Managing the Simultaneous Execution of Coupled Phases in

advertisement
IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, VOL. 43, NO. 2, MAY 1996
210
ltaneous Exec
Viswanathan Knshnan
Abstruct- Concurrent engineering (CE) calls for the simultaneous execution of coupled product development phases. One
approach to simultaneous execution involves performing a downstream information absorbing product development phase concurrently with an information supplying upstream phase. However, such simultaneous execution of coupled phases, in the
absence of careful management, can lead to substantial deterioration in the product development performance. In this paper, we
present the managerial implications of a model-based framework
to manage the risks involving the simultaneous execution of
coupled development phases. We describe how the framework
would apply to an automobile instrument panel (Up) development
process, and discuss the changes in organizational perspective
required for successful simultaneous product development.
I. INTRODUCTION
EVELOPING and delivering a complex, technological
product to the market requires the completion of hundreds of closely coupled tasks, grouped into phases, through
which the concept, configuration and other technical details of
the product are generated, narrowed, and finalized. The couplings and information dependencies among product development phases are manifested in the exchange of product-specific
information between development teams-for
example, the
transfer of customer needs and preferences from the marketing to the design group which helps designers generate and
finalize the product concept and detailed design. In fact this
information dependency among phases was one of the reasons
for the traditional, sequential “pass-the-baton’’ execution of
phases [26];the information-absorbing downstream development phases waited to begin until the required information was
“fully” available or until the information-generating upstream
phase was completed. Breaking the sequential mind set and
pattern of execution therefore requires that the downstream
phases be able to operate concurrently with the upstream
phases by using early upstream information.
Beginning downstream development phases before the product or market information they require is available in a
finalized form (so that the upstream and downstream phases
are simultaneously executed) is inherently risky; changes in the
exchanged information may have to be incorporated through
additional work in the downstream phase [13], [19]. If the
concurrent process is not properly managed, significant deterioration in product development performance can ensue due
Manuscript received October 3 , 1994. Review of this manauscript was
arranged by Guest Editor G. Susman. This work was supported in part by the
Massachusetts Institute of Technology Leaders for Manufacturing Program.
The author is with the College of Business Administration, University of
Texas, Austin, TX 78712-1 174 USA.
Publisher Item Identifier S 0018-9391(96)04877-5.
to the expensive and/or time consuming downstream rework
required to absorb the changes in the upstream-generated
information. These risks involved in the concurrent execution
of coupled product development phases are exemplified by the
process of development of instrument panels (Up) at one of our
study companies-a U.S. auto maker.
A. Risks Facing Simultaneous Development:
Instrument Panel Example
The ilp of an automobile consists of components such as the
steering column, instrument cluster, air bags, base panel, trim
panel, and the glove box. Fig. 1 shows the base panel-one
of the most time consuming components to develop in the
ilp system because: 1) almost all other components including
the steering column, instrument cluster, air bags, trim panel,
and the glove box mount onto the base panel (changes in
any of these components cause changes in the base panel);
and 2) the design of the base panel itself is technically
challenging because it is the load-bearing component in front
collisions, and it has many intricate formations to house other
components.
A study of the development process of a base panel reveals
that there are two lengthy phases in the critical path of
the process, called the “design phase” and the “mockup
construction phase.” During the design phase, which lasts 10
weeks, the spaces, clearances, and the attachment schemes for
each of the ilp components are defined. Mockup construction, which takes 15 weeks, involves the development and
assembly of fiberglass parts from a wooden mold of the base
panel using the engineering drawings of the base panel. In a
sequential process, therefore, both stages together would take
25 weeks. Using our previous terminology, the downstream
phase (mockup construction) may not begin until it receives
finalized information (panel engineering drawings) from the
upstream phase (see the base panel design in Fig. 2).
In the highly competitive automotive industry, where development lead times are constantly shrinking [SI, it is very
desirable to expedite the completion of the base panel development process. One approach to accelerate this process is to
seek to overlap in time the base panel design and prototyping
processes so that the simultaneous development time can help
reduce the base panel lead time. This, however, means that
mockup construction begins with unfinalized drawings, and
changes in the design be accommodated as the construction
progresses. In the development process of the i/p, it is quite
possible for the design changes to be significant enough to
cause substantial rework of the mockup. Because of the tight
coupling among the i/p components, a small change in the
0018-9391/96$05,00 0 1996 IEEE
KRISHNAN: MANAGING THE SIMULTANEOUS EXECUTION OF COUPLED PHASES IN CONCURRENT PRODUCT DEVELOPMENT
21 1
Fig. 1. The base panel of an automobile.
design information about one of the components can lead to
a “stackup” that would require a major revision of the base
panel design. In fact, overlapping mockup construction with
base panel design, amidst significant high impact changes
in the base panel design information, can increase both the
duration of the mockup construction phase by as much as 10
weeks, and the development cost by hundreds of thousands
of dollars-without leading to significant lead time savings in
the base panel development process.
11. A FRAMEWORK
TO ENABLE
THE
CONCURRENT EXECUTION
OF COUPLED PHASES
The example presented in Section I illustrates the risks
involved in simultaneously executing coupled phases. To be
successful in concurrent product development requires not
only the early critique of upstream decisions by downstream
personnel, but also the careful overlapping of the development
phases-which forms the focus of this paper as it involves
an in depth understanding of the phase couplings. In an
earlier paper [17], we introduced a mathematical model of
overlapping based on two aspects of the coupling between the
phases, called upstream information evolution and downstream
sensitivity to changes in the exchanged information. In a
subsequent paper [18], we further enhanced the model to cover
different types of overlapping, and developed a conceptual
framework to capture the appropriateness of different types
of overlapping. The contribution of this paper is to discuss
the managerial implications of the model-based framework,
illustrate its applicability to i/p development, and provide
insights on the organizational changes needed to facilitate
successful concurrent development. We begin by describing
upstream evolution, downstream sensitivity, and the conceptual framework.
A. Upstream Information Evolution and
Downstream Sensitivity
From the i/p example, we see that it would be risky to
start downstream phases with upstream information which is
1.5 wppks
ckup Construction1
Fig. 2. The blase panel design and prototyping in a sequential process.
preliminary (or not yet attained its final form). Nor would
it be advisable to freeze the upstream design information
early, just for the sake of passing it downstream, without
knowing how closely the product information meets technical
and market specifications. The notion of upstream information
evolution is introduced to refer to the rate at which the
exchanged information reaches its final form. Evolution of
upstream generated information is fast when the information
gets close I O its final form rapidly, and is capable of being
frozen and passed downstream early in the upstream process
without much penalty for the upstream phase (see Fig. 3).’
The evolution is said to be slow, however, if finalizing product
information early in the upstream process is either impossible
or involves a huge quality penalty for the upstream phase. Such
penalty may ensue when the various quality dimensions (such
as performance, aesthetics. and reliability) are at lower than
desired levels at the time of freeze. This paper is restricted
to the concepts and their managerial implications, and the
mathematical models presented in [ 181.
It is noteworthy that evolution refers to the time availability
of the upstream information-information that evolves faster
is available at a given level of completeness earlier to the
downstream phase than information that evolves slowly. It
is also interesting to note that the amount of change the
exchanged information undergoes is a function of its evolution
for processes that enjoy strictly nondecreasing evolution [ 171;
‘Degree of evolution measures how close the unfinalized upstream information is to its final form. Mathematical definitions presented in [18] normalize
degree of evolution such that it is zero at the beginning of the upstream phase,
and builds up to a value of one at the end of the upstream phase.
IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, VOL. 43, NO. 2, MAY 1996
212
Degree t o which exchanged
information is in final form
Downstream
Iteration
Duration
t
sensitivity case
Magnitude of Change in
Exchanged Information
Beginning of
Upstream Task
Time
End of
Upstream Task
Fig. 3. Upstream information evolution at the fast and slow extremes.
for such processes, the faster the evolution, the smaller the
amount of change exchanged information undergoes at the
end of the upstream phase.
In the concurrent execution of product development phases,
changes in the exchanged information occurring during the
earlier period of the upstream phase have different implications
for the downstream phase than those happening near the
later period of the upstream phase. If large changes were to
happen until the completion of the upstream phase and if the
downstream phase were to be conducted simultaneously, then
substantial rework is needed to accommodate these changes.
If the changes in the exchanged information are, however,
such that major changes in the exchanged information happen
during the initial period of the upstream phase, and only
minor changes occur near its completion, then simultaneous
execution is much more likely to be able to reduce the
development lead time. Thus we note that simultaneous execution is more viable when the evolution of the upstream
design information is fast (or when only minor changes in
the exchanged information occur near the completion of the
exchanged phase) than when it is slow.
Evolution of the upstream information describes whether
the exchanged parameter undergoes small or large changes
near the finish of the upstream phase. The consequence of
the upstream changes to the downstream phase is captured
by downstream sensitivity, which measures the duration of
downstream work required to accommodate changes in the
upstream information (see Fig. 4). Downstream phases are
highly sensitive when the phases are so closely coupled that the
downstream work required to incorporate even small changes
in the upstream information is large. The sensitivity may also
be high to changes in decisions that are strategic in nature
or fall at a relatively high level in the design hierarchy [7].
By decoupling the phases using design techniques such as
product modularity [22],and by enriching the communication
between the phases, it may be possible to keep the downstream sensitivity low. For example, it may be possible for
downstream designers to make design decisions such that the
choices made are robust against variations in the upstream
information [25]. Anticipation by downstream designers for
Fig. 4. Downstream sensitivity at the high and low extremes.
changes in the upstream information can also help reduce the
amount of downstream work required to absorb changes in
the upstream information [101.
111. A CONCEPTUAL
FRAMEWORK
BASEDON EVOLUTION
AND SENSITIVITY
From the i/p example, we note that while offering the potential of reduced development time, concurrent execution of
coupled phases runs the risk of increased downstream rework.
The conceptual framework, introduced in [ 181 and summarized
in this paper, helps manage this risk and answer questions
such as when to use preliminary upstream information for
downstream development, and how the interaction between
the phases must be managed.
As shown in Fig. 5 , four extreme situations are likely
to arise-when the upstream evolution is fast or slow and
when the downstream sensitivity is high or low. For each of
these combinations of evolution and sensitivity, the interaction
between the phases for concurrent development needs to
be managed differently as described by different types of
overlapping below.
1 ) Itemtive Overlapping: When the downstream phase
sensitivity is low, preliminary upstream information can
be processed by the downstream phase without much risk
of significant rework. Even if changes in the exchanged
information are large, their effects on the downstream
phase are not likely to be substantial. The upstream
product information, however, evolves slowly-it
cannot
be finalized until late into the upstream process. The
interaction between the phases in this case is called iterative
overlapping where the downstream phase begins early with
preliminary information, and design changes are incorporated
in subsequent downstream iterations. In this case, the upstream
design information is not finalized until the completion of the
upstream phase, as doing so may result in a large quality
penalty for the upstream phase.
2) Preemptive Overlapping: The opposite case to iterative
overlapping occurs when the downstream phase sensitivity to
changes in exchanged information is high, and the upstream
information evolves fast. In such a case, the exchanged information is to be precipitated to its final value/form at an earlier
point in time. In other words, the upstream problem solving
is accelerated and information frozen ahead of the normal
KRISHNAN: MANAGING THE SIMULTANEOUS EXECUTION OF COUPLED PHASES IN CONCURRENT PRODUCT DEVELOPMENT
Degree of
Evolution
Increase in &wnstream
Duration
Low Sensitivity
Design
hange
lnmease in h m : j t r e a m
Duration
High Sensitivity Case
Design
Change
Fig. 5
213
Degree of
Evolution
Iterative Overlapping
Concurrency by exchange
of preliminary p r o d u c t
information
D
m
Poor prospects for
g
-
Distributive Overlaming
B’oth preliminary
information exchange and
early finalization lead to
simultaneity
--preemptive O v e r l a p p i n g
Concurrency by Early
concurrency. I n f o r m a t i o n Finalization of upstream
is disaggregated to
iin f o rma t ion.
improve prospects.
The framework for concurrcnt development. (Adapted from [lS].)
time of freeze. This is called preemptive overlapping (because
iterations are preempted by finalizing upstream information
early), and can help reduce development time by starting
the downstream phase earlier-but with jinalized upstream
information. Note that there are no subsequent downstream
iterations. It may result in some quality loss for the upstream
phase due to its losing the ability to make changes up to
its original completion time. The penalty for preemptive
overlapping may be low, however, when due to the short lives
of product generations within a product type it is better to get
to the market quickly, assess customer reactions, and follow-up
rapidly with improved product versions.
3) Distributive Overlapping: Consider the case when both
the upstream information evolves f i s t and the downstream
phase sensitivity is low. In such a case, it is possible to
both start downstream phase with preliminary information
and finalize the exchanged upstream information early. This
situation is called distributive overlapping, because the impact of overlapping is distributed between the upstream and
downstream phases. It is noteworthy that distributive overlapping represents a combination of iterative and preemptive
overlapping approaches, and can be useful in situations where
multiple parameters are exchanged between the development
phases, with different evolution and sensitivity characteristics.
4 ) Divisive Overlapping: The last scenario occurs when
the downstream phase sensitivity is high and the upstream evolution is slow. Here, it is neither desirable to start downstream
phase with preliminary information (because incorporating
changes can be expensive) nor feasible to precipitate the exchanged information to its final form at an earlier point in time.
In such a case, the exchanged information is disaggregated
into components to see if any of the components evolve faster
or if transferring any of the components in their preliminary
form to the downstream phase is practical. Often the evolution
and sensitivity of the components may be different from
the aggregated information. Because the disaggregation of
exchanged information is typically based on physical or functional division of the upstream and downstream phases (as will
be illustrated later with the i/p example), this approach is called
divisive overlapping. If neither of the parts evolve quickly,
nor can they be used by the downstream phase in preliminary
form, then it is preferable that the phases not be simultaneously
executed with existing evolution and sensitivities.
A. Application of the Framework to Instrument
Panel Development
Earlier we briefly discussed the risks involved in the simultaneous execution of base panel design and prototyping
(“mockup construction”) phases. We will now illustrate how
the conceptual framework developed above may be applied to
the base panel development process.
As discussed earlier, the (upstream) base panel information
changes significantly until the very end of the design phase,
so the base panel design may be said to evolve slowly. Data
collected from the craftsmen who construct the mockups in
our study company indicates that changes made in the base
panel will have an enormous effect, (some times requiring
that the prototype be started all over again), thereby making
the “mockulp construction” a highly sensitive phase to changes
in the base panel design information. Thus the phase interface falls under the slow evolution-high sensitivity category
(the lower left quadrant). The framework suggests that the
exchanged iinformation be disaggregated. We examine if parts
of the base panel design information differ in evolution or
sensitivity by dividing the base panel into the driver, center
and passenger sections, respectively.
Interviews with the professionals who are involved in constructing the mockups indicates that mockup construction
phase is highly sensitive to changes in the design information
about all the three sections-so disaggregation does not alter
the sensitivity of the downstream prototyping phase to changes
in the exchanged information. However, field study of the base
214
IEEE TR4NSACTIONS ON ENGINEERING MANAGEMENT, VOL. 43, NO. 2, MAY 1996
B. Pe$ormance Trude-offs for Different Types of Overlapping
Fig. 6. The simultaneous execution of base panel design and mockup
construction.
panel development process shows that the different sections
evolve at different rates; the passenger section (right side)
of the base panel evolves relatively faster than the driver
(left) and center sections as it does not interface with many
changing components. The steering column and instrument
clusters, which are change-prone during the design process,
drive the changes in the driver section to which they mount.
The changes in the location of the ducting for heat, ventilation,
and air conditioning cause changes in the center section. In
comparison to the other two sections, the passenger section
of the base panel is relatively free of change as the design
process progresses, and can afford to be frozen early.
With disaggregation, we find that the design and mockup
stages exchange the design information about the (relatively)
fast evolving base panel passenger section, and the slow
evolving center and driver sections (all of which are of high
sensitivity). The framework suggests that the right section be
frozen early and passed downstream (the design and mockup
stages be preemptively overlapped as Fig. 6). In the studied
process, such a disaggregation was found to have the potential to save two to three weeks in development time. It is
noteworthy that such disaggregation of product information
was possible because of the relatively weak interdependence
between the different sections of the base panel. If it is
not possible to disaggregate the exchanged information into
relatively weakly coupled parts with different patterns of
evolution and/or sensitivity, it may not be possible to overlap
the phases with the existing design concept and configuration
choices.
The ilp development process illustrates the way in which
divisive and preemptive overlapping help manage the risk associated with overlapping the base panel development phases.
Iterative overlapping is illustrated with an automobile door
development process example in [ 171 where the sensitivity
data is highly nonlinear in n a t u r e 4 h a n g e s above a certain
point cause significantly larger iterations than changes below
that point. Iterative overlapping is made possible by identifying
a point in the design process when the preliminary door design
information is sufficiently well evolved, and can be put to use
by downstream developers without the risk of further changes
causing large downstream iterations.
An example of distributive overlapping is the pager development process at Motorola [ 181, where the pager dimensions fall
under the fast evolution-high sensitivity category, and shape
details fall under the slow evolution, low sensitivity category.
Industrial Design and Engineering Design phases are overlapped by exchanging preliminary information about the shape
details, and early commitment to the dimensions-thereby
exemplifying distributive overlapping.
Note that the four different types of interactions described
above result in different trade-offs among the performance
parameters. In iterative overlapping for instance, increased
downstream effort (required to incorporate upstream design
changes) is associated with increased development cost and
is traded-off against lead time savings, while in preemptive
overlapping, upstream quality loss (due to early finalization
of upstream information) is traded-off against savings in lead
time due to concurrent execution.
The concepts of upstream evolution and downstream sensitivity, described above, have been formulated rigorously
[18] leading to a mathematical model that helps schedule
the coupled development phases for concurrent execution.
Such a model helps decide when to begin the downstream
information-absorbing phase, how many iterations to perform,
and when to start each of these individual iterations.
Iv. ASSESSMENTAND MANAGEMENT
OF EVOLUTION
AND SENSITIVITY
In the above discussion, we have focused mainly on how the
evolution and sensitivity information may be used to determine
the pattern of overlapping among phases. In what follows,
we discuss how evolution and sensitivity may be assessed
and modified to the firm’s benefit in specific engineering
management situations, and why there exists differences in
the evolution and sensitivity of different products.
In practice, evolution and sensitivity functions may be
obtained as input from product development professionals
based on their experience. This is especially true for redesigned
products, which form a significant portion of products designed
in many companies [20]. The evolution data can be constructed
based on how much the intermediate upstream phase tasks help
refine the exchanged design parameter [ 181. The sensitivity
data may be obtained by examining the impact of upstream
design changes on the downstream phase. Researchers have
observed that the process of designing a new product is
often not radically different from that of an earlier generation
product [4], [20]. In such cases, the documentation and understanding of the established process can be used to describe the
evolution and sensitivity functions.
If the change in the development process is substantial or
the product is entirely new, then forecasts of evolution and
sensitivity need to be made based on the particular design
concept, product architecture, and underlying technologies. It
is noteworthy that the conceptual framework presented in the
previous section does not require detailed functional forms of
the evolution and sensitivity but merely qualitative data as
to whether evolution is fast or slow and sensitivity is high
or low. Yet we recognize that the ability to assess evolution
and sensitivity will be lower for such highly unpredictable
processes.
Different processes differ in their evolution because customer preferences and market conditions related to some
products are much more time varying and uncertain than others. In a highly competitive market for an evolutionary product
(which undergoes gradual improvements in performance and
KRISHNAN: MANAClNG THE SIMULTANEOUS EXECUTION OF COUPLED PHASES IN CONCURRENT PRODUCT DEVELOPMENT
size but minor changes in the concept), many of the product’s
specifications get narrowed to their final value rather quickly.
These specifications that are invariant are tightly constrained
by factors such as the characteristics of the target market (e.g.,
infrastructure), core technology or industry standards which
makes their evolution rapid.’ In the case of a revolutionary
product (such as HDTV), however, the product specifications
undergo substantial revisions during the design process due to
the lack of many constraints (such as industry standards or
established technologies), making the evolution of the product
relatively slow. There are also intrafirm factors such as the
rate of design problem solving (the rate at which the upstream
designers attain convergence) that decide the rate of upstream
evolution.
Sensitivity of the downstream phases to changes in exchanged information, however, appears to be very much
dependent on internal factors such as the communication
among engineers, anticipation of the downstream engineers,
and the flexibility of the development methodologies used.
With good communication and anticipation, and a flexible
design and prototyping process, the sensitivity of downstream
phases to changes in exchanged information can be kept
low. This leads us to the interesting idea that evolution and
sensitivity, which are determinants of the pattern of concurrent
execution, can to some extent be altered themselves. As we
observed in our study of the pager development process at
Motorola, effective communication between the upstream and
downstream product developers helped reduce the downstream
sensitivity [ 181. Also, computer tools and better design practice
can ensure that the upstream design information evolves faster
and the effects of changes on the downstream phase are
reduced.
The evolution-sensitivity map in Fig. 7 shows the desirable
direction of movement for a development process; from the
point of view of overlapping, it is desirable that the evolution
of the design information become faster and downstream
phase sensitivity lower (move upwards and rightward on the
map). Once again it needs to be cautioned that for processes
whose evolution and sensitivity are hard to estimate, the
practical ability to modify evolution and sensitivity is limited,
and attempt should be made to understand the pattern of
evolution of the upstream information and the sensitivity of
the downstream tasks.
v.
CHANCES IN
MANAGERIAL
PERSPECTIVE FOR CE
The conceptual framework presented above and the mathematical models underlying the framework offer engineering
managers rough guidelines on when and how to simultaneously execute coupled phases. However, these guidelines
cannot in isolation help in the successful implementation of
concurrent development. Prior to using the framework, it is
essential that the managerial perspective that contributed to
the sequential mind set be revisited.
2Tn some industries, such as consumer electronics, key product specifications are severely constrained by existing products in the market. (e.g., the
size of a new camcorder may be no larger than an existing product.)
Slow
215
Fast
Evolution
.,ne*k
t-
Design-Practice,
Better Communication
Sensitivity
Fig. 7. The evolution-sensitivity map
In the conventional sequential mode of project execution, it
was assumed that the product design information was available
for exchange, only in its finalized form, and that too at
the completion of the information-generating phase-leading
to the much written about “throw-over-the-wall” mode of
developmenl. To break out of this mode, managers need to
recognize the availability and utility of preliminary product
information which may be exchanged at intermediate points
in the execution of generating phases. However, using preliminary information requires that changes be incorporated
in future iterations which further need to be planned in the
development process.
Rather than view development iterations as “unavoidable
evils” that lead to cost escalations and arise out of designer errors, managers may want to also leverage the beneficial aspects
of iterations--especially their potential to enable simultaneous
development by absorbing preliminary information in earlier
iterations. Performing multiple iterations can also enable the
downstream phase to offer early feedback to its upstream
counterpart, thereby helping to improve the design quality.
Other changes in managerial perspective that are required
for successfiul simultaneous development are listed in Table 1.
One of the contributors to the sequential mind set is the
application of the traditional network project management
approaches [6], as the interrelationships between phases in
these tools are described by constraints relating only a phase’s
start and finiish events (no intermediate events are recognized).
In other words, the project management paradigm unintentionally contributes to the notion of one-shot transfer of finalized
information (at start and finish times). As we have seen in
this paper, that would be required only when the upstream
evolution is slow and downstream phase sensitivity is high.
By considering three other combinations of evolution and
sensitivity, the above described framework represents one
of the points of departures from the project management
paradigm [ 111.
Implementing these changes in perspective requires fundamental changes in the organizational structure and measurement systems. The barriers posed to successful simultaneous
development by a traditional hierarchical and/or functional
IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, VOL. 43, NO. 2, MAY 1996
216
From a Sequential Engineering View
Design information is transferred only once-in finalized form.
Product information is available for exchange only at the finish of the
generating phase.
Phase durations are constants, as phases do not consist of any planned
iterations.
Intermediate points in the evolution of product information are not
recognized.
The effect of changes in the exchanged information on the downstream
phase is ignored.
The communication between the upstream and downstream phases is
minimal with a one-time “dump” of finalized information by the upstream
nhase.
organizational structure has been widely written about in
the research literature [5], [ll], [12]; many firms also continue to use outdated measurement systems that discourage
concurrent product development, as illustrated by one company we studied, which had made a significant investment
on implementing “concurrent development” by collocating
product developers and forming cross functional teams. The
development managers in this company, however, continued to
be evaluated only on the basis of the development expense for
their engineering function. As we discussed earlier, concurrent
product development may require resorting to more iterations
in the interest of absorbing preliminary information; more
downstream iterations may translate into an increase in cost
in order to obtain the lead time savings. Because the measurement system did not take into consideration savings in lead
time achieved by concurrent development (rewards were only
on the basis of development expense), managers continued
to emphasize costs and waited to begin development until
upstream designers “released” finalized information. Despite
the formation of teams, the development process in this
firm continued to resemble the “pass the baton” sequential
information hand-over process.
VI. CONCLUSIONS AND FUTURE
WORK
This paper presents the managerial implications of a conceptual framework to enable the simultaneous execution of
coupled phases in industrial product development. The framework is based on the nature of the coupling between phases
described by the concepts of upstream evolution and downstream sensitivity, and offered different ways to overlap the
coupled phases for concurrent development. Existing work
on managing CE has focused on incorporating downstream
feedback in the upstream decision making process [ 141. However, in order for the downstream phase to be able to provide
feedback before an upstream decision is made, the phases need
to be able to exchange and process unfinalized information,
which involves considerable risks. Besides highlighting the
risks involved in the concurrent execution of coupled phases
(using the i/p example), this paper offered some guidelines for
managing this risky process. We observed that simultaneity
can be achieved by preliminary product information exchange
when the downstream phase sensitivity is low, and by early fi-
To a CE View
Information may be beneficially exchanged many times, well before it is
finalized.
Information exchange may occur at intermediate points in the execution of
generating phases.
Phase durations may vary to incorporate upstream design changes in future
iterations.
Knowledge about upstream evolution is profitably utilized in executing the
phases simultaneously.
Downstream sensitivity to the exchanged information is used to overlap the
coupled phases.
Communication between the phases is frequent and high in bandwidth. Not
only is the content of the information exchanged many times, but also
meta-level details such as evolution and sensitivity.
nalization of product information when the upstream evolution
is fast (with different types of performance trade-offs).
Managing the interaction between phases for simultaneous
execution using the concepts of evolution and sensitivity
corresponds to the detailed management of CE, but is by
itself not sufficient to successfully implement CE. Proper
managerial perspective, conducive organizational structure,
and measurement systems are equally, if not more, important.
A (nonexhaustive) list of changes in the manager’s perspective
of the engineering process essential for making the transition
from sequential to CE was presented. This list contained,
among others, a recommendation to expand the communication among developers to include metalevel information such
as evolution and sensitivity apart from frequent exchange of
the information itself.
The work presented in this paper was focused on modeling
the coupling between two product development phases, and
must be extended to cover the interactions among multiple
product development phases. Though the evolution-sensitivity
based framework does not require detailed quantitative data, its
applicability would be limited when the ability to assess even
qualitative evolution and sensitivity data is low-as might be
the case for a breakthrough product being developed by a
new organization. Future work needs to focus on developing
coordination mechanisms for such situations to help address
the risk in simultaneous product development.
This paper builds on much existing research in the area of
product development. Interdependence among phases, which
makes it difficult to overlap the phases, has been discussed
in considerable detail by Thompson [27]. Thompson classified
interdependence into sequential, pooled, and reciprocal interdependence. The framework presented in this paper pertains
more to the concurrent execution of phases which are sequentially interdependent upstream and downstream phases. When
one considers concurrent execution of tasks within phases,
one may experience reciprocal interdependence Le., the tasks
may be connected in a cyclic manner. For such situations, it
may be necessary to consider both upstream and downstream
evolutions and sensitivities to determine how to overlap the
phases.
Several researchers have also noted the informationintensive nature of the task of managing product development
KRISHNAN: MANAGING THE SlMULTANEOUS EXECUTION OF COUPLED PHASES IN CONCURRENT PRODUCT DEVELOPMENT
[9], [16], [24], [28], and have studied the utility of design
tools to perform the information processing functions.
Rosenthal and Tatikonda [21] have noted the criticality of
timing of access to (design) information to new product
development, and identified communication acceleration as a
key information-processing function that needs the application
of design tools. The concepts of evolution and sensitivity
developed in this paper should help address the “element of
risk” in communication acceleration they caution managers
against. Adler proposes an interesting causal model relating
(product development phase) interdependence, coordination
mechanisms, outcome, and managerial action [2], [3]. He
suggests that based on how effective (defective) the outcome
is, managers seek to modify the coordination mechanisms
or the pattern of interdependence itself. In this context,
the concept of evolution and sensitivity can be thought
of as modeling the interdependence, and using them to
overlap phases as a mechanism to coordinate the phases.
When managers seek to alter the evolution and sensitivity
themselves (as in Fig. 7), they are in the act of modifying the
interdependence itself-in line with Adler’s causal model.
Several authors have pointed out that faster development
processes are more overlapped in time [IO], [151, [23], [26]. To
achieve time overlapping, Clark and Fujimoto recommended
frequent, face-to-face, bilateral communication of preliminary
information instead of late release of complete information.
This work offers 1) a more detailed description of the characteristics of the information exchanged to operationalize
overlapping, and 2) a method to map product development
processes (based on evolution and sensitivity) which will be
useful in practice both to evaluate existing process capabilities,
and to transform product development processes into more
effective ones.
ACKNOWLEDGMENT
The author benefited greatly from discussions with S. D.
Eppinger and D. E. Whitney at the Massachusetts Institute of
Technology (MIT) for developing the ideas contained in this
paper.
REFERENCES
P. S. Adler, A. Mandelbaum, E. Schwerer, and V. Nguyen, “From
project to process management: An empirically-based framework for
analyzing product development time,” Manage. Sei., vol. 41, no. 3, pp.
458484, 199.5.
P. S. Adler, “Managing DFM: Leaming to coordinate product and
process design,” in Integrating Design and Manufacturing For Competitive Advantage, C. I. Susman, Ed. New York: Oxford, 1992.
-, “Interdepartmental interdependence and coordination: The case
of the design/manufacturing interface,” Organizat. Sei., vol. 6, no. 2,
pp. 147-167, 1995.
R. E. Albano and J. P. Keska, “Is design realization a process? A case
study,” in 7th Int. IEEE Elec. Manufact. Tech. Symp., 1989.
D. G. Ancona and D. E. Caldwell, “Cross-functional teams: Blessing or
curse for new product development,” in Transforming Organizations, T.
Kochan, Ed. New York Oxford, 1991.
J. D. Blackburn, “New product development: The new time wars,” in
Time-Based Competition. Homewood, IL: Business One Irwin, 1991.
217
[7] K. B. Clark, “The interaction of design hierarchies and market concepts
in technological evolution,” Res. Policy, vol. 14, pp. 235-251, 1985.
[8] K. B. Clark, B. Chew, and T. Fujimoto, “Product development in the
world auto industry: Strategy, organization and performance,” Brookings
Papers on Economic Activity, vol. 3 , pp. 729-771, 1987.
[9] K. B. Clark and T. Fujimoto, “Overlapping problem solving in product
development,” Harvard Bus. Sch. Working Paper No. 87-048, 1987.
[ 101 __, PayoductDevelopment Performance: Strategy, Organization, and
Management in the World Auto Industry. Boston: Harvard Bus. Sch.
Press, 1991.
[ l l ] D. Clausing, Total Quality Development: A Step-by-step Guide to World
Class Concurrent Engineering. New York: ASME, 1994.
[12] D. Douglerty, “New products in old organization: The myth of the
better mousetrap in search of the beaten path,” Ph.D. dissertation, MIT,
Cambridge, MA, 1987.
[13] R. M. Eastman, “Engineering information release prior to final design
freeze,” I.EEE Trans. Eng. Manage., vol. EM-27, pp. 3 7 4 1 , May 1980.
[14] A. Ha and E. Porteus, “Optimal timing of reviews in concurrent design
for manufacturability,” Manage. Sci., vol. 41, no. 9, pp. 1431-1447,
1995.
[1S] K. Imai, I. Nonaka, and H. Takeuchi, “Managing the new product
development process: How the Japanese companies learn and unlearn,”
in The Uneasy Alliance, K. B. Clark, R. H. Hayes, and C. Lorenz, Eds.
Boston: Harvard Bus. Sch. Press. 1985.
[16] H. M. Kirandikar, J. Rao, and F. Mistree, “Sequential vs. concurrent formulations for the synthesis of engineering designs,” in ASME Advances
in Design! Automat., 1991.
[17] V. Krishrian, S. D. Eppinger, and D. E. Whitney, “Accelerating product
development by the exchange of preliminary product design information,” ASME J. Mechan. Design, vol. 117, pp. 491498, Dec. 1995.
[ 181 __, “A model-based framework to overlap product development
activities,” Manage. Sei., 1995; Univ. Texas at Austin, Dept. Manage.,
Working Paper No. 95-96-1.
[ 191 E. Mansfield et al., Research and Innovation in the Modern Corporation,
1st ed. New York: W. W. Norton, 1971.
[20] M. E. McGrath, M. T. Anthony, and A. R. Shapiro, Product Development: Success Through Product and Cycle-Time Excellence. Boston:
Butterworth-Heinemann, 1992.
[21] S. R. Rosenthal and M. V. Tatikonda, “Competitive advantage through
design tools and practices,” in Integrating Design and Manufacturing
For Competitive Advantage, G. 1. Susman, Ed. New York: Oxford
Univ. Press, 1992.
[22] G. V. Shirley, “Modular design and the economics of design for manufacturing,” in Integrating Design and Manufacturing for Competitive
Advantage, G. I. Susman, Ed. New York: Oxford, 1992.
[23] P. G. Smith and D. G. Reinertsen, Developing Products in Halfthe Time.
New York: Van Nostrand Reinhold, 1991.
[24] G. I. Susman, Ed., Integrating Design and Manufacturing For Competitive Advantage. New York: Oxford, 1992.
[251 G. Taguchi and D. Clausing, “Robust quality,” Harvard Bus. Rev., pp.
6.5-75, Jan./Feb. 1990.
[26] H. Takeuchi and I. Nonaka, “The new product development game,”
Harvard Bus. Rev., pp. 137-146, Jan./Feb. 1986.
[27] J. D. Thompson, Organizations in Action. New York McGraw-Hill,
1967.
[28] J. M. Utterback, “Innovation in industry and diffusion of technology,”
in Readings in the Management of Innovation, M. Tushman and W. L.
Moor, Eds. Marshfield, MA: Pitman, 1982.
Viswanathan Krishnan received the B.Tech. and
M.E degrees from the Indian Institute of Technology at Madras, India, and Carnegie Mellon University, Pittsburgh, PA, respectively. He received the
Ph.D degree from the Massachu3etts Institute of
Technology, Cambridge, MA
He is currently an Assistant Professor of Management at the University of Texas, Austin, TX
His research focused on the performance improvement of product development, commercialization,
production, and processes.
Download