A multi-criteria approach to assess the impact of regulations and voluntary programs on product architecture

ABSTRACT

A methodology to find an optimal architecture for appliances undergoing periodic governmental regulations and energy-efficiency improvements is proposed. For instance, manufacturers producing home appliances are continuously challenged by new rules and regulations to meet minimum energy efficiency requirements. These requirements are either enforced by new regulations or recommended by voluntary programs, depending on governmental policies. The methodology proposed in this study initially identifies all regulatory and voluntary factors that affect product architecture. These factors include product performance, economic and environmental considerations. Using these factors, the analytic hierarchy process (AHP) is applied to evaluate trade-offs among alternative product architectures with varying degrees of modularity. Lastly, AHP assists in identifying the optimal product architecture based on the formerly identified factors and criteria. Using this methodology, appliance manufacturers are able to choose the most sustainable architecture, one that is environmentally friendly, and which meets governmental regulations.

KEYWORDS:

• Product architecture
• product innovation
• energy conservation
• environmental regulation
• analytic hierarchy process
• sustainable operations

1. Introduction

Proper evaluation and effective selection of product architecture can have a significant impact on production and operations. For instance, Volkswagen has been able to save 1.7 USD billion because it is able to choose the most appropriate product architectures (Dahmus et al., 2001). Ulrich (1995) defines product architecture as ‘the scheme by which the function of a product is allocated to physical components’. Thus, every product consists of a set of components and each component delivers certain functions. Each component is linked with other components through certain interfaces. A product’s architecture is more modular if it tends to have a one-to-one mapping from each physical component to each function of the product. However, complex mappings (e.g. many-to-one) result in a more integral architecture.

The architecture of a product has important implications on its design changes (Ulrich, 1995). Various factors drive design changes. The change can be customer-based, competition-based, and/or regulation-based. Regarding regulation-based changes, some features could be forced to comply with a set of pre-specified minimum requirements called the Minimum Energy Efficiency Standard (MEES). In the United States, for instance, MEES is enforced by the Department of Energy (DOE). In addition, there are other circumstances where voluntary programs are designed to motivate manufacturers to adopt new changes. Firms often consider these changes as opportunities by which they can simultaneously innovate their products designs and optimize them for environmental requirements and customer needs.

Regulations have had a long history around the world. For instance, in the U.S., efforts for enforcing appliance efficiency standards started in the 1960's at the state-level. However, it was not until the Energy Policy Conservation Act (EPCA) of 1975 that those efforts were coordinated at the federal level. After the introduction of the EPCA, the U.S. government periodically adds new regulatory requirements.Usually prior to each regulation, extensive studies are conducted by the government to evaluate such a regulation’s economic, environmental, and technical impact.

DOE regulations typically require appliance manufacturers to increase the MEES for some of their products. The DOE usually specifies the new MEES and sets a date for the regulation to take effect. Customarily, after the effective date, manufacturers are not allowed to produce any units that do not satisfy the newly enforced standard. However, all the items produced prior to that date can be sold.

Worldwide regulations take various forms and can be classified in different ways. For example, while the U.S. focuses on enforcing laws with MEES, other countries use voluntary programs, product labeling, and financial incentives, etc. Regulations can also be statistics-based or technical-based. As for the former, statistical studies are conducted to understand the distribution of existing products based on their energy efficiency level. Then a certain percentage of these products is banned (say the lowest 25%). In the latter case, technical and economic studies are made to check the technological and economic feasibility of the upgrade.

Table 1 presents a list of regulated appliances in various countries. While the U.S. and Canada regulate all four listed appliances, Russia regulates only water heaters and refrigerators. This non-uniformity in the enforcement of the regulations for appliances in various countries has implications for both local manufacturers and international companies which sell their products on the global markets.

Table 1. Regulated appliances in different countries*

Appliance/Country	USA	EU	China	Canada	Australia	Russia
Central Air Conditioners	✓		✓	✓	✓
Water Heaters	✓			✓	✓	✓
Refrigerators	✓	✓	✓	✓	✓	✓
Washing Machines	✓		✓	✓

* Source: Nadel (2002)

One additional important factor to consider is the frequency of regulations – i.e. various products may have different periodic patterns of regulations. For example, considering the U.S., washing machines and refrigerators are regulated more frequently than central air conditioners (Table 2). Table 2 shows that U.S. washing machine manufacturers would need to apply frequent product changes as a priority factor when choosing their product architectures. In addition, different appliances can have different intensities of required efficiency improvement. For example, for the U.S. regulations for refrigerators in 1993 and 2001, no refrigerators satisfied the standard prior to its enforcement (Nadel, 2002). This reflects a relatively intense regulation. Interestingly, however, for many other products, brands satisfying the rules had already existed before even enforcing the rules.

Table 2. Years of effective new regulations in the U.S. for some appliances*

Appliance/Year	‘88	‘90	‘92	‘93	‘94	‘00	‘01	‘04	‘06	‘07
Central Air Conditioners			✓						✓
Water Heaters		✓						✓
Refrigerators		✓		✓			✓
Washing Machines	✓				✓			✓		✓

* Source: Nadel (2002)

These regulations are justified because reports show that they have significant benefits. Lowenberger et al. (2012) states that in 2010, in the U.S., standards compliance resulted in a saving of 278 terawatt hours (TWh) and reduced electricity consumption by seven percent. Lowenberger et al. (2012) also calculated energy savings of up to 682 TWh by 2025, which could save consumers and businesses more than 1.1 trillion USD by 2035, excluding the reduction in carbon dioxide emissions. It should also be noted that although regulations are generally expected to negatively impact manufacturers by increasing production costs, several studies found that proper management of product upgrades can and do result in a positive return on equity (ROE) for manufacturers (McMahon et al., 1996).

To assist in choosing the optimal architecture for home appliances, this study initially performs an in-depth literature search to identify all the pertinent criteria and sub-criteria for evaluating different architectures for appliances. Using the ascertained criteria/sub-criteria, the analytic hierarchy process (AHP) is employed to identify viable alternative architectures. AHP evaluates them to determine the optimal architecture. The proposed methodology allows decision-makers to perform a pairwise assessment of the criteria. It also provides an overall ranking of the criteria when developing alternative architectures. AHP allows for incorporating both tangible and intangible factors when comparing criteria and evaluating various alternatives. Specifically, in this study, AHP (thus the suggested methodology) lets decision-makers evaluate various modular and integral architectures so that they can choose particular upgrades based on the strategy of the firm. In this way, they can choose priorities of its strategic objectives, considering environmental, economic and technical factors. A noticeable advantage of the suggested methodology is its practicality and ease of use.

The remainder of this paper is structured as follows: All the significant criteria/sub-criteria that affect the choice of optimal architecture(s) for appliances are identified through the literature review. The AHP is briefly introduced and an in-depth explanation of the suggested methodology follows. To demonstrate its practicality and inherent benefits, an illustrative example uses the suggested methodology and applies it to a real-world situation. Next, by performing a sensitivity analysis of the findings/results, various real-world scenarios are introduced to identify some viable alternative architectures. Finally, a discussion of the results and conclusions are provided.

2. Determinant factors affecting product architecture

This section explores the literature on products architecture and optimal designs. Product architecture choices can have considerable implications on product changes. Ulrich (1995) discusses this effect in two situations of change: (1) for a product, and (2) for new generations of a product. In the latter case, which is the focus of this paper, if for instance, a modular architecture is applied, functional elements can be added by changing only one or a few components of the product. On the contrary, if an integral architecture is chosen, most of the components may need to be changed. However, for the integrated products, due to their integrality, global optimization of system functions, performance and attributes are more likely to be achieved. For example, if optimizing size or weight of a product is a primary goal, then an integral architecture could more suitably serve the purpose. This is because each functional element can be mapped to multiple components of the product, unlike for a modular architecture. Yet, modular architectures can provide notable advantages on cost, quality, operational excellence, variety of product configurations (Ulrich & Eppinger, 1988), flexibility of the manufacturing processes (Jacobs et al., 2007), organizational learning (Sanchez, 2000), and the buyer-supplier relationships (Hoetker et al., 2007). For instance, if offering a high variety is a priority, then a modular architecture is the more feasible option. Furthermore, inherent standardization in modular architectures can provide cost and efficiency advantages. It should be noted, however, due to the combinatorial richness derived from a mix and match among various modules, the modular architecture is not a ‘silver bullet’ and can be quite complex (Simchi-Levi, 2010). Consequently, there are various trade-offs along the modularity-integrality spectrum when deciding to choose an appropriate product architecture.

Several attempts have been made to quantify the degree of modularity in a product. J. Mikkola (2006) argues that the degree of modularity can be measured by observing four factors: types of components (standard versus new-to-firm), interfaces, degree of coupling, and substitutability. These factors are then incorporated in a mathematical model called the ‘modularization function’ that quantifies the degree of modularity (J. Mikkola, 2006). Although the approach ignores some of the qualitative aspects of architecture like types of interface (e.g. electrical, mechanical, etc.), it offers a solid method to measure product architectures under certain assumptions. Interestingly, there is a body of literature that suggests a continuum of architectures between the integrality at one end of the spectrum and modularity at the other end. In addition, researchers claim that any design has, to some extent, a degree of modularity (Campagnolo & Camuffo, 2010). While modularity is often measured at the product design level, some researchers address it at the production system and/or organizational design level. Yet, another study states that modular architectures require standardization of parts and processes (Swaminathan, 2001).

Several studies attempt to identify the trade-offs for choosing modular versus integral architectures. For instance, functional, socio-economic and technological factors have been considered ‘competing factors’ that influence architecture choice and help identify ‘successful designs’ (Clark, 1985). Also, design choice can be evaluated based on criteria such as financial, commercial, and operational performance (Ravasi & Stigliani, 2012). Schilling (2000) identified eleven factors that affect a tendency towards or away from modularity. Some of these factors are the component specificity, degree of difficulty of assessing interaction among components, assembly complexity, differential capabilities among firms, diversity in technological options, customer heterogeneity in desired function, speed of technological change, competitive intensity and synergies, and interactions between these factors. Other studies focus on certain aspects of the product or supply chain. For instance, Howard and Squire (2007) used an empirical method to find an association between product modularization and buyer-supplier collaboration. Thus, nature of the supply chain entities relationships is impacted depending on the product architecture choice. Ülkü and Schmidt (2011) employed a mathematical model to find that in contrary to general belief, integral and modular architectures are not necessarily linked to vertically integrated supply chains and decentralized supply chains. They conclude that other factors like supplier product development capabilities, supply chain relationship types, and technical collaboration penalties play a significant role. Takeishi and Fujimoto (2001) investigated the auto industry. They argued that modularization is multi-faceted and involves the product, production and supplier systems. They conclude that there are ‘tensions’ between the three levels (product, production and supplier systems) of modularization and the interrelationship between their modularity choices. Pandremenos and Chryssolouris (discuss how the degree of modularity of a product can facilitate its customization and Gauss et al. (2019) suggest module-based design to deal with product variety and customization, utilizing reconfigurable manufacturing systems (RMS). Similarly, Fixson (2005) operationalized the concept of product architecture using a framework where product, process, and supply chain dimensions were considered. Ahmed (2018) relates the modularity of a product to complexity of managing its supply chain.

On the effect of regulation, Mauer et al. (2013) examined changes of performance, features and price for regulated products before and after a regulation’s effective date. The authors indicate that, after the enforcement of regulations, the performance of a product usually improved or stayed the same. In addition, for all the nine products that they evaluated, it was noticed that manufacturers had either introduced new features or expanded their availability. Considering the price of the product, Mauer et al. (2013) found that in cases where the price did not decline or stayed the same (five out of nine products), from customers’ perspectives, the price increase was outweighed with energy saving benefits. Desroches et al. (2013) noticed that the net present value realized by manufacturers undergoing energy efficiency regulations is undermined in technical analysis support for these regulations. If an experience curve is considered, regulation events have a more positive impact on manufacturers than predicted. To date, most research has focused on the effect of the regulations on different performance measures. There lacks any study which addresses its impact on product architecture.

Another contemporary perspective on product architecture, particularly more so when considering governmental regulations, is sustainability. It is common to associate modular product architectures with sustainability merits. These merits could be attributed to environmental, business, or technological dimensions (Sandborn & Myers, 2008). For instance, modular architectures by nature allow for recyclability and reuse of decomposable components. Considering sustainability, Ma and Kremer (2014) suggest a methodology to find an optimal modular product design (MPD). Using a multi-criteria integer programming model, the authors suggest an optimal design based on three life cycle factors: cost, energy consumption, and labor hours. These criteria are intended to reflect three sustainability dimensions: economic, environmental, and social sustainability. Ma and Kremer (2014) used a sustainability index to evaluate different modular structures to find the most sustainable structure. While the research on sustainability focuses on its impact related to each of the factors, the proposed model in this paper focuses on architectural implications of these factors.

Pashaei and Olhager (2017) considered the optimal architecture problem in a global setting and study firms that have manufacturing facilities located across the globe. The authors studied three real world cases and tested several viable propositions. The primary research questions posed by Pashaei and Olhager (2017) concentrated on how characteristics of global operations can affect the optimal degree of modularity. Based on their research, the component and assembly plants designed for economies of scope favor more of a modular architecture, whereas those designed for economies of scale favor integral architectures. Furthermore, Pashaei and Olhager (2017) suggest that since integral architectures are harder to manage through external parties, having the key technologies in-house enables more appropriate integral architectures than when key technologies rely upon external parties.

There are numerous studies that use multi-criteria approaches to find optimal product designs. For instance, Chang et al. (2007) utilized the analytic hierarchy process (AHP) to select the best slicing machine based on machine-related, human-related, management and measurement factors. However, they do not classify slicing machines based on their generic design architecture. Baumann et al. (2002) state that most research studies do not provide enough links between green product designs and business factors. The authors also proposed a life cycle assessment (LCA) as a major analytic tool which can be used for measuring the environmental performance of products. LCA is integrated with other factors like cost in a multi-objective optimization to find the best design. Baumann et al. (2002) suggest that environmental optimization should be conducted on a higher level and that governmental regulation should be linked to the optimal design problem. Chu et al. (2009) argued that most of the studies fail to identify the importance of product structure as a critical factor in the reduction of negative environmental impacts. Therefore, they proposed a CAD-based approach and utilize the genetic algorithm to find the optimal product architecture from a set of designs.

Smith and Yen (2010) treated the issue of environmentally friendly product design by modularizing design to meet green constraints such as material compatibility. Thus, green requirements can be applied to relevant modules. They employed atomic theory and touch matrix to represent relations between pairs of components. However, they did not identify other factors like economic and performance implications of the design. Shi et al. (2001) proposed a framework to optimize product design. While customer preferences represent the criteria in their model, technological and economic feasibility are assumed to be valid for all the considered designs set and not considered explicitly. In addition, these authors utilized the nested partitions (NP) method and mathematical programming to solve the design problem. They build on other studies in which greedy algorithms and dynamic programming (Kohli & Krishnamurti, 1989), genetic algorithm (P.S. Balakrishnan & Jacob, 1995; P.V. Balakrishnan & Jacob, 1996), and divide and conquer (Green & Krieger, 1985) were utilized for the design problem in addition to a class of problems called the share-of-choices problems. However, these studies did not consider the degree of modularity as a crucial architectural decision. Neither did the studies include environmental attributes in the design decision.

Chen (2001) considered the conflict between traditional and environmental attributes and applied theories of optimal product design at the supply side, and preferences and conjoint analysis on the demand side. He evaluated the environmental and economic impact of regulations imposed by the government. Thus, there was an emphasis on a product’s performance and economics versus environmental issues as the three major forces and built on the well-known study by Lancaster (1966) for defining products as bundles of attributes. Even though Chen (2001) segmented the markets into ordinary and green customers, he did not consider product architecture as an essential element for product design.

The review of the literature on product design reveals that no research has been performed to investigate the impact of regulations on the product architecture so that viable alternative architectures can be configured. Yet, obeying governmental laws and regulations is of the utmost importance for any company and should be explicitly considered in product design. Previous researchers have studied either a specific product’s design change aspects (rather than a holistic approach) or have indirectly addressed optimal architecture without considering regulatory effects. This study explicitly considers the regulation effect on product architecture and identifies economic, performance, market and industry factors that affect architecture choice. By allowing trade-offs among them, generating viable alternative designs is possible. It should be noted that factors affecting product architecture choices are very diverse and application dependent. These factors have not been fully investigated in earlier research and need a comprehensive approach to understand and analyze (Campagnolo & Camuffo, 2010). Another important trade-off considered in this research is the impact of a regulation’s intensity and frequency on product architecture. For example, a high frequency of regulations can affect economies of scale (Pashaei & Olhager, 2017). This is because frequent regulations lead to a short lifetime for products, which reduces total sales for an architecture. As another example, one may consider situations where the material requirements change for a product, leading to a change in the unit cost. Thus, regulatory factors could directly affect architectural choice and indirectly affect it through affecting other factors.

Some studies emphasize corporate social responsibility (CSR). CSR can be defined as the actions taken by firms so that the business has a positive impact on society (in this case through following environmental regulations). These studies suggest that CSR should be treated explicitly in the firm’s strategy (see Porter & Kramer, 2006; Simchi-Levi, 2010). These studies indicate that CSR is certainly a value-added component and can create ample business opportunities. Based on the previous discussion, an exhaustive list of major factors that affect product architecture choices is provided in Table 3.

Table 3. Factors and sub-factors affecting product architecture

Factors	Sub-factors	Literature
Generic	Components	Ulrich, 1995,
	Interfaces	Schilling, 2000,
	Degree of Coupling	J.H. Mikkola, 2003,
	Substitutability	J. Mikkola, 2006,
		Simchi-Levi, 2010
Economic	Economies of scale (market size)	Ulrich, 1995,
	Economies of Scope (Product Variety,	Pashaei & Olhager, 2017
	Market segments)
	Unit cost
	Durability of the product
Technical	Performance	Ulrich, 1995,
	Product Upgrades (Product Change)	Sommer & Iyer, 2007,
	Key technologies	Lowenberger et al., 2012,
	Size of the product	Mauer et al., 2013
	Weight of the product
	Output power level (e.g. lighting power, cooling power)
	Noise
Supply chain	Number of key suppliers	Lee & Sasser, 1995,
Characteristics	Number of capable plants	Takeishi & Fujimoto, 2001,
And design	Suppliers capabilities	Fixson, 2005,
	Assembly plant focus	Howard & Squire, 2007,
	Component plant focus	Ülkü & Schmidt, 2011,
	Distance between key suppliers and	Simchi-Levi, 2010,
	Component plants	Pashaei & Olhager, 2017
	Distance between assembly plants and
	Markets
Environmental	Intensity of regulation demands	Nadel, 2002,
Regulation	Frequency of regulation rules	Porter & Kramer, 2006,
Characteristics	Materials change requests	Sommer & Iyer, 2007,
		Lowenberger et al., 2012
		Simchi-Levi, 2010
Other	Organizational Structure	Sosa et al., 2004
	Ease of replicability (imitability)

The literature search summarized in Table 3 shows:

• Products are not solely modular or exclusively integral. Products often have some degree of modularity and integrality in them.

• Contemporary organizations take a holistic approach to determining the optimal architecture for products. To this end, they include a few of the sub-factors (if not all) for each factor presented in Table 3.

• Frequency of the introduction of new regulations as well as the frequency of the introduction of new products will have an impact on the extent of modularity. A higher frequency of change in regulations or in the introduction of new products favors a higher degree of modularity. This is because modularity provides a high degree of product flexibility.

• An integral architecture for the product is preferred when the frequency of change is low.Thus, less flexibility is needed for a foreseeable time period.

• Environmental factors play a significant role in product design. Most notably, pressured by government regulations and competition, contemporary manufacturers treat environmental factors explicitly and extend it to include the impact of the entire supply chain on the environment.

• Corporate social responsibility must be treated explicitly and can have a significant business advantage for the firm.

The forthcoming sections provide a brief overview of AHP followed by a detailed outline of the proposed methodology with an illustrative example.

3. Analytic Hierarchy Process (AHP)

Introduced by Saaty (1980), AHP is an approach to solve complex problems by breaking them down into their parts. So, at the top level, the overall objective is specified, then criteria are determined at intermediate levels. While criteria can consist of factors and sub-factors forming intermediate levels, the lowest level is reserved for the alternatives which are simply the options that, from among them, the decision-maker will choose the best one.

The elements of the pairwise comparison matrix (PWC) are chosen by the decision-maker(s) from the scale (Table 4)(suggested by Saaty, 1980).

Table 4. Saaty scale for pairwise comparison matrix

= 1	if two criteria are equally important
= 3	if one criterion is moderately more important then the other
= 5	if strongly more important
= 7	if very strongly more important
= 9	if extremely more important
= 2,4,6,8	when a compromise between the above values is needed
	for all and = 1,2,.,

Solving the PWC matrix and obtaining Eigen values provides normalized priority weights for various criteria and alternatives. Free software available on the Internet can easily ascertain the priority weights. In addition, an inconsistency index (Saaty, 1980) is computed to check for any bias in the judgement when the PWC matrices were constructed. Thus, a high inconsistency index for a pairwise comparison matrix reflects a bias in the subjective judgements of the decision-maker(s) when constructing the PWC matrix and requires revising the entries of the matrix. An index value of 10% or less is acceptable and denotes consistency in the judgements.

4. An outline of the proposed methodology

Based on the literature review presented in the previous section, it is evident that product architecture related decisions involve complex models and multiple conflicting objectives/criteria. Different products share some common components that a change in any of them might affect various products. Eliminating a component from a product may require the design of a completely new component. Consequently, it is often the case that some manufacturing processes will be impacted by architectural changes. In the methodology presented in this study, it is assumed that the manufacturer will select from a discrete number of possibilities for a product’s architecture. In view of the recommendations for the product architectures articulated in the previous section, this study, without loss of generality, considers the following three alternative architectures:

(1) highly modular, (2) moderately modular, and (3) highly integral

Figure 1 presents an overview of the proposed methodology followed by a detailed explanation of the steps involved in the methodology. AHP is utilized by the suggested methodology to help decision-makers incorporate factors that impact various architectural alternatives. It also allows them to choose the optimal architecture.

Figure 1. Outline of the proposed methodology

Figure 1 shows the steps of the proposed methodology and are explained as follows:

Step 1. Consider the value proposition of the firm’s and customers’ perception of the products manufactured by this firm. Determine the main objective(s) of the design/redesign of the product under consideration. It is important at this step to set the right objectives. The objectives must be derived by (a) the customer value proposition that the firm attempts to offer and (b) the customers’ valuation/perception of the product to be designed (redesigned).

Consider the product innovation (e.g. in light of new government regulations, it may be an excellent opportunity to innovate a product to meet the newly introduced regulations).

Consider other factors such as value-added features and services, cost savings, future growth, customer relationship (e.g. design an innovative product that will prevent customers from switching to competitors), and impact on the customer experience.

Step 2. Using the information obtained in Step 1 and the factors and sub-factors identified in Table 3, determine the pertinent criteria to be considered when evaluating different architectures and develop viable alternative product architectures.

Step 3. Use the information in Steps 1 and 2 to construct an AHP hierarchy.

Step 4. Referring to the AHP hierarchy (Step 3), construct a pairwise comparison matrices for the criteria, sub-criteria and the alternatives. Determine the weights for all the pairwise comparison matrices. Augment all the weights to calculate the overall weights for the alternative architectures identified in Step 2.

Step 5. Select the best architecture based on the overall weights obtained for all alternatives. The alternative with the highest weight is the preferred architecture.

Step 6. Perform sensitivity analysis to investigate the trade-offs among different alternatives.

5. An illustrative example

Getachoo Manufacturer (GM) is a large-sized U.S.-based company that makes air conditioners and must follow different governmental regulations (that usually change every three or four years) in addition to other economic and technical factors. GM faces huge challenges to maintain the high quality of its product (e.g. performance, variety, durability). Meanwhile, the company must meet the environmental regulations and uphold its excellent Customer Service Record (CSR). GM is an excellent candidate for the illustrative example in this study due to the importance and widespread use of air conditioners. For example, air conditioners usually consume the largest amount of energy among home appliances and compared to other home appliances, they relatively go through more frequent changes in environmental, economic and technical requirements.

In a recent meeting, a top level committee composed of GM’s VP production and operations, production managers, quality manager, design engineersagreed to follow the steps of the proposed methodology to identify viable architectures for its product family so as to determine an optimal architecture for it. (Hereby, it is referenced as ‘the committee’). Below is a systematic discussion of the activities that took place to execute the proposed methodology.

The committee members used Table 4 to determine the elements of the PWC matrix in any of the following steps that involved constructing a PWC matrix.

Step 1. The customer value proposition of GM has always been ‘by buying our air conditioners, you get the best value for the price that you pay. GM is determined to use all the newly required regulations as well as technical and economic factors to deliver to customers the best value for the money they pay for GM’s air conditioners. To this end, in the redesign of the current air conditioners, ‘product innovation’ is the rule of the game for GM.

Step 2. The committee decided that for the environmental factors, intensity, frequency and refrigerant change are used as the sub-factors. Following the new EPA regulations, GM decided to use Puron (R410A) as the new refrigerant in their new air conditioners,discontinuing Freon refrigerant (HCFC-22) due to its harmful effect on the ozone. The committee also decided to consider the unit cost and durability of the product as economic factors and sales volume as the sub-factor. Since an air conditioner’s variety was important for GM in providing its value proposition to the customers, sharing common components and a higher degree of modularity were desired. Thus, economies of scale and economies of scope are essential sub-factors. This also proves to be important given the findings of Pashaei and Olhager (2017).

Finally, the technical factors were included to represent performance and attribute level of air conditioners. For example, size of air conditioners was expected to double due to the 2006 energy efficiency rise to 13 SEER. Thus, weight is also expected to increase. As in Ulrich (1995), size and weight are important factors that can be influenced by product architecture. Another important aspect considered by the committee was noise level. The committee also decided that an air conditioner’s cooling power level is an important sub-factor that reflects the performance level. For example, it is reasonable to argue that units with high cooling power require more of an integral architecture since they allow for global rather than local optimization of the components of the product. Table 5 presents the factors and sub-factors for the redesign of GM’s air conditioners, as determined by the committee.

Table 5. Factors and sub-factors for the air conditioners example with acronyms

Level 1		Level 2
Environmental	Intensity	Frequency	Refrigerant	-
(ENV)	(INTN)	(FREQ)	Change (REF)
Economic	Economies of	Economies of	Durability	Unit Cost
(ECON)	Scope (SCP)	Scale (SCL)	(DUR)	(COST)
Technical	Cooling Power	Noise Level	Size (SIZE)	Weight
(TECH)	(PWR)	(NSE)		(WGHT)

Step 3. Using the criteria and sub-criteria identified in the previous steps, an AHP hierarchy was constructed (Figure 2).

Figure 2. AHP hierarchy for air conditioners

Step 4. The next step is to construct the pairwise comparison matrices for AHP to evaluate different architectures. Initially, in a subsequent meeting, the committee determined the relative importance of factors at all levels. Then for each pairwise comparison matrix, weights were calculated, and the inconsistency ratio was checked to assure reliability in judgements. In the end, all the weights were augmented to calculate the overall weight for the three alternative architectures: highly modular, moderately modular and highly integral. The architecture with the highest weight is the optimal choice. Table 6 shows the pairwise comparison matrix for the criteria. Referring to Table 6, since all entries are one, this implies that the committee considers the environmental, economic and technical factors as equally important for GM.

Table 6. Pairwise comparison matrix for criteria

CRITERIA	ENV	ECON	TECH
ENV	1	1	1
ECON	1	1	1
TECH	1	1	1

Table 7 provides the PWC matrix to obtain the weights (and importance ranking) for the sub-criteria for technical factors. Similarly, two other PWC matrices were constructed (not shown here) for the sub-criteria of the environmental and economic criteria.

Table 7. PWC matrix for technical factors

TECH	PWR	NSE	SIZE	WGHT
PWR	1	2	1/3	½
NSE	½	1	1/6	¼
SIZE	3	6	1	3
WGHT	2	4	1/3	1

Table 8 shows the PWC matrix, comparing the three alternative architectures. These are highly modular (HM), moderately modular (MM), and integral (INTG). They are considered in regards to Refrigerant Change (REF). Since there are eleven sub-criteria, the committee constructed ten additional PWC matrices. Similar to the PWC shown in Table 7, each of the additional ten PWC matrices that the committee constructed compared the three alternatives, considering one sub-criteria at a time.

Table 8. PWC matrix for alternatives considering Refrigerant Change

REF	HM	MM	INTG	.
HM	1	2	6
MM	½	1	3
INTG	1/6	1/3	1

Step 5. Select the best architecture based on the overall weights obtained for all alternatives. The alternative with the highest weight is the preferred architecture.

Table 9 provides the combined weights for the three criteria, considering the weights for all criteria and sub-criteria. So, as shown in the table, the integral architecture is the most suitable architecture because, considering all criteria and sub-criteria, it gets the highest weight of 37%. However, highly modular architecture weight (34%) is only marginally less than the integral architecture. The sensitivity analysis, performed in the next step, plays a decision support system role by providing a deeper insight into the robustness of the results and allowing for a ‘what-if’ analysis.

Table

Integral Architecture	37%
Highly Modular	34%
Moderately Modular	29%

Table 9. Overall priority scores for product architectures

Architecture	Overall Weight
Highly Modular	34%
Moderately Modular	29%
Integral	37%

Notice that the committee checked the inconsistency ratio for all the PWC matrices. Since all the ratios were below 10%, no significant inconsistencies were noted.

Step 6. Perform sensitivity analysis to investigate the trade-offs among different alternatives.

The suggested last step is sensitivity analysis. It helps decision-makers explore and evaluate other viable product architectures. In the last GM committee meeting, the members identified what they consider ‘three feasible and practical scenarios’. The first scenario considered the case of a manufacturer with a product that faces highly intense regulatory requirements. The second scenario dealt with a situation in which environmental factors are given relatively low priority (with voluntary rather than mandatory regulations). The third scenario was a case for a manufacturer that shifted the focus to economies of scope, considering it a predominant factor.

Scenario 1: Intensity of Regulations versus Frequency

As discussed previously, different appliances tend to have regulations with different characteristics. Some require relatively small changes, and some require radical modifications. Moreover, some appliances are under more frequent and ongoing sequence of rules and regulations, while others barely need to undertake any change, even in a decade. In this scenario, the GM committee investigate how increasing the intensity of upgrades, while decreasing its frequency, would affect the optimal architecture. This is a highly conceivable scenario, since decreasing the frequency of updates is very likely for some products that have reached some maturity in their efficiency levels (Nadel, 2002).

The committee adjusted the PWC matrix, comparing the environmental sub-factors against each other (Table 10).

Table 10. Updated PWC matrix – scenario 1

ENV	INTN	FREQ	REF
INTN	1	2	4
FREQ	½	1	2
REF	¼	½	1

The AHP result in Table 11shows the overall weight for the three alternative architectures. Table 11 indicates that changing the intensity and frequency of regulations increases the gap between modular and integral architectures. In fact, in this scenario, the integral architecture prevails as the best architecture.

Table 11. Overall priority scores – scenario 1

Architecture	Overall Weight
Highly Modular	31%
Moderately Modular	29%
Integral	39%

One further remark regarding Scenario1 is that for certain products, the frequency of regulations is chosen in relative harmony with typical life cycles of products. In such situations, the relative importance of the frequency of regulations could be significantly undermined in the analysis compared to the intensity of the required change.

Scenario 2: Voluntary Environmental Rules

Given the highly competitive market for home appliances, occasionally when regulation is not mandatory, some manufacturers may decide not to adopt the newly suggested upgrades by regulatory authorities. The GM committee decided to assign a lower weight to the environmental factors compared to both economic and technical factors, as shown Table 12.

Table 12. Updated PWC matrix – scenario 2

CRITERIA	ENV	ECON	TECH
ENV	1	½	½
ECON	2	1	1
TECH	2	1	1

As a result of favoring the economic and technical factors over the environmental factor (Table 13), once more the integral architecture prevails (similar to Scenario 1). However, it should be noted that a modular architecture would be expected to be the most likely choice if the frequency were more important in the environment pairwise comparison matrix.

Table 13. Overall priority scores – scenario 2

Architecture		Overall Weight
Highly Modular		33%
Moderately Modular		29%
Integral		38%

Scenario 3: Economy of Scope versus Economy of Scale

Under this setting, the GM committee considered a reduction in the appliance’s unit cost by sharing more common modules among its product lines and focusing on the gains provided by the economy of scope. Table 14 provides the PWC matrix for this scenario.

Table 14. Updated PWC matrix – scenario 3

ECON	SCP	SCL	DUR	COST
SCP	1	6	9	1
SCL	1/6	1	3	1/6
DUR	1/9	1/3	1	1/9
COST	1	6	9	1

Table 15 shows the new AHP-weights obtained for the three architectures based on the new preferences of the GM committee. For this scenario, as indicated in Table 15, the highly modular architecture becomes more significant, and its total weight is now equal to the weight of the integral architecture.

Table 15. Overall priority scores – scenario 3

Architecture	Overall Weight
Highly Modular	35%
Moderately Modular	30%
Integral	35%

6. Discussion and conclusion

The proposed methodology is a unique tool to assist in identifying and choosing the best product architecture. It is a multi-criteria approach and allows various criteria and sub-criteria to be considered in the architecture selection process. Using an illustrative example, several product architectures were developed and evaluated using the methodology. Sensitivity analysis, as the last step, can assist product designers to gain more insights into architectural alternatives and make appropriate changes, considering the degree of modularity desired in the final product. Although some prior studies attempted to characterize the optimal product design, they either did not take a holistic approach or ignored a very important factor: the frequency of change of governmental regulations. This study overcomes the deficiencies of previous models.

A limitation of the proposed methodology is that the recommendation for the optimal architecture is largely dependent on assumptions about when integral versus modular architecture is preferable in view of factors and sub-factors. Another limitation can be the tedious process of constructing numerous PWC matrices. This latter drawback, however, may be a strength, considering that constructing numerous PWC matrices forces those involved in the process to gain a deeper understanding of the firm’s objectives, priorities, and alternative architectures. This in turn will foster a better understanding of the trade-offs among various alternative architectures.