descrip-
tion of decision-making in set-based development and our con-
ceptual design of it as an optionas teams wait longer to start
eliminating design alternatives, they gain access to more and
more accurate quality information and, therefore, make wiser
choices, on average.
But quality levels off after a certain pointdecisions do not
improve after a certain amount of delay. In our model and in
practice, waiting longer, developing multiple alternatives, and
gathering more information after a certain point may increase
managerial condence, but it does not necessarily improve
decision-making. Interestingly, quality does not increase con-
tinuously, indicating transitory periods of development work
that do not yield additional useful information, followed by
periods of improvement. We speculate later on potential causes
of these transitions.
B. Average Development Cost
The expected point-based development cost [ in (6)]
is $73 million. As the convergence initiation time increases the
development costs [ in (6)] increase roughly linearly
to $239 million when convergence begins on day 900. More
alternatives kept alive longer translate into higher development
costs.
C. Value of Flexibility in Set-Based Development
Fig. 5 displays the marginal values of exibility in set-based
development over point-based development for increasing
Authorized licensed use limited to: MONTANA STATE UNIV BOZEMAN. Downloaded on October 20, 2009 at 15:44 from IEEE Xplore. Restrictions apply.
FORD AND SOBEK: ADAPTING REAL OPTIONS TO NEW PRODUCT DEVELOPMENT 183
convergence initiation times.4Note that the increase in value as
increases from 180 to 360 days is $50 million, which is close
to the calibration estimate of $43 million described earlier.
Similarly, value increases by a similar magnitude again as
increases from 360 to 450 days. This provides some validation
that the model results are reasonable.
Fig. 5 illustrates how the value of exibility initially increases
as convergence initiation time increases due to better alternative
selection that outweighs increases in development costs. After
convergence delays have improved alternative selection nearly
as much as possible (450 days of delay under this set of condi-
tions), further delaying alternative elimination incurs additional
cost without increased quality, and the net value of set-based
exibility decreases.
Fig. 5 also displays an oscillatory effect, where the marginal
value of increasing convergence time can actually decrease for
a period, then increase again. These periods are associated with
the latter portions of development phases. Inspection of model
behavior indicates that during these times development work oc-
curs in parallel, thus increasing cost, but most of the quality is-
sues that can be found in the phase have been discovered and,
therefore, the additional information is insufcient to improve
alternative selection enough to add more value than the cost of
parallel development. New information becomes available rela-
tively suddenly when a new development phase begins, thereby
signicantly increasing the probability of selecting a better al-
ternative. This characteristic seems consistent with PD practice.
In summary, the simulation results demonstrate how a real
options approach can be adapted and operationalized to model
design alternative selection. Set-based development decision-
making rules can be modeled like exercise decisions used in tra-
ditional options, but in forms that allow for multiple dimensions
of performance and variety in strategies [(1)(3)]. Modeling that
reects many of the complexities of PD practice can provide
a means of developing insight about how options impact per-
formance (e.g., Fig. 4) and can capture the value of exibility
[(4)(6) and Fig. 5].
VII. MANAGERIAL IMPLICATIONS
The current work has several potential implications for PD
managers, despite the preliminary nature of the investigation.
First, structuring set-based exibility with real options concepts
suggests that managers can use options to improve their recog-
nition, descriptions, and modeling of managerial exibility, a
potentially useful PD approach that to date has been tacit and
difcult to improve and exploit [11]. To the extent that adap-
tations of real options improve managersmental and formal
models of exibility, their understanding and insight about the
effectiveness of development plans can also improve. In addi-
tion, practicing managers at Toyota and elsewhere [11] seem to
implicitly recognize the value of hedging their bets to acquire
more and better information on alternatives, but many compa-
nies do not have a culture that values broad searches of alterna-
tive solutions. Real options adaptations that closely reect prac-
4From a more traditional options perspective, Fig. 5 is analogous to a graph
of different times to the exercise date (convergence initiation time) versus the
value of a European option (set-based exibility) on an asset that pays very large
dividends (development costs).
tice may help managers recognize that developing alternatives
in parallel (and options in general) has utility, with potential for
signicant improvements in PD management. Still, these results
are preliminary and more work is needed to incorporate the op-
tions perspective into actionable tools to fully test these potential
impacts.
Second, real options may provide a partial explanation
for how delayed convergence adds value and contributes to
Toyotas sustained competitive advantage. Explaining the
Paradox with an adaptation of established tools and methods
that are more general than Toyota (real options) provides a
foundation for customizing Toyotas process to other PD con-
texts. For example, under the conditions modeled, increasing
convergence initiation time beyond a certain point does not add
value. Under other conditions, this point of negative returns for
holding the option longer may be much earlier (e.g., low end
consumer goods) or much later (e.g., high cost of failure, as
in space programs). Although further investigation is needed,
the current work suggests that adapting and operationalizing
real options may be helpful in effectively transferring set-based
practices to other PD contexts.
Finally, the shape of the curves in Figs. 4 and 5 suggests both
a promise and a warning for PD managers. One possible insight
for managers is the existence of an optimal convergence initi-
ation time for maximizing the value added by the exibility in
set-based development. For the conditions modeled here, the op-
timal lifespan is nearer the middle of the project life than the
beginning or end of the project. The results can help managers
recognize the importance of estimating the size of the optimal
lifespan.
But, transitory periods may exist early in a project when the
quality of decision-making and value-adding may stall or even
decline, and that delaying convergence a little longer (e.g., into
a new development phase) may restart improvements because
enough new knowledge is being gained to improve decisions.
More generally, this warns managers that using real options to
manage PD is not necessarily straightforward. Managers are
warned to consider carefully and thoroughly if and how real op-
tions can be effectively designed and used before attempting to
exploit their potential to improve insight and add value.
VIII. CONCLUSION
We have proposed an adaptation of the real options approach
to managing PD for the purpose of developing insight into how
strategies to manage uncertainty through exibility impacts
project behavior, performance, and value. We operationalized
the adaptation to model design alternative selection in set-based
development. Managerial decisions about changing develop-
ment strategies were structured as decision rules based on one
used in traditional options and specied for the elimination of
inferior alternatives. A formal PD model simulated projects that
retain exibility for different periods of time to investigate how
project performance and value changes with different exible
strategies. Results described how, and to some extent why,
delaying the initiation of design convergence impacts decision
quality and project value.
The current work makes at least two valuable contributions.
First, we have demonstrated that real options concepts can be
Authorized licensed use limited to: MONTANA STATE UNIV BOZEMAN. Downloaded on October 20, 2009 at 15:44 from IEEE Xplore. Restrictions apply.
184 IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, VOL. 52, NO. 2, MAY 2005
adapted to reect the process and managerial complexity of PD
practice, and that these adaptations can be operationalized to
develop insight into how and why exible development plans
impact project behavior, performance, and value. In doing so,
we explore an options modeling method that relaxes common
assumptions used in some traditional real options valuation
models and that can more closely reect PD practice. However,
considering the complexity and specicity of the set-based de-
cision-making modeled here, we also conclude that effectively
designing and implementing real options to increase PD project
value requires a deep understanding of development processes,
managerial and developer behavior, option designs, and how
their interactions impact project value. This work also suggests
methods that may be used to expand real options models to a
broader range of assets, such as those with uncertainties that
are proactively managed by those who hold options on those
assets.
The current work also contributes to the existing body of work
on new PD and technology management. We propose and ini-
tially test the applicability of real options as a framework for
improving the understanding and management of PD projects.
We provide a preliminary test of the ability of real options con-
cepts to explain the Second Toyota Paradox. We also expand the
set-based theory from a binary description (point-based or set-
based) into a continuum of set-based development plans based
on the timing of design convergence.
Much work, however, remains to be done, both from PD and
real options perspectives. Validation would be improved by cal-
ibrating the model more tightly to actual development statis-
tics, particularly across multiple projects, and validating the re-
sults empirically. Although data collection and analysis would
be challenging, the results would signicantly improve con-
dence in the conclusions. Future work will investigate the de-
sign of set-based development with other parameters as man-
agerial levers to control PD projects (e.g., probation period, or
selecting the best alternative versus eliminating the worst), ef-
fects of the characteristics of uncertainty, and how PD man-
agement might differ with respect to specic characteristics in
the set of alternatives (e.g., Do optimal decision points change
if the set includes a fall-backalternative?). Impacts of fac-
tors that are important in traditional real options models but
less important in set-based decision-making (such as discount
rates and exercise costs) or that are important in set-based deci-
sion-making but less so in traditional real options models (such
as development costs) will also be investigated to improve the
understanding of what project components to utilize in real op-
tions application, and how. Finally, we hope this work will cul-
minate in the development of tools and methods that apply real
options to PD management and help development teams per-
form more effectively.
ACKNOWLEDGMENT
The authors would like to thank C. Mailet and A. Tresarrieu
for modeling assistance, and J. Liker and the reviewers for their
careful review and insightful feedback.
REFERENCES
[1] D. N. Ford, D. Lander, and J. Voyer, The real options approach to
valuing strategic exibility in uncertain construction projects,Con-
struction Manage. Econ., vol. 20, no. 4, pp. 343352, 2001.
[2] V. Krishnan and K. Ulrich, Product development decisions: A review
of the literature,Manage. Sci., vol. 47, no. 1, pp. 121, 2001.
[3] M. Amram and N. Kulatilaka, Real Options, Managing Strategic Invest-
ment in an Uncertain World. Boston, MA: Harvard Business School
Press, 1999a.
[4] ,Disciplined decisions: Aligning strategy with the nancial mar-
kets,Harvard Bus. Rev., vol. 77, pp. 95104, 1999b.
[5] A. Ward, J. K. Liker, J. J. Cristiano, and D. K. Sobek, The second
Toyota paradox: How delaying decisions can make better cars faster,
Sloan Manage. Rev., vol. 36, no. 3, pp. 4361, 1995.
[6] R. Brealey and S. Myers, Principles of Corporate Finance. New York:
McGraw-Hill, 2000.
[7] A. K. Dixit and R. S. Pindyck, Investment Under Uncer-
tainty. Princeton, NJ: Princeton Univ. Press, 1994.
[8] M. Benaroch, Option-based management of technology investment
risk,IEEE Trans. Eng. Manage., vol. 48, no. 4, pp. 428444, 2001.
[9] M. J. Brennan and L. Trigeorgis, Project Flexibility, Agency, and Com-
petition, New Developments in the Theory and Application of Real Op-
tions. New York: Oxford Univ. Press, 2000.
[10] A. G. Z. Kemna, Case studies on real options,Fin. Manage., vol. 22,
no. 3, pp. 259270, Autumn 1993.
[11] R. Miller and D. Lessard, The Strategic Management of Large Engi-
neering Projects: Shaping Institutions, Risks, and Governance. Cam-
bridge, MA: The MIT Press, 2000.
[12] L. Trigeorgis, Ed., Real Options in Capital Investment. New York:
Prager, 1995.
[13] C. Y. Baldwin and K. B. Clark, Design Rules, the Power of Modu-
larity. Cambridge, MA: MIT Press, 2000.
[14] P. Sharpe and T. Keelin, How Smithkline Beecham makes better
resource-allocation decisions,Harvard Bus. Rev., pp. 4557, Mar. 1,
1998.
[15] S. Bounkendour and R. Bah, The guaranteed maximum price contract
as call option,Construction Manage.Econ., vol. 19, pp. 563567, 2001.
[16] H. Courtney, J. Kirkland, and P. Viguerie, Strategy under uncertainty,
Harvard Bus. Rev., pp. 6779, Nov.Dec. 1997.
[17] H. Bierman and S. Smidt, The Capital Budgeting Decision—Economic
Analysis of Investment Projects, 8th ed. New York: Macmillan, 1992.
[18] J. W. Kensinger, The capital investment project as a set of exchange
options,Manage. Fin., vol. 14, no. 2/3, pp. 1627, 1988.
[19] D. Ford, D. Lander, and J. Voyer, Business strategy and real options in
the context of large engineering projects,J. Competitiveness, 2004, to
be published.
[20] R. M. Bookstaber, Option Pricing and Strategies in In-
vesting. Reading, Ma: Addison-Wesley, 1982.
[21] J. C. Cox, S. A. Ross, and M. Rubinstein, Option pricing: A simplied
approach,J. Fin. Econ., vol. 7, pp. 383402, 1979.
[22] F. Black and M. Scholes, The pricing of options and corporate liabili-
ties,J. Political Econ., vol. 81, no. 3, pp. 637654, 1973.
[23] D. M. Lander and G. E. Pinches, Challenges to the practical implemen-
tation of modeling and valuing real options,Quart. Rev. Econ. Fin., vol.
38, pp. 537567, 1998.
[24] M. J. Garvin and C. J. Cheah, Valuation techniques for infrastructure in-
vestment decisions,Construction Manage. Econ., vol. 22, pp. 373383,
May 2004.
[25] T. Alessandri, D. Ford, D. Lander, K. Leggio, and M. Taylor, Identi-
fying and valuing uncertainty in capital budgeting projects: Beyond dis-
counted cash ow analysis,Quart. Rev. Econ. Fin., to be published.
[26] J. K. Liker, D. K. Sobek II, A. C. Ward, and J. J. Cristiano, Involving
suppliers in product development in the United States and Japan: Evi-
dence for set-based concurrent engineering,IEEE Trans. Eng. Manage.,
vol. 43, no. 2, pp. 165178, May 1996.
[27] A. Yassine, N. Joglekar, D. Braha, S. Eppinger, and D. Whitney, Infor-
mation hiding in product development: The design churn effect,Res.
Eng. Design, vol. 14, pp. 145161, 2003.
[28] D. K. Sobek, J. K. Liker, and A. C. Ward, Principles of Toyotas set-
based concurrent engineering,Sloan Manage. Rev., vol. 40, no. 2, pp.
6783, 1999.
[29] S. Pugh, Total Design: Integrated Methods for Successful Product En-
gineering. Wokingham, U.K.: Addison-Wesley, 1991.
[30] K. T. Ulrich and S. D. Eppinger, Product Design and Development, 2nd
ed. New York: McGraw-Hill, 1999.
Authorized licensed use limited to: MONTANA STATE UNIV BOZEMAN. Downloaded on October 20, 2009 at 15:44 from IEEE Xplore. Restrictions apply.
FORD AND SOBEK: ADAPTING REAL OPTIONS TO NEW PRODUCT DEVELOPMENT 185
[31] D. K. Sobek II, Principles that shape product development systems: A
Toyota-Chrysler comparison,Ph.D. dissertation, Univ. Michigan, Ann
Arbor, MI, 1997.
[32] A. Strauss and J. Corbin, Basics of Qualitative Research: Grounded
Theory Procedures and Techniques. Newbury Park, CA: Sage, 1990.
[33] J. W. Forrester, Industrial Dynamics. Reading, MA: Productivity
Press, 1961.
[34] J. D. Sterman, Business Dynamics, Systems Thinking and Modeling for
a Complex World. New York: McGraw-Hill, 2000.
[35] D. Ford and J. Sterman, Overcoming the 90% syndrome: Iteration
management in concurrent development projects,Concurrent Eng.
Res. Appl., vol. 11, no. 3, pp. 177186, 2003.
[36] T. Abdel-Hamid, Understanding the 90% syndromein software
project management: A simulation-based case study,J. Syst. Softw.,
vol. 8, pp. 319330, 1988.
[37] K. G. Cooper, Naval ship production: A claim settled and a framework
built,Interfaces, vol. 10, no. 6, pp. 2036, 1980.
[38] D. N. Ford and J. D. Sterman, Dynamic modeling of product develop-
ment processes,System Dynamics Review, vol. 14, no. 1, pp. 3168,
1998.
[39] J. E. Smith and K. F. McCardle, Options in the real world: Lessons
learned in evaluating oil and gas investments,Oper. Res., vol. 47, no. 1,
pp. 115, 1998.
[40] R. L. McDonald, Real options and rules of thumb in capital budgeting,
in Project Flexibility, Agency, and Competition, M. Brennan and L. Tri-
georgis, Eds. New York: Oxford Univ. Press, 2000.
[41] Toyota Motor Corporation. (2004, Jul.) 2003 Annual report. [Online].
Available: http://www.toyota.co.jp/en/ir/reports/annual_reports/03/
[42] Ford Motor Corporation. (2004, Jul.) 2003 Annual report. [Online].
Available: http://www.ford.com/en/company/investorInformation/com-
panyReports/annualReports/2003annualReport/default.htm
[43] J. Forrester and P. M. Senge, Tests for building condence in system
dynamics models,Manage. Sci., TIMS Series, vol. 14, pp. 209228,
1980.
David N. Ford (M86) received the B.S. and M.S.
degrees from Tulane University, New Orleans, LA,
and the Ph.D. degree from the Massachusetts Institute
of Technology, Cambridge.
He is a Professor in the Construction Engineering
and Management Program, Department of Civil
Engineering, Texas A&M University, College
Station. Prior this position, he was on the Faculty of
the Department of Information Science, University
of Bergen, Bergen, Norway, where he researched and
taught in the System Dynamics Program. For over
14 years, he designed and managed the development of constructed facilities in
industry and government. His current research interests include the dynamics
of development supply chains, strategic managerial exibility, and resource
allocation policies.
Dr. Ford is a member of INFORMS, ASCE, and other professional
organizations.
Durward K. Sobek, II received the A.B. degree
in engineering science from Dartmouth College,
Hanover, NH, and the M.S. and Ph.D. degrees
in industrial and operations engineering from the
University of Michigan, Ann Arbor.
He is currently an Associate Professor of Indus-
trial and Management Engineering, Montana State
University, Bozeman. His current research interests
include product development, engineering design ed-
ucation, and lean applications to health care.
Dr. Sobek, II is a member of the American Society
of Engineering Education (ASEE) and the Institute of Industrial Engineers (IIE).
Authorized licensed use limited to: MONTANA STATE UNIV BOZEMAN. Downloaded on October 20, 2009 at 15:44 from IEEE Xplore. Restrictions apply.
... The resulting design remediation costs may become prohibitively expensive for rework occurring late in product development. Decision delay helps to limit design remediation requirements and cost by reducing early cost commitment and retaining design flexibility later into the production development lifecycle [15]. ...
... However, there exist methodologies providing insights relevant to this research. These methods include the use of real options [15], Markov decision processes [16], integrated sequential decision process frameworks [1], and multilevel multiobjective decision models [12]. ...
Article
System design is an exercise in sequential decision-making, with the objective of developing resilient and affordable systems. Throughout the design process, engineering and program managers must balance several competing objectives, such as ensuring design feasibility, minimizing cost, schedule, and performance risk, while simultaneously achieving stakeholder value. Thus, engineering and program managers require design and analysis methods enabling complex and multiobjective design decisions under uncertainty. Unfortunately, there exists limited research providing quantitative methodologies specifically enabling program management decisions using quantitative set-based design. We, therefore, present a quantitative set maturation and uncertainty resolution decision methodology using value-focused multiobjective decision models and model-based engineering practices. This methodology assesses and quantifies uncertainty regarding stakeholder value, cost, requirements, and design maturity for each design set. These metrics facilitate the calculation of the program manager value, which when combined with design set feasibility entropy, enable tradeoff analysis informing design maturation and uncertainty resolution prioritization decisions. We develop and demonstrate our methodology using a model-based unmanned aerial vehicle case study implemented in the ModelCenter modeling environment. This methodology provides program managers an efficient, cost effective, and defensible approach to inform system design maturation and uncertainty resolution decisions enabling the development of resilient and affordable systems.
... L'approche de la dynamique des systèmes, comme outil d'exploration des tensions L'approche de la dynamique des systèmes a été proposée par Jay Forrester, à la fin des années 50, afin d'analyser la nature complexe des systèmes socio-économiques (Bloodgood et al., 2015). Elle a été appliquée à divers champs de recherche pour étudier des problèmes complexes, tels que les stratégies d'alliance internationales (Kumar et Nti, 2004), les relations contractuelles entre clients et fournisseurs (Marquez et Blanchar, 2004), le processus et les implications sur la performance d'un mouvement stratégique de diversification (Gary, 2005), la prise de décision dans les projets de développement de nouveaux produits (Ford et Sobek, 2005), le développement de l'entrepreneuriat en entreprise (Bloodgood et al., 2015), etc. ...
Research
Full-text available
Mémoire HDR La gestion des paradoxes stratégiques : vers une méta-capacité à développer face à la complexité
... The third tenet uses a slow set-narrowing process that includes increasing detail, a commitment to the selected set, and an uncertainty management process that uses process gates as elements to "establish feasibility". These steps and characteristics are consistent with the work of other researchers [18][19][20][21][22][23][24][25][26][27][28][29][30][31][32][33][34][35]. However, Specking et al. [36] performed a SBD literature search that demonstrated a lack of quantitative techniques in the previous SBD literature. ...
Article
Full-text available
System designers, analysts, and engineers use various techniques to develop complex systems. A traditional design approach, point-based design (PBD), uses system decomposition and modeling, simulation, optimization, and analysis to find and compare discrete design alternatives. Set-based design (SBD) is a concurrent engineering technique that compares a large number of design alternatives grouped into sets. The existing SBD literature discusses the qualitative team-based characteristics of SBD, but lacks insights into how to quantitatively perform SBD in a team environment. This paper proposes a qualitative SBD conceptual framework for system design, proposes a team-based, quantitative SBD approach for early system design and analysis, and uses an unmanned aerial vehicle case study with an integrated model-based engineering framework to demonstrate the potential benefits of SBD. We found that quantitative SBD tradespace exploration can identify potential designs, assess design feasibility, inform system requirement analysis, and evaluate feasible designs. Additionally, SBD helps designers and analysts assess design decisions by providing an understanding of how each design decision affects the feasible design space. We conclude that SBD provides a more holistic tradespace exploration process since it provides an integrated examination of system requirements and design decisions.
... The five articles addressing this paradox build on the design theory known as 'set-based concurrent engineering' and they do not consider cultural aspects in their analysis (Biazzo 2009;Malak, Aughenbaugh, and Paredis 2009). Nevertheless, some authors recognise the important role of OC in dealing with the changes resulting from new product development and in successfully managing this paradox (Belay, Welo, and Helo 2014;Ford and Sobek II 2005). ...
Article
This study investigates how the effective management of cultural paradoxes and dilemmas can support lean implementation in a service organisation. We have conducted an in-depth single case study employing the paradox theory as a theoretical lens. The findings offer a dynamic analysis of how organisational culture (OC) influences lean implementation, and in turn, how lean implementation directly impacts the OC. The study shows that in order to achieve a successful lean implementation, the organisation must reshape the OC by identifying when to manage cultural tensions as dilemmas or as paradoxes. We also discuss the existing defensive mechanisms (sources of resistance) and managerial actions taken to manage those tensions and resistances. Besides contributing to the lean literature, this study contributes to the operations management field more broadly by highlighting the need to replace the long tradition of the trade-off perspective by a more embracing perspective that accepts the co-existence of dilemmas (either-or choices) and paradoxes (balancing opposed goals). The practical implications rely on the identification of lean practices that counterbalance specific cultural traits, offering recommendations to managers dealing with the cultural clashes inherent to a lean implementation.
... Kennedy [12] presented the lean product development principles in the form of a novel. Principles from diverse fields such as queuing theory, information theory, and options pricing have been applied to understand and improve product development processes [13,14]. Since 2010, there has been increased literature published accompanied by industry case studies, for example [15][16][17]. ...
Conference Paper
Lean thinking is a successful management philosophy that originated in the production dominated mid-20th century. More recently, these principles have been contextualized in a product development setting. However, few companies have successfully adopted lean product development. The gap between the existing knowledge within the academic community and the current practices within the industry motivated an industry-academia research roundtable to identify the successes and challenges in implementing lean product development. This paper shares the successful practices, and the implementation challenges that industry faces in the context of lean product development, which include nomenclature, visual management and metrics, knowledge repository and lean education & training. The paper concludes with a roadmap of research required to enable the adoption of lean product development in the industry.
... Uncertainty in NPD projects creates significant challenges for managers who are under intense competitive pressures to reduce development time and costs (Ford and Sobek, 2005). Uncertainty can come in various forms. ...
Article
Full-text available
Given the complexity of sourcing choices in R&D, driven by conflicting points of difference, scholars have struggled to formulate clear theoretical predictions about which sourcing choices lead to superior performance. Our study contributes to the elucidation of this question by studying both the antecedents and performance consequences (project cost and duration) of alternative sourcing choices in the context of new product development, thus bridging two rarely intersecting streams of research to discover unique insights. We draw from the transaction cost economics (TCE) and organizational learning theory to account for the interdependence of cost-and non-cost related factors in firms' sourcing choices. Results from our empirical analysis, using data on clinical trials in the global biopharmaceutical industry, show that greater project complexity, project stage uncertainty, and prior sourcing experience determine the sourcing choices a firm makes. We also establish that alternative sourcing choices vary in their ability to minimize project cost versus project duration, and show that prior experience with a sourcing choice is an important moderator of project performance, in addition to being one of the determinants of sourcing choice.
Thesis
Full-text available
This dissertation comprises a body of research facilitating decision-making and complex system development with quantitative set-based design (SBD). SBD is concurrent product development methodology, which develops and analyzes many design alternatives for longer time periods enabling design maturation and uncertainty reduction. SBD improves design space exploration, facilitating the identification of resilient and affordable systems. The literature contains numerous qualitative descriptions and quantitative methodologies describing limited aspects of the SBD process. However, there exist no methodologies enabling the quantitative management of SBD programs throughout the entire product development cycle. This research addresses this knowledge gap by developing the process framework and supporting methodologies guiding product development from initial system concepts to a final design solution. This research provides several new research contributions. First, we provide a comprehensive SBD state-of-practice assessment identifying key knowledge and methodology gaps. Second, we demonstrate the physical implementation of the integrated analytics framework in a model-based engineering environment. Third, we develop a quantitative methodology enabling program management decision making in SBD. Fourth, we describe a supporting uncertainty reduction methodology using multiobjective value of information analysis to assess design set maturity and higher-resolution model usefulness. Finally, we describe a quantitative SBD process framework enabling sequential design maturation and uncertainty reduction decisions. Using an unmanned aerial vehicle case study, we demonstrate our methodology’s ability to resolve uncertainty and converge a complex design space onto a set of resilient and affordable design solutions.
Conference Paper
There is an increasing supply of obsolete buildings across the U.S. that have surpassed their useful life. These buildings become obsolete, either because of aging systems or shifts in market demand for the services and amenities the building provides. As a result, obsolete buildings have a higher cost of ownership and can have a negative effect on occupant productivity and morale. In the manufacturing sector, the rate of building obsolescence is especially high. Because the buildings in this sector are typically customized for a specific product or manufacturing process, they struggle to accommodate change over their lifetime. One solution to this problem is to make the buildings themselves more flexible. This solution requires an understanding of the types of uncertainty facing the project during design, construction, and operation, as well as the potential options that may become available to the owner if modifications to the building are necessary. This paper presents the results of a real options analysis (ROA) to valuate flexible building strategies in the manufacturing sector, where uncertainty is high. Specifically, we consider three strategies—dedicated, general purpose, and scalable—and develop real options for each strategy related to the variable capacity of the project, project life, and timing of project construction or renovation. We demonstrate that certain strategies ultimately offer owners of manufacturing buildings a higher long-term value when considering uncertainty about the future demand of their products. By recognizing this uncertainty and evaluating potential design strategies with ROA, this research provides decision guidance to owners that will increase use of flexible design strategies and reduce instances of building obsolescence.
Article
Full-text available
Set-based design (SBD), sometimes referred to as set-based concurrent engineering (SBCE), has emerged as an important component of lean product development (LPD) with all researchers describing it as a core enabler of LPD. Research has explored the principles underlying LPD and SBCE, but methodologies for the practical implementation need to be better understood. A review of SBD is performed in this article in order to discover and analyse the key aspects to consider when developing a model and methodology to transition to SBCE. The publications are classified according to a new framework, which allows us to map the topology of the relevant SBD literature from two perspectives: the research paradigms and the coverage of the generic creative design process (Formulation–Synthesis–Analysis–Evaluation–Documentation–Reformulation). It is found that SBD has a relatively low theoretical development, but there is a steady increase in the diversity of contributions. The literature abounds with methods, guidelines and tools to implement SBCE, but they rarely rely on a model that is in the continuum of a design process model, product model or knowledge-based model with the aim of federating the three Ps (People–Product–Process) towards SBCE and LPD in traditional industrial contexts.
Article
Increasing system complexity has provided the impetus to develop new and novel systems engineering methodologies. One of these methodologies is set-based design (SBD), a concurrent design methodology well suited for complex systems subject to significant uncertainty. Since the 1990s, numerous private, public, and defense sector design programs have successfully implemented SBD. However, concerns regarding SBD's complexity, tendency towards qualitative methods, and lack of quantitative tools have limited its use. To address these issues, our research surveys 122 refereed journal articles and conference papers to assess SBD's state-of-practice and identify relevant research opportunities. To accomplish these tasks, we perform a structured literature review to identify and assess relevant and influential research. We found that SBD's state-of-practice relies heavily upon decision and tradespace analysis with increasing emphasis on uncertainty modeling and MBSE. We found that the majority of SBD research consists of quantitative methodologies focusing on component and small system applications. We also found that complex system applications used mostly qualitative methodologies. We identify SBD research opportunities for requirements development, MBSE, uncertainty modeling, multiresolution modeling, adversarial analysis, and program management. Finally, we recommend the development of a comprehensive SBD methodology and toolkit, suited for complex system design across all stages of the product development life cycle.
Article
Full-text available
How Toyota's product design and development process helps find the best solutions and develop successful products.
Article
Fully updated and revised by international authorities on the topic, this new version of a classic and established text returns to its roots as a clear and concise introduction to this complex but essential topic in corporate finance.
Article
Program overruns, contract disputes, and legal confrontation between defense contractors and the government escalated seriously over the 1970's. The author led the development and application of a computer simulation model to resolve a $500 million shipbuilder claim against the US Navy. By using the model to diagnose the causes of cost and schedule overruns on two multibillion-dollar shipbuilding programs. Ingalls Shipbuilding quantified the costs of disruption stemming from Navy-responsible delays and design changes; in June 1978, the Navy agreed out of court to pay $447 million of the claim. Use of the model (which was the basis for at least $200–300 million of the settlement) broke new legal ground, providing the defense and legal communities with a means by which adversary relationships can be avoided and equitable settlements of contract cost disputes achieved. Ingalls Shipbuilding (a division of Litton Industries, Inc.) now has extended the model to aid strategic decision making in managing its shipyard operations. Each phase of several shipbuilding programs—acquisition and utilization of manpower, scheduling and performance of work, and managerial decisions throughout the program—can be accurately simulated. Executives find it valuable as a test bed for evaluating the consequences of alternative policies in bidding and marketing, contract management, program work scheduling, resource management, and cost forecasting. Also, it provides a needed technique in the avoidance of contractor claims.
Article
A five-year study reveals that the highly successful Toyota Motor Corporation seems to follow a different paradigm of design than other US and Japanese auto companies. This paper outlines 11 principles that appear to form the foundation of Toyota's use of "Set-Based Concurrent Engineering." Discussion of the principles includes illustrations from Toyota.