1. INTRODUCTION
Today's highly competitive global marketplace
is redefining the way many companies do business. The new form of
competitive advantage is mass customization, and is, as Pine
(1993a, p. xiii) says, “a new
way of viewing business competition, one that makes the identification
and fulfillment of the wants and needs of individual customers
paramount without sacrificing efficiency, effectiveness, and low
costs.” In his seminal text on mass customization, Pine (1993a, p. 6) argues that “customers
can no longer be lumped together in a huge homogeneous market, but are
individuals whose individual wants and needs can be ascertained and
fulfilled.” He attributes the increasing attention on product
variety and customer demand to the saturation of the market and the
need to improve customer satisfaction: new products must be different
from what is already in the market and must meet customer needs more
completely. Sanderson and Uzumeri (1997, p.
3) add that “the emergence of global markets has fundamentally
altered competition as many firms have known it” with the
resulting market dynamics “forcing the compression of product
development times and expansion of product variety.” Findings
from studies of the automotive industry (Womack et
al., 1990; MacDuffie et al., 1996;
Alford et al., 2000) and empirical surveys of
manufacturing firms (Chinnaiah et al., 1998;
Duray et al., 2000) confirm these trends.
Similar themes pervade the text by Wortmann et al. (1997), who examine industry's response in
Europe to the “customer-driven” market.
Because many companies typically design new products one at a time,
Meyer and Lehnerd (1997, p. 2) have found
that the focus on individual customers and products results in “a
failure to embrace commonality, compatibility, standardization, or
modularization among different products or product lines.” Mather
(1995, p. 378) finds that “rarely does
the full spectrum of product offerings get reviewed at one time to
ensure it is optimal for the business.” The end result is a
“mushrooming” or diversification of products and parts that
can overwhelm customers (Stalk & Webber,
1993; Mather, 1995; Huffman & Kahn, 1998). Nissan, for example,
reportedly had 87 different varieties of steering wheels for one of
their cars (Chandler & Williams, 1993).
Although offering a wide variety of products has both positive and
negative effects (cf., Galsworth, 1994; Anderson & Pine, 1997; Ho
& Tang, 1998), the proliferation of product variety can
incur substantial costs within a company (Lancaster,
1990; Child et al., 1991; Ishii, Juengel, & Eubanks, 1995). “The
imperative today,” write Anderson and Pine (1997, p. 3), “is to understand and fulfill
each individual customer's increasingly diverse wants and
needs—while meeting the coequal imperative for achieving low
cost.”
Many companies are using product families and platform-based product
development to provide sufficient variety for the market while
maintaining economies of scale and scope within their manufacturing
processes. In general terms, a product family is a group of related
products that is derived from a product platform to satisfy a variety
of market niches. As Robertson and Ulrich (1998, p. 20) point out, “by sharing
components and production processes across a platform of products,
companies can develop differentiated products efficiently, increase the
flexibility and responsiveness of their manufacturing processes, and
take market share away from competitors that develop only one product
at a time.”
Platform-based product development offers a multitude of benefits
including reduced development time and system complexity, reduced
development and production costs, and improved ability to upgrade
products. Platforms also promote better learning across products and
can reduce testing and certification of complex products such as
aircraft (Sabbagh, 1996), spacecraft (Caffrey et al., 2002c), and aircraft
engines (Rothwell & Gardiner, 1990). In
the automotive industry, platforms enable greater flexibility between
plants and can increase plant usage (sharing underbodies between models
can yield a 50% reduction in capital investment, especially in welding
equipment) and can reduce product lead times by as much as 30% (Muffatto, 1999). Firms using a platform-based
product development approach in the automotive industry recently gained
a 5.1% market share per year whereas firms that did not lost 2.2%
(Cusumano & Nobeoka, 1998).
Methods for platform-based product development have progressed
remarkably in the past decade, and this paper provides a comprehensive
review of the flurry of research activity that has occurred during that
time to facilitate product family design and platform-based product
development for mass customization. In the next section, definitions
and examples of product families and product platforms are given to set
the stage for the discussions that follow. Sections 3 and 4 describe
strategies for leveraging product platforms across different market
segments and metrics for assessing product platforms, respectively.
Sections 5 and 6 review optimization approaches and artificial
intelligence (AI) techniques, respectively, for designing and
configuring families of products. The relationship between product
platforms and mass customization is examined further in Section 7, and
Web-based systems for product platform customization are also
discussed. Finally, closing remarks and avenues of future research are
outlined in Section 8.
2. PRODUCT FAMILIES AND PLATFORMS:
DEFINITIONS AND EXAMPLES
There are two basic approaches to product family design (Simpson, Maier, & Mistree, 2001). The first is
a top-down (proactive platform) approach wherein a company
strategically manages and develops a family of products based on a
product platform and its derivatives. For instance, Sony has
strategically managed the development of its Walkman®
products using carefully designed product platforms and derivatives
(Sanderson & Uzumeri, 1997). Similarly,
Kodak's product platform-based response to Fuji's
introduction of the QuickSnap® single-use camera in 1987
enabled them to develop products faster and more cheaply, allowing them
to regain market share and eventually overtake Fuji (Wheelwright & Clark, 1995). The second is a
bottom-up (reactive redesign) approach, wherein a company
redesigns or consolidates a group of distinct products to standardize
components to improve economies of scale. For example, after working
with individual customers to develop 100+ lighting control products,
Lutron redesigns its product line around 15–20 standard
components that can be configured into the same 100+ models from which
customers could initially choose (Pessina &
Renner, 1998). Black & Decker (Lehnerd,
1987) and John Deere (Shirley, 1990)
have benefited from similar redesign efforts to reduce variety in their
motor and valve lines, respectively.
The key to success in either approach is the product platform from
which the product family is derived. A product platform can be either
narrowly or broadly defined as
- “a set of common components, modules, or parts from which a
stream of derivative products can be efficiently developed and
launched” (Meyer & Lehnerd, 1997,
p. 7);
- “a collection of the common elements, especially the
underlying core technology, implemented across a range of
products” (McGrath, 1995, p. 39);
and
- “the collection of assets [i.e., components, processes,
knoweledge, people and relationships] that are shared by a set of
products” (Robertson & Ulrich,
1998, p. 20).
As an example, a platform at Volkswagen consists of the floor group,
drive system, and running gear, along with the unseen part of the
cockpit as shown in Figure 1. This
platform is shared across several models as well as all of its brands
(i.e., Volkswagen, Audi, Seat, and Skoda). According to Bremmer (1999), Volkswagen owned three of the six automotive
platforms that successfully achieved production volumes over 1 million
in 1999. The number of million-unit platforms is expected to reach 16
by 2004, with Volkswagen leading the way with its A04 and A4/A5
platforms.
Volkswagen's platform definition. Adapted from Wilhelm (1997).
The prominent approach to platform-based product development, be it
top-down or bottom-up, is through the development of a module-based
product family, wherein product family members are instantiated by
adding, substituting, and/or removing one or more functional
modules from the platform. An alternative approach is through the
development of a scale-based product family, wherein one or
more scaling variables are used to “stretch” or
“shrink” the platform in one or more dimensions to satisfy
a variety of market niches. Examples and methods for module-based
product family design are described in the next section, followed by
examples of scale-based product families in Section 2.2.
2.1. Module-based product families
There are numerous examples of module-based product families in the
literature; some of the more frequently quoted examples follow.
- Sony builds all of its Walkmans® around key
modules and platforms and uses modular design and flexible
manufacturing to produce a variety of quality products at low cost,
which allowed them to introduce 250+ models in the United States in the
1980s (Sanderson & Uzumeri,
1997).
- Nippondenso Co. Ltd. makes an array of automotive
components for a variety of automotive manufacturers using a
combinatoric strategy that involves several different modules with
standardized interfaces; for instance, 288 different types of panel
meters can be assembled from 17 standardized subassemblies (Whitney, 1993).
- Hewlett Packard successfully developed several of their
ink jet and laser jet printers around modular components to gain the
benefits of postponing the point of differentiation in their
manufacturing and assembly processes (Feitzinger
& Lee, 1997).
- Bally Engineering Structures offers an almost infinite
variety of environmentally controlled structures that can be readily
assembled from one basic modular component—the pre-engineered
panel—that can be produced in a variety of shapes and sizes and
customized with options, attachments, and finishes to fit into any size
structure (Pine, 1993b).
These successful examples resulted from careful attention to customer
needs and the underlying product architecture in the family. Ulrich
(1995, p. 420) defines the product
architecture as “(1) the arrangement of functional
elements; (2) the mapping from functional
elements to physical components; (3) the
specification of the interfaces among interacting physical
components.” A product architecture is classified as either
modular, if there is a one to one or many to one mapping of
functional elements to physical structures, or integral, if a
complex or coupled mapping of functional elements to physical
structures and/or interfaces exists. For example, personal
computers (PCs) are highly modular. Baldwin and Clark (2000) trace the development of the IBM's
System/360, the first modular computer family. Automotive
architectures, on the other hand, are predominantly integral
(cf. Siddique et al., 1998; Muffatto, 1999), but modularity has become a major
strategic focus for future product development within many automotive
companies (Kobe, 1997; Shimokawa et al., 1997; Cusumano
& Nobeoka, 1998). For instance, Volkswagen's Golf II
comprises several modules to facilitate assembly (see Fig. 2a), and the rolling chassis module produced
by the Dana Corporation (Fig. 2b) saved
DaimlerChrysler nearly $700 million when developing their new Dodge
Dakota facility (Kimberly, 1999). The rolling
chassis module consists of brake, fuel, steering, and exhaust systems;
suspension; and drive-line assembled to the frame. It is the largest,
most complex module provided by a supplier, accounting for 25% of the
vehicle content.
(a) Modules in the Golf II. Adapted from Wilhelm (1997). (b) A rolling chassis module. Adapted from
Kimberly (1999).
Modularity is an important topic in many product design textbooks
(see, e.g., Pahl & Beitz, 1996; Ulrich & Eppinger, 2000; Otto & Wood, 2001) and is the sole focus in
several texts (Ericsson & Erixon, 1999;
O'Grady, 1999; Baldwin & Clark, 2000). Approaches for
developing modular product architectures and module-based product
families abound in the engineering design literature. For instance,
Mattson and Magleby (2001) discuss concept
selection techniques for managing modular product development in the
early stages of design. Wood and his coauthors (McAdams et al., 1999; Stone et
al., 2000b; McAdams & Wood,
2002) present a methodology for representing a functional model
of a product in a quantitative manner to assist in developing product
architectures and facilitate the identification of a core set of
modules for a product family. As part of their work, Stone et al.
(2000a) present a heuristic method
to identify modules for these product architectures; heuristics to
identify functional and variational modules within a product family are
introduced by Zamirowksi and Otto (1999).
Their work is foundational to the methods for developing modular
product architectures developed by Otto and his coauthors (Dahmus et al., 2001; Otto,
2001; Sudjianto & Otto, 2001).
A method for incorporating customer demand into the development of
the modular product architecture is discussed in Yu et al. (1999), and a method for assessing value in a
module-based product family using real options concepts in the presence
of uncertainty has also been developed (Gonzalez–Zugasti et al., 2001). Sundgren
(1999) proposes a method for managing
interfaces between modules within a product family after studying
several product family development projects in the Swedish
manufacturing industry over a period of 3 years; a method for
developing robust interfaces for modular products is also introduced in
Blackenfelt and Sellgren (2000). Comparisons
of methods for modularizing product architectures can be found in Guo
and Gershenson (2003) and Holtta and Salonen
(2003), and researchers have investigated
modular design approaches specifically for electronic products (Tseng & Jiao, 1997b), digital circuits
(Kusiak & Huang, 1997), and mechatronics
products (Huang & Kusiak, 1999). Finally,
Schilling (2000) is developing a general
theory of modular systems based on causal models developed from
studying systems research in many engineering and nonengineering
disciplines.
Modularity also plays a key role in product evolution,
upgradeability, and retirement (Ishii, Lee, &
Eubanks, 1995). Zhang et al. (2001)
study the impact of modularity on product retirement costs, and life
cycle cost issues associated with modular product architectures are
discussed in Ulrich (1995), Riitahuta and
Andreasen (1999), and Dahmus and Otto (2001). Meanwhile, Newcomb et al. (1998) present a decomposition algorithm to
partition architectures into modules based on different life cycle
viewpoints, whereas both Coulter et al. (1998) and Umeda et al. (1999) have proposed methodologies for designing
modules for evolving families of products subject to life cycle
concerns. The impact of modularity on component reuse is discussed in
Kimura et al. (2001), and Allen and
Carlson–Skalak (1998) develop a
methodology for designing modular products that involves identifying
and reusing modules from previous generations of products. Similarly,
Martin and Ishii (2002) consider multiple
generations of products when presenting their approach for designing
modular product platform architectures. Their approach is one of
several that uses Quality Function Deployment (QFD) to help identify
modules within a product family (Cohen, 1995;
Erixon, 1996; Ericsson
& Erixon, 1999; Sand et al.,
2002). Erixon (1996) has extended QFD
into Modular Function DeploymentTM, a five-step process that
utilizes the Module Identification MatrixTM to help generate
module concepts (see also, Ericsson & Erixon,
1999); Huang and Kusiak (1998)
introduce a similar modularity matrix. Techniques for clustering
modules based on functional requirements using design structure
matrices are discussed in Suh (1990), Pimmler
and Eppinger (1994), Blackenfelt (2000a) Stake and Blackenfelt (2000), Kusiak (2002), and
Sharman et al. (2002). Optimization-based
approaches and AI techniques for configuring and sizing modules are
discussed in Sections 5 and 6, respectively.
2.2. Scale-based product families
As stated previously, scale-based product families are developed by
scaling one or more variables to “stretch” or
“shrink” the platform and create products whose performance
varies accordingly to satisfy a variety of market niches. Although some
consider scale-based product families to be a subset of module-based
product families (see, e.g., Fujita & Yoshida,
2001), platform scaling is a common strategy employed in many
industries. For example:
- Honda developed an automobile platform that can be
stretched in both width and length to realize a “world
car,” which was developed after failing to satisfy the Japanese
and American markets with a single platform (Naughton et al., 1997).
- Boeing developed many of its commercial airplanes by
“stretching” the aircraft to accommodate more passengers,
carry more cargo, or increase flight range (Sabbagh,
1996).
- Rolls Royce scaled its RTM322 aircraft engine by a factor
of 1.8, as shown in Figure 3, to realize a
family of engines with different shaft horsepower and thrust (Rothwell & Gardiner, 1990).
A family of scale-based aircraft engines; HP, horsepower. Adapted
from Rothwell and Gardiner (1990).
A frequently quoted example of a successful scale-based product
platform is Black & Decker's universal electric motor.
According to Lehnerd (1987), in the 1970s,
Black & Decker developed a family of universal motors for its power
tools in response to a new safety regulation: double insulation. Prior
to that, they used different motors in each of their 122 basic tools
with hundreds of variations. Through redesign and standardization of
the product line, they were able to produce all of their power tools
using a line of motors that varied only in the stack length and the
amount of copper wrapped within the motors. As a result, all of the
motors could be produced on a single machine with stack lengths varying
from 0.8 in to 1.75 in, and power output ranging from 60 to 650 W. By
paying attention to standardization and exploiting platform scaling
around the motor stack length, material costs dropped from $0.77 to
$0.42 per motor while labor costs fell from $0.248 to $0.045 per motor,
yielding an annual savings of $1.82M per year. Tool costs decreased by
as much as 62%, boosting sales, increasing production volumes, and
further improving savings. Furthermore, new designs were developed
using standardized components such as the redesigned motor, which
allowed products to be introduced, exploited and retired with minimal
expense related to product development. This electric motor example has
served as a test problem for several optimization-based approaches for
product family design as noted in Section 5. Meanwhile, the idea of
sharing components across different market segments (e.g., power tools,
lawn tools) leads us to the topic of platform leveraging.
3. PLATFORM LEVERAGING STRATEGIES
Regardless of whether the platform is modular or scalable, the basic
development strategy within any product family is to leverage the
product platform across multiple market segments or niches. Early
attempts at mapping the evolution of a product family based on
extensions and upgrades to a product platform can be found in
Wheelwright and Sasser (1989) and Meyer and
Utterback (1993), but it was not until Meyer
(1997) introduced the market segmentation
grid that platform leveraging strategies were clearly articulated. As
shown in Figure 4, market segments are
plotted horizontally in the grid while price/performance tiers are
plotted vertically; each intersection of a market segment with a
price/performance tier constitutes a market niche that is served by
one or more of a company's products. Three platform leveraging
strategies can be identified within the grid as shown in Figure 4: horizontal leveraging, vertical
leveraging, and the beachhead approach, which combines both. Meyer and
Lehnerd (1997) discuss the advantages and
drawbacks of each leveraging approach, and examples of market
segmentation grids can be found in Caffrey et al. (2002b) for spacecraft and avionics systems
and in Meyer and Lehnerd (1997) for
computers, data storage systems, power tools, and office furniture.
Platform leveraging strategies. Adapted from Meyer (1997).
The market segmentation grid is useful for both platform development
(i.e., as part of a top-down approach to product family design), as
well as product family consolidation (i.e., as part of a bottom-up
approach). For instance, Farrell and Simpson (2003) use the market segmentation grid to identify
potential platform leveraging strategies for a line of flow control
valves using historical sales data. Although most horizontal leveraging
strategies take advantage of modular platforms, the relationship
between vertical leveraging strategies and scalable platforms is
discussed in (Simpson, Maier, & Mistree,
2001). Finally, Meyer describes adaptations of the market
segmentation grid for platform-based development approaches to
nonassembled products (Meyer & Dalal,
2002) and the design and renewal of services (Meyer & DeTore, 2001). Metrics for measuring
the success of platforms and platform leveraging strategies are
discussed next.
4. METRICS FOR PRODUCT PLATFORMS AND
PRODUCT FAMILIES
An important measure of success of a product platform is how quickly
and cheaply new products can be developed from it. To help determine
when to renew or refocus product platform efforts, Meyer et al. (1997) introduced metrics for platform efficiency
and effectiveness. Platform efficiency assesses how much it
costs to develop derivative products relative to how much it costs to
develop the product platform within the product family. Platform
effectiveness measures the ratio of the revenue a product platform
and its derivatives create to the cost required to develop them. A
similar approach is taken in Schellhammer and Karandikar (2001), wherein a project ranking index,
which combines an investment index and a revenue
index, is introduced to assist in project planning. Their approach
extends the aggregate project planning concepts of Wheelwright and
Clark (1992) for managing platform projects,
derivative projects, breakthrough projects, advanced research and
development projects, and partnership projects and alliances.
During product platform design, much of the focus revolves around the
trade-off between commonality and distinctiveness: designers must
balance the commonality of the products in the family with the
individual performance (i.e., distinctiveness) of each product in the
family. For instance, Airbus has enjoyed a competitive advantage
over Boeing due to improved commonality, particularly in the cockpit.
The A330 cockpit is common to all other Airbus types while
Boeing's 767-400 cockpit is common only with the 757. This has
enabled the A330-200, a less efficient “shrink” of a larger
aircraft, to outsell Boeing's 767-400ER, a more efficient
“stretch” design of a smaller aircraft (Aboulafia, 2000). Commonality can also adversely
impact a company's reputation: in the late 1980s, engineers at
Chrysler were accused of having “fallen asleep at the typewriter
with our finger stuck on the K key” (Lutz,
1998, p. 17) due to overusage of the K-car platform and lack of
distinctive new products.
Despite the potential drawbacks of commonality, numerous indices have
been developed to measure commonality in the management science and
operations research community (see, e.g., Rutenberg,
1969; Collier, 1981; Baker et al., 1986; Trelevan
& Wacker, 1987; Thomas, 1992;
Lee & Billington, 1994; McDermott & Stock, 1994; Vakharia et al., 1996; Kim &
Chhajed, 2000). The degree of commonality index
proposed by Collier (1981) was one of the
first such indices that uses information contained in the
company's bills of materials to assess commonality for a single
end item, a product family, or an entire product line. Jiao and Tseng
(2000) extend Collier's commonality
index to create indices for component part commonality and process
commonality, overcoming the limitations of his index (cf. Wacker & Trelevan, 1986). Siddique et al.
(1998) propose separate indices for measuring
component commonality and connection commonality,
applying them to automotive underbodies, which are predominantly
integral architectures. Finally, Kota and coauthors (2000) introduce a product line commonality
index to capture the level of commonality within a product family
based on size and shape, materials and manufacturing processes, and
assembly and fastening schemes. Comparisons of these various
commonality metrics are lacking in the literature.
Martin and Ishii (1996, 1997) also introduced a commonality index similar
to Collier's, along with indices for measuring setup
costs and the point of product differentiation, which
correlate with many of the indirect costs of providing variety. Martin
and Ishii (2002) most recently proposed a
generational variety index to help identify which components
are likely to change over time to meet future market requirements and a
coupling index to measure the coupling between these
components. The importance of minimizing the coupling in a product
architecture has been studied extensively by Suh (1990). A functional similarity index was
introduced by McAdams et al. (1999; McAdams & Wood, 2002) to assist in concept
development and modular product design. Finally, indices for measuring
the degree of variation within a scale-based product family
have also been proposed (Simpson, Seepersad, &
Mistree, 2001; Nayak et al., 2002;
Messac et al., 2002b); these indices
are useful for product family optimization as discussed next.
5. OPTIMIZATION-BASED APPROACHES FOR
PRODUCT FAMILY AND PRODUCT PLATFORM DESIGN
Several optimization approaches have been developed within the
engineering design community to help determine the best design variable
settings for the product platform and individual products within the
family; a summary of these approaches is given in Table 1. In looking at the table, the
approaches are split evenly between module-based and scale-based
product families, while the work by Fujita and Yoshida (2001) specifically targets both. Almost two-thirds
of the approaches require specifying the platform a priori to
the optimization to reduce the design space and make the problem more
tractable. This is not ideal, however, because designers would like
to use optimization to explore varying levels of platform
commonality to help identify which variables to make common and unique
within the family (cf. Simpson &
D'Souza, 2002). More than half of the approaches use
multiobjective optimization to accomplish this. Three assumptions are
often made when using multiobjective optimization to design a product family:
- maximizing each product's performance maximizes its demand,
- maximizing commonality among products minimizes production costs,
and
- resolving the trade-off between assumptions 1 and 2 yields the most
profitable product family.
However, without explicitly modeling the market demand for
the products in the family and their associated manufacturing
costs, these assumptions may lead to suboptimal product families.
The universal electric motor example from Lehnerd (1987) employed in Simpson, Maier, and Mistree
(2001), Messac et al. (2002b), and Nayak et al. (2002) provides a realistic case of when this can
occur. The objective is to design a family of 10 motors based on a
scalable platform. The initial formulation scaled the motors around the
stack length of the motor (Simpson, Maier, &
Mistree, 2001), but maximizing commonality in the family using
two different approaches revealed that the motor platform should be
scaled by the radius to maximize performance (Messac
et al., 2002b; Nayak et al.,
2002). According to Lehnerd (1987),
the best choice is stack length, and through discussions with
experienced motor designers, production costs, not performance, drive
the use of stack length as the scaling variable (Simpson, Maier, & Mistree, 2001). In the table,
note that only about half the approaches integrate manufacturing costs
directly within the formulation while less than one-third incorporate
market demand/sales. Also, note that the majority of approaches
that include costs or sales in their formulation use single objective
optimization, rather than multiobjective, where the objective is to
either maximize profit or minimize cost.
Summary of engineering optimization approaches for product family
design
Although not specifically noted in the table, most of the approaches
that incorporate uncertainty in the formulation model it in the market
demand and future sales of the products in the family (Seepersad et al., 2000; Gonzalez–Zugasti et al., 2001; Jiang & Allada, 2001; Allada
& Jiang, 2002; Li & Azarm,
2002). Uncertainty in customer requirements has also been used
to develop robust product platforms. Chang and Ward (1995) were among the first to use robust design
techniques to develop a family of products that were insensitive to
design changes. Simpson and coauthors use robust design techniques to
develop scale-based platforms for General Aviation Aircraft (Simpson et al., 1999), electric motors (Simpson, Maier, & Mistree, 2001), and
absorption chillers (Hernandez et al., 2001).
Blackenfelt (2000b) uses robust
design techniques to maximize profit and balance commonality and
variety within a family of lift tables.
The number of stages in the optimization approach is another
interesting statistic. Single-stage approaches seek to optimize the
product platform and corresponding family of products simultaneously
while two-stage approaches optimize the platform first and then
instantiate the individual products within the family during the second
stage; multistage approaches are those that involve more than two
stages. Single-stage and two-stage approaches are employed almost
equally in the literature, and the reader is referred to the table for
examples. Although both approaches are effective at determining the
best design variable settings for the product platform and product
family, single-stage approaches will yield the best overall performance
of the product family because the optimization is not partitioned into
two or more stages (cf. Messac et al.,
2002a). The dimensionality of single-stage optimization
problems, however, is considerably higher than in two-stage approaches,
which often leads to many computational challenges (cf. Messac et al., 2002b). It is also worth
noting that a modification to the two-stage approach is introduced by
Nelson et al. (2001) and used by Fellini et
al. (2000), Fellini, Kokkolaras, Michelena,
et al. (2002), and Fellini, Kokkolaras,
Papalambros, et al. (2002): the first stage
involves individually optimizing each product while the second stage
involves optimizing the product family with constraints on performance
losses due to commonality. Meanwhile, only two multistage approaches
have been developed. First, Hernandez et al. (2002, 2003) develop a
multistage optimization approach by viewing the product platform design
problem as a problem of access in a geometric space. Second, Allada and
Jiang (2002) introduce a dynamic programming
(DP) model for configuring module instances within an evolving family
of products. Their DP-based approach is used to plan module
introduction (i.e., which modules to introduce when) to maximize the
total profit in a given planning horizon. An alternative classification
of optimization approaches based on the extent of the optimization
(i.e., module attributes, module combinations, or both) is discussed in
Fujita (2002).
Based on the variety of optimization algorithms listed in the table,
there does not appear to be a preferred algorithm for product family
design. Linear and nonlinear programming algorithms (e.g., sequential
linear programming, sequential quadratic programming, nonlinear
programming, generalized reduced gradient) are employed by many
researchers, as are derivative-free methods such as genetic algorithms
(GAs), simulated annealing, pattern search, and branch and bound. When
the design space is small enough, exhaustive search techniques (Hernandez et al., 2003) or orthogonal arrays (Chang & Ward, 1995; Blackenfelt, 2000b) can be used to
enumerate different combinations of parameter settings and modules.
However, very few problems involve so few options that such an approach
can be taken, and many researchers advocate the use of GAs for product
platform design due to the combinatorial nature of the product family
design problems (Gonzalez–Zugasti & Otto,
2000; Fujita & Yoshida, 2001;
Li & Azarm, 2002; D'Souza & Simpson, 2003). Finally,
algorithm choice is often mandated by the selected framework, for
example, Decision-Based Design (DBD, Li & Azarm,
2002), Target Cascading (Kokkolaras et al.,
2002), 0–1 integer programming (Fujita
et al., 1999), Physical Programming (Messac
et al., 2002a), and the Compromise Decision Support
Problem (Simpson et al., 1999).
Finally, these optimization approaches have been tested on a variety
of product families as noted in the last column of the table. These
product families range from 2 to 16 products, and include consumer
products such as drills (see, e.g., Li &
Azarm, 2002), vacuum cleaners (Jiang &
Allada, 2001) and automobiles (Fellini,
Kokkolaras, Michelena, et al., 2002; Kokkolaras et al., 2002); industrial
products such as chillers (Hernandez et al.,
2001) and flow control valves (Farrell &
Simpson, 2003); and complex systems such as aircraft
(see, e.g., Fujita & Yoshida, 2001; Simpson & D'Souza, 2002) and spacecraft
(see, e.g., Gonzalez–Zugasti et al.,
2000). Detailed analyses for the universal electric motor
problem can be found in Simpson, Maier, and Mistree, (2001); it has been used to benchmark a variety of
optimization approaches, as noted in the table. The commercial aircraft
problem found in Fujita et al. (1998) and
Fujita and Yoshida (2001) uses aircraft
analyses available in the literature in combination with their own
models for design and development, facility, and production costs and a
profit model for the manufacturer. The nail gun (Nelson et al., 2001), vacuum cleaner (Jiang & Allada, 2001), and power screwdriver
and electric knife (Rai & Allada, 2002)
examples are comprehensive as well. The automotive example used in
Fellini, Kokkolaras, Michelena, et al. (2002)
and Kokkolaras et al. (2002) is based on a
detailed vehicle body structural model that is currently unavailable to
the public; simpler models of the automotive vehicle frame can be in
Fellini, Kokkolaras, Papalambros, et al. (2002) and Cetin and Saitou (2003). The analyses for the absorption chiller
problem (Seepersad et al., 2000; Hernandez et al., 2001) are not publicly available
either.
6. AI IN PRODUCT PLATFORM DESIGN AND
CUSTOMIZATION
AI techniques for product platform design and customization lag
behind optimization-based approaches; however, they have been
successfully employed and shown great promise for automatic product
configuration and automatic computer-aided design (CAD) modeling and
geometry generation. Sabin and Weigel (1998)
recently reviewed rule-based and model-based techniques for automatic
product configuration; they state that the acquisition of the rules or
constraints on which the reasoning depends is one of the major
challenges in knowledge-based configuration systems. One such system is
the Product Module Reasoning System developed by Rosen (1996), which reasons about sets of product
architectures, translates design requirements into constraints on these
sets, compares architecture modules from different viewpoints (e.g.,
material, connections, covers), and directly enumerates all feasible
module combinations. His approach uses discrete mathematics
(combinatorics and set theory) and provides the foundation for the
Product Family Reasoning System developed by Siddique and Rosen (2000, 2001), which
reasons about families of products in addition to individual product
architectures. A configuration framework for mass customization of
products that employs the Unified Modeling Language is introduced by
Felfernig, Friedrich, and Jannach (2001) and
Felfernig, Friedrich, et al. (2001). Claesson
et al. (2001) use
function–means–trees and a chromosome model to create
configurable components that represent a parameterized set of design
solutions; the concept is currently being deployed at Saab automobile
to help control product variety. Finally, agent-based systems offer
many advantages for concurrent product and process design and
configuration (Shen et al., 2001), and
agent-based approaches for module-based product family design are being
explored by some researchers (Allada & Rai,
2002; Liang & Huang, 2002; Rai & Allada, 2002). However, these approaches
are not yet as mature as the aforementioned optimization-based
approaches.
Sabin and Weigel (1998) also discuss
several case-based reasoning techniques for automatic product
configuration. Such techniques generally involve eliciting customer
requirements, retrieving a configuration from a pool of stored cases,
and adapting the case to satisfy the new situation. A case-based
approach for mass customization of goods was proposed by Tseng and Jiao
(1997a). Their approach organizes
information around the common features and platform within a product
family architecture to facilitate case retrieval; adaptation of cases
occurs either by reinstantiating the case if it is similar enough or by
providing specific adaptation knowledge (configurational or
topological) to modify aspects of the case. In Tseng and Jiao (1997c), a two-phase methodology is
presented for recognizing patterns of functional requirements in
similar existing products to help define new products.
To reduce data redundancy when modeling families of products, the
Generic Bill-of-Material (GBOM) concept developed at the Eindhoven
University of Technology (Hegge & Wortmann,
1991; Erens et al., 1992; van Veen, 1992; Erens &
Hegge, 1994) allows all variants of a product family to be
specified only once. Within a GBOM, the Primary Generic Product (PGP)
represents the set of all variants of a particular primary product and
Generic Subassembly Products (GSPs) describe the sets of subassemblies.
Parameter values of the PGP are passed through the levels of the GBOM
and are inherited by lower level GSPs. McKay et al. (1996) combine the GBOM concept with product
modeling concepts and software to reduce data redundancy when
considering multiple views (e.g., sales, manufacturing, assembly). Jiao
and coauthors (2000) have extended the GBOM
concept to include operations, and De Lit et al. (2001) use GBOMs to support assembly planning for
families of products. Cheng et al. (2002)
extended the GBOM concept for product family development within an
extended enterprise.
Finally, grammar-based approaches have been developed for automating
the generation of products within a family as well as automating the
generation of CAD models. Agarwal and Cagan (1997, 2000) first
popularized the approach, developing shape grammars to create
a variety of coffeemakers. Siddique and Rosen (1999) use the same coffeemaker example to
demonstrate a graph grammar-based approach for product
platform design and instantiation. Within their graph grammar, graphs
are used to represent the core function and structure (i.e., the
platform), and grammars are used to specify the relationships between
the core and the options (i.e., product variants). Meanwhile, Du et al.
(2001b) have also developed a graph
grammar-based approach for modeling product families. They use a
Programmed Attribute Graph Grammar to specify the design space of the
product family, which is then customized by varying modules according
to a control diagram that captures the complex relationships and
configuration constraints between modules. Their approach is extended
in Du et al. (2002) by implementing a graph
rewriting process to transform product family graphs into product
variant graphs; a family of office chairs is used to illustrate their
approach. Graph grammars are being used by Siddique and Shao (2001) to develop a Web-based system for product
family reasoning and Siddique and Yanjiang (2002) describe a template-based approach that
automatically generates CAD models for each member in the product
family. The approach was expanded to include parametric design, mating
relationships, and modularity (Martinez–Larrosa & Siddique, 2002), and
is one of many Web-based approaches for customization that is discussed
next.
7. PRODUCT PLATFORMS AND (WEB-BASED)
CUSTOMIZATION
In addition to improving economies of scale and scope, a product
platform can facilitate customization by enabling a variety of products
to be quickly and easily developed to satisfy the needs and
requirements of distinct market niches (Pine,
1993a). Although flooding the market with a variety of
products derived from a platform may satisfy some customers by
providing a substitute for customization, variety is not
customization. Variety provides choices for customers but does not
enable the customer to specify the product. A customized product, on
the other hand, is designed to meet the specific needs of a particular
customer; therefore, customers must be involved at one or more points
in the product realization process for the product to be truly
customized (see Fig. 5). This
distinction is overlooked in much of the mass customization literature
as noted by Duray et al. (2000), who study
and classify companies that customize products based on the point of
customer involvement. Customized products can be either made to order,
tailored to order, assembled to order, or made to stock, each of which
has different implications for product platform development and the
associated information technologies needed to deliver that product (cf.
Duray & Milligan, 1999). Regardless of
the stage of customer involvement, product platforms play an
integral role in facilitating the product customization process.
Tseng, Jiao, and coauthors have extensively studied the relationship
between product family architecture and mass customization (Tseng et al., 1996; Tseng &
Jiao, 1998; Jiao & Tseng, 1999,
2000; Du et al., 2000, 2001a),
including a framework for virtual design for mass customization (Tseng et al., 1997).
Points of customer involvement for product customization. Adapted
from Duray and Milligan (1999).
Choi and Whinston (1999) assert that the
“computer-mediated market will accelerate the process of
customization through its technologies.” Efficient information
technology for product customization is one of several important
research directions facing many of today's firms (Da Silveira et al., 2001; Zipkin, 2001). For instance, Web-based
customization opens new paradigms for one to one marketing (Gilmore & Pine, 1999), and it can provide
valuable information for companies to improve customer demand estimates
and determine whether to contract or expand product variety (Kotha, 1995).
Some prototypical Web-based systems for platform customization have
been developed. For example, Simpson et al. (2003) created a Web-based system for customizing
refiner plates for pulp and paper processing based on predefined
platforms; the system exploits a parametric and feature-based modeling
scheme for refiner plates developed in earlier work (Kulvatunyou et al., 2000). Flores et al. (2002) present a Web-based system for customizing
coated steel belt sheaves, extending early parametric modeling
capability developed in their lab (Rohm et al.,
2000). A Web-based system that incorporates fuzzy geometric
customization and fuzzy reasoning is presented in Chen et al. (2001) and demonstrated for customizing wineglasses
and furniture. A Web-based knowledge system to support product family
design has been developed by Zha and Lu (2002) and tested with a power supply product
family. Researchers now recognize that suppliers are playing an
increasingly important role in new product development, particularly in
the automotive industry (MacDuffie et al.,
1996; Gupta & Krishnan,
1998a; Fisher et al., 1999),
and Web-based systems are being developed to include suppliers in the
design process (Huang, Huang, & Mak,
2000; Huang & Mak, 2000). Huang
and his colleagues have also been developing systems to support
collaborative product development and Design for X capabilities over
the Internet (Huang & Mak, 1999; Huang, Shi, & Mak, 2000; Huang, Lee, & Mak, 2001; Huang, Shen, & Mak, 2001). Finally, commercial
software developers are staking their claims in this rapidly growing
area of research. For instance, Windchill®
DynamicDesignLinkTM enables dynamic, collaborative,
Web-based product customization for design to order products (Parametric Technologies Corporation, 2002). The
software allows customers to create custom products via the Web by
providing guided product selection and configuration, automated product
and process selection and generation, and integration with enterprise
business systems.
8. CLOSING REMARKS AND FUTURE RESEARCH
DIRECTIONS
As evidenced by this comprehensive review of product family and
product platform design research, there has been considerable progress
in planning, modeling, designing, and assessing product platforms and
the families of products derived from them. Be it a module-based or a
scale-based product family, there are now a variety of
optimization-based and AI-based techniques to support the design and
decision making process. Although optimization-based approaches are
more prevalent in the literature, AI-based techniques show great
promise for product family design and customization. Web-based systems
are also becoming more prominent to promote customer involvement in the
product realization process and facilitate individual product
customization via a well-defined product platform. Thanks to this
flurry of research activity, this nascent field of study has matured
rapidly in the past decade; however, considerable research is still
needed to help bridge the gap between planning and managing families of
products and designing and manufacturing them. Toward this end, several
research thrusts are identified in the remainder of this section to
help guide future efforts in this burgeoning field of research.
8.1. Product family planning and platform development
In an empirical study of 108 new product development projects,
Tatikonda (1999) found that platform and
derivative projects are executed in a similar manner and do not differ
in terms of project success even though they do differ in task
characteristics (i.e., complexity and amount of new technology
introduced). This suggests that some firms can manage single product
and platform projects in similar ways, but system-level designers must
still “address the problem of what product architecture should be
used to deliver the different products while sharing parts and
production steps across the products” (Robertson & Ulrich, 1998, p. 21). Product
family maps and the market segmentation grid support product family
planning from a management perspective but offer little support to
product and process engineers who must determine each product's
architecture and the appropriate levels of platform commonality and
component sharing within the family (cf. Maier &
Fadel, 2001). Information management systems to support product
family development and foster component/module reuse are needed,
and platform planning processes such as given in Robertson and Ulrich
(1998) need to be developed, tested, and
refined to help support engineering decision making during
platform-based product development. Lessons learned from the set-based
approaches employed successfully for years at Toyota (Ward et al., 1995) also show promise as models for
product platform planning and development. Approaches for determining
the extent and number of platforms to offer within a family (Seepersad
et al., 2000, 2002;
de Weck et al., 2003) also need to be further
investigated.
8.2. Quantifying the benefits and drawbacks of
platform-based product development
The metrics discussed in Section 4 provide surrogates for measuring
the financial impact of platform development; however, quantifying the
economic benefit of platform-based product development is important for
strategic decision making (cf., Meyer et al.,
1997; Schellhammer & Karandikar,
2001). Some preliminary work in this area includes the use of
activity-based costing to estimate the cost savings (Siddique & Repphun, 2001) and reduction in
development time (Siddique, 2001) for
developing a hard disk drive spindle motor platform for a family of
hard disks. Although many researchers espouse the benefits of
platforms, platform-based approaches can impose additional costs on
product development. The fixed costs of developing a product platform
can be enormous, as evidenced by Ulrich and Eppinger (2000), who note that developing a product platform
can cost 2–10 times more than a single product, and sharing
components across low-end and high-end products can increase unit
variable costs due to overdesigned low-end products (Gupta & Krishnan, 1998a; Fisher et al., 1999). In the automotive industry,
Muffatto (1999) found that up to 80% of total
vehicle development cost is spent on platform development (including
engine and transmission); others argue that platform development
accounts for only 60% of these costs (Sundgren,
1999). Krishnan and Gupta (2001)
develop a mathematical model to examine some of the costs of
platform-based product development and find that platforms are not
appropriate for extreme levels of market diversity or high levels of
nonplatform scale economies.
8.3. Modeling customer demand for product families
As discussed in Section 7, incorporating customers in the product
realization process is critical to successful platform customization,
and recent work in DBD has revealed the importance of formulating a
proper objective function to reflect the interests of both consumers
and producers (Hazelrigg, 1996, 1998; Chen et al., 2000).
Georgiopoulos et al. (2002) integrate
engineering and business models for evaluating portfolio decisions
involving a premium-compact and sport utility vehicle, and a formal
framework for DBD was proposed by Hazelrigg (1996, 1998) and
implemented for product design (Gu et al.,
2000; Li & Azarm, 2000; Wassenaar & Chen, 2001) and product line design
(Li & Azarm, 2002). Within their
framework, Li and Azarm (2000, 2002) use conjoint analysis (see, e.g., Louviere, 1988; Green &
Srinivasan, 1990) to estimate customer demand within their DBD
framework. Tseng and Du (1998) have also used
conjoint analysis techniques for soliciting customer input for mass
customized products that are based on a product family. Conjoint
analysis has been widely used in marketing and management science for
product platform design (Moore et al., 1999),
product line design (see, e.g., Green & Krieger,
1985; McBride & Zufryden, 1988;
Kohli & Sukumar, 1990; Dobson & Kalish, 1993), and product line
redesign (Page & Rosenbaum, 1987);
however, Chen and Hausman (2000) note that
many of these conjoint based approaches are mathematically intractable
or NP-hard. Consequently, alternatives to conjoint analysis such as
choice-based conjoint analysis (Azarm et al.,
2003) and discrete choice analysis (Wassenaar
& Chen, 2001) are being investigated for use with DBD, but
these techniques have not yet been extended for product family
design.
8.4. Design for manufacturing and assembly (DFMA)
Most DFMA techniques have been developed for single products (cf.
Bralla, 1999; van Vliet et
al., 1999; Shah & Wright, 2000;
Boothroyd et al., 2002), and they do not
effectively support product family design. For instance, creating
modular product architectures runs counter to the DFMA principle of
reducing part count by integrating parts, which often makes the
architecture more integral; however, modularity can facilitate
assembly. Conversely, increasing commonality will reduce the overall
part count within a product family, and it will have the added benefit
of delaying the point of product differentiation within the
manufacturing and assembly process (see, e.g., Lee
& Tang, 1997; Gupta & Krishnan,
1998b; He et al., 1998; Desai et al., 2001; Ma et al.,
2002). Approaches for concurrent design of product families and
assembly systems have been developed (Stadzisz &
Henrioud, 1995; Stadzisz et al., 1995;
Martinez et al., 2000; Fouda et al., 2001). Kusiak (2000) discusses the importance of modular design in
developing agile manufacturing systems, and an approach for optimizing
modules for reconfigurable manufacturing systems was recently
introduced by Yigit et al. (2002). These DFMA
guidelines need to be formalized to support product family design and
platform-based product development.
8.5. Support for small- and medium-size manufacturers
Although much of the research in product family design has focused on
large corporations (e.g., Volkswagen, Boeing, Kodak, HP), product
platforms and customization are becoming particularly important for
small- and medium-size firms. Small manufacturers lack an
“adequately trained technical workforce” and do not have
the “deep pockets” and “financial float” that
is available in larger companies (Maupin &
Stauffer, 2000). Developing product platforms that can be
leveraged across a variety of products is critical to remaining
profitable, and tools such as Variant Mode and Effects Analysis, which
have been tested at 10 medium-sized German automotive manufacturers,
are useful in reducing product complexity and unwanted product variety
(Schuh & Tanner, 1998). Preliminary
efforts to reengineer and improve commonality within a product family
at small manufacturing firms can also be found in Maupin and Stauffer
(2000), Berti et al. (2001), and Farrell and Simpson (2003).
8.6. Overcoming organizational barriers to
platform-based product development
Designing a product platform and corresponding family of products
embodies all of the challenges of product design while adding the
complexity of coordinating the design of multiple products in an effort
to increase commonality across the set of products without compromising
their distinctiveness. For instance, Sanchez and Mahoney (1996) and Pederson (1999)
discuss the impact of modularity and platforms, respectively, on the
organization. Despite the benefits and advantages of platform-based
product development, many companies remain hesitant to embrace product
families and product platforms. For instance, despite the success and
growth of the PC industry due to open, modular product architectures
and product families (cf. Baldwin & Clark,
1997; Christensen & Verlinden,
2002), many aerospace organizations are resistant to such an
approach because they fear losing their competitive edge in a
relatively small market (cf. Caffrey et al., 2002a, 2002b). Erens (1997, p. 2) notes that “if sales engineers
and designers focus on individual customer requirements, they feel that
sharing components compromises the quality of their products.”
Dismantling organizational complexity is the first step in reducing the
negative effects of product variety (cf. Galsworth,
1994), and innovative approaches are needed to overcome the
corporate inertia that develops within many companies. As an example,
Chrysler's dramatic reorganization in 1988 around five platform
teams (small car, large car, Jeep, truck, and minivan) with
crossfunctional “tech clubs” (e.g., chassis, engine, HVAC,
etc.) achieved significant reductions in development costs, lead time,
and variable costs (Lutz, 1998).
Organizational structures at other automotive companies (e.g., GM,
Ford, Toyota, Volkswagen, Nissan, Fiat, Mazda, Honda) to promote
platform development are documented and discussed in (Shimokawa et al., 1997; Cusumano
& Nobeoka, 1998). Meanwhile, Kotha (1995) examines an innovative approach used by
Japan's National Bicycle Industrial Company for combining mass
production and mass customization. Due to the impressive gains
resulting from such innovations, we must strive to understand the
organizational impact of platform-based product development better.
ACKNOWLEDGMENTS
This work was supported by National Science Foundation Career Grant
DMI-0133923. Any opinions, findings, and conclusions or recommendations
presented in this paper are those of the author and do not necessarily
reflect the views of the National Science Foundation.
