Journal of Operations Management 16 1998.

569581

Technical note

Determination of an optimal set of design requirements using house of quality

Taeho Park
b

a, )

, Kwang-Jae Kim

a Organization and Management Department, San Jose State Uniersity, One Washington Square, San Jose, CA 95192, USA Department of Industrial Engineering, Pohang Uniersity of Science and Technology, San 31, Hyoja-Dong, Nam-Gu, Pohang, Kyungbuk, 790-784, South Korea

Received 23 May 1996; accepted 1 September 1997

Abstract Quality Function Deployment QFD. has been used to translate customer needs and wants into technical design requirements in order to increase customer satisfaction. QFD utilizes the house of quality HOQ., which is a matrix providing a conceptual map for the design process, as a construct for understanding Customer Requirements CRs. and establishing priorities of Design Requirements DRs. to satisfy them. Some methodological issues occurring in the conventional HOQ are discussed, and then a new integrative decision model for selecting an optimal set of DRs is presented using a modified HOQ model. The modified HOQ prioritization procedure employs a multi-attribute decision method for assigning relationship ratings between CRs and DRs instead of a conventional relationship rating scale, such as 139. The proposed decision model has been applied to an indoor air quality improvement problem as an illustrative example. q 1998 Elsevier Science B.V. All rights reserved.
Keywords: Quality; Product development; Mathematical programming

1. Introduction Quality has become one of the critical competitive strategies in todays global market. To ensure the improvement of quality and productivity, firms have adopted Total Quality Management TQM. as a key element of their business goals and initiated the use of TQM methods, such as Quality Function Deployment QFD., design for manufacturability, and statistical process control. Among these TQM methods, QFD has been used to translate customer needs and wants into technical design requirements by integrat)

Corresponding author.

ing marketing, design engineering, manufacturing, and other relevant functions of an organization Akao, 1990; Ansari and Modarress, 1994.. QFD originated in 1972 at Mitsubishis Kobe shipyard site, and then Toyota and its suppliers developed it further for a rust prevention study Hauser and Clausing, 1988; Wasserman, 1993.. After the concept of QFD was introduced in the US through auto manufacturers and part suppliers Sullivan, 1987., many US firms, such as Procter & Gamble, Raychem, Digital Equipment, HewlettPackard, AT & T, ITT, GM and Ford, applied QFD to improving communication, product development, and measurement of processes and systems Ansari

0272-6963r98r$ - see front matter q 1998 Elsevier Science B.V. All rights reserved. PII: S 0 2 7 2 - 6 9 6 3 9 7 . 0 0 0 2 9 - 6

570

T. Park, K.-J. Kim r Journal of Operations Management 16 (1998) 569581

and Modarress, 1994; Bounds and Yorks, 1994.. Furthermore, as cited by McElroy 1989., QFD can be applied to whatever process you have control over: new product design, business plans, engineering proposal systems, even reducing die transition time, said Norman Morrell, corporate manager of quality-product reliability at Budd. The basic concept of QFD is to translate the desires of customers into product design or engineering characteristics, and subsequently into parts characteristics, process plans and production requirements. Each translation uses a matrix, called the house of quality HOQ., for identifying Customer Requirements CRs. and establishing priorities of Design Requirements DRs. to satisfy the CRs Hauser and Clausing, 1988.. See Griffin and Hauser, 1993 for a discussion on the issues of identifying, structuring, and providing priorities for CRs.. HOQ is presented in a matrix form which shows: 1. CRs in rows and DRs in columns, 2. their relationships within the matrix, and 3. the correlations i.e., dependencies. of DRs at the top of the matrix See Fig. 1.. The conventional HOQ employs a rating scale e.g., 139, or 159. to indicate the degree i.e., weakmediumstrong. of strength between CRs and DRs. Although the conventional approach to prioritizing DRs is easy to understand and use, there are several methodological issues associated with it, namely 1. determination of the degree of impor-

Fig. 1. A typical HOQ matrix with a 139 rating scheme.

tance of CRs, 2. assignment of the relationship ratings between CRs and DRs, 3. adjustment of the relationship ratings between CRs and DRs, called normalization, in order to insure a more meaningful representation of the DR priorities, 4. incorporation of the correlations between DRs to a decision process for determining appropriate DRs, and 5. consideration of cost trade-offs among DRs. Some research have been done to resolve these methodological issues. Lu et al. 1994. and Armacost et al. 1994. applied the Analytical Hierarchy Process AHP. Saaty, 1980; Saaty and Kearns, 1985. to determine the degree of importance of CRs. It should be noted that AHP is a powerful and widely-used multi-criteria decision-making technique for prioritizing decision alternatives of interest Saaty, 1980.. Wasserman 1993. presented a linear integer programming model for maximizing customer satisfaction subject to a cost i.e., limited budget. constraint with a linear function and a procedure for normalizing the relationship ratings between CRs and DRs. However, further studies are necessary on development of a better relationship rating scheme between CRs and DRs other than the conventional ordinal rating scale, and integration of the correlations between DRs to a decision model for determining appropriate DRs to satisfy CRs. The relationship ratings used in the HOQ prioritization process have a strong and direct impact on the technical importance ratings of DRs. Thus, the choice of a relationship rating scheme is critical in QFD applications. Thirty QFD applications that were published in the literature were selected at random and examined as to their rating schemes. The study revealed that various rating scales have been employed; seventeen used a 139 scale, five used 135, and the remaining eight used other scales such as 159, 124, or 169. None of the applications provided an explicit justification for the choice of such a rating scale. There is no scientific basis for any of the choices, added Cohen 1995.. The conventional relationship rating scheme primarily shows ordinal ranks of relationships between CRs and DRs rather than a continuum of rating values indicating a sliding scale of relationship strength. As a result, the absolute importance ratings of DRs in the conventional HOQ present ordinal importance ranks of DRs in their contribution to customer satis-

T. Park, K.-J. Kim r Journal of Operations Management 16 (1998) 569581

571

faction. Thus, it is rather more meaningful and necessary to show the differences of DRs in contributing to customer satisfaction in their magnitude than just in the ordinal importance ranks. The absolute importance ratings of DRs derived from the relationship ratings with cardinal numbers can be used in further analyses requiring the magnitude of each DRs importance in satisfying CRs e.g., selection of appropriate DRs to maximize customer satisfaction with a limited budget.. Although the conventional QFD obtains information about correlations between DRs in the triangular-shaped correlation matrix at the top of HOQ, the correlation information was not used in calculating priorities of the DRs and determining appropriate DRs for a design problem. Since some DRs are in fact highly correlated with others i.e., if the level of one DR is changed, the levels of the other DRs are changed collaterally., it is necessary to devise a mechanism for accommodating the dependencies of DRs in calculating importance ratings of DRs, and to incorporate the correlation between DRs into the decision process of determining appropriate DRs subject to some organizational constraints, such as cost and time. For example, when two DRs with a high correlation are selected at the same time, there may be cost savings in installing them in a product. Thus, in this paper, we will present: 1. a new integrative decision model for selecting an optimal set of DRs which incorporates all of the above issues into the conventional prioritization processes in QFD, 2. a modified HOQ with a better approach for assigning relationship ratings between CRs and DRs other than the current rating scheme, and 3. a quadratic integer programming formulation for considering cost trade-offs among DRs and correlations between DRs during the process of selecting an optimal set of DRs. An indoor air quality improvement problem is presented to provide a thorough numerical understanding of the proposed new integrative decision model using the modified HOQ. 2. Description of HOQ Since QFD was introduced to US firms in 1983, it was applied to many industry problems e.g., product design, strategic planning, renovation of a computer workroom facility, and improvement of customer

service. using the conventional QFD processes for prioritizing DRs. See Maddux et al., 1991; Griffin, 1992; Mallon and Mulligan, 1993; Ansari and Modarress, 1994; Glushkovsky et al., 1995; Lockamy and Khurana, 1995 for some recent application examples.. An HOQ typically contains information on CRs, relative importance of the CRs, DRs for satisfying the CRs, relationships between CRs and DRs, and correlations between DRs, as shown in Fig. 1. In conventional QFD applications, a cell i , j . in the relationship matrix of HOQ i.e., i th row and j th column of HOQ. is assigned 1, 3, 9 or 1, 5, 9. to represent a weak, medium, or strong relationship between i th CR called CR i . and j th DR called DR j ., respectively. The absolute and relative importance of DRs are computed using the relative importance of CRs and the relationship ratings i.e., 139 or 159.. For each DR, the absolute importance rating is computed as:
m

AI 1 s

Wi R i j ,
is 1

1.

where AI j s absolute technical. importance rating of DR j , j s 1, . . . , n, Wi s degree of importance i.e., importance weight. of CR i , i s 1, . . . , m, R i j s relationship rating representing the strength of the relationship between CR i and DR j . The absolute importance rating can then be transformed into the relative importance rating, RI j , as AI j RI j s n . AI k
ks 1

The larger the RI j , the more important is DR j . Thus, without consideration of any other constraints e.g., cost and time., DRs should be incorporated into the product of interest in the order of their relative importance rating to achieve more customer satisfaction.

3. A new model for selecting design requirements in QFD In this section, a new integrative HOQ model will be presented to determine an optimal set of DRs. The new model, shown in Fig. 2, integrates the following

572

T. Park, K.-J. Kim r Journal of Operations Management 16 (1998) 569581

Fig. 2. A new integrative HOQ model.

two aspects into Wassermans QFD planning process model 1993. and Lu et al.s integrative HOQ model 1994.: 1. employing a new rating scheme for the relationship between CRs and DRs, using a most commonly-used multi-attribute decision method i.e., swing method., and 2. considering the correlation between DRs in the mathematical programming for selecting an optimal set of DRs. In Phase 1, CRs, pairwise comparison of CRs, DRs, correlations between DRs, organizational resource constraints such as budget, and benchmarking of the product should be collected using the conventional procedures of HOQ, AHP, and benchmarking. Since the conventional procedures of the methods

can be obtained from the literature, their detailed descriptions are not included here. See Akao, 1990 and Cohen, 1995 for HOQ; Saaty, 1980 and Saaty and Kearns, 1985 for AHP; Lu et al., 1994 and Camp, 1989 for benchmarking.. Rather, we will present a step-by-step procedure of a method i.e., swing method. used for obtaining relationship ratings between CRs and DRs. Methods for eliciting utility weights in multi-attribute decision making can be used as a better alternative to the subjective weighting scale e.g., 139 scale. which has been used in past practical problems. Considering that the relationship ratings between CRs and DRs in HOQ represent the contri-

T. Park, K.-J. Kim r Journal of Operations Management 16 (1998) 569581

573

bution of DRs to the satisfaction of CRs, methods of viewing weights as the relative importance of multiple attributes may be more appropriate for QFD applications. SMART Simple Multi-Attribute Rating Technique., originally presented by Edwards 1977., is one technique which assesses attribute weights through the responses provided by the decision maker concerning hisrher tradeoffs among the attributes. von Winterfeldt and Edwards 1986. presented a modified version of SMART for explicitly considering the range of consequences on the given attributes. The swing method, which is a part of the SMART procedures, is used in this paper to obtain the relationship ratings between CRs and DRs. Eppel 1990. showed the following steps of the swing method to elicit the attribute weights through an interaction with the decision maker: Step 1. Show the decision maker two alternatives. One leads to the worst consequence on each attribute, and the other one leads to the best consequence. Step 2. Ask the decision maker to imagine that he or she is under a condition of the worst alternative i.e., the one leading to the worst consequence on each attribute. and to rank the attributes, one at a time, by indicating which attribute will improve an objective utility function e.g., customer satisfaction. most if its level swings from the worst to the best consequence of the attribute. Step 3. Assign a relative value of 100 to the most significant attribute which was ranked first in step 2. Rate swings of all other attributes on a 0100 scale. Thus, an attribute whose swing leads to half the improvement compared to the most significant swing would get a weight of 50, and an utterly irrelevant attribute would get a weight of 0. Step 4. Normalize the weights so that they add up to one. The above steps can be applied to CRs, one by one, to assess the relationship ratings between CRs and DRs in QFD. The worst alternative in Step 1 is a situation where no DR will be included into a product or service to satisfy a CR suppose it is CR i . whose relationship ratings are being sought. Attributes should be replaced with DRs, and the weights obtained in Step 3 will correspond to the assessed relationship ratings for the CR i i.e., R i j s in the i th row of the relationship matrix in HOQ chart.. An

example of the swing method applied to a QFD problem is illustrated in Appendix A. After all data involved in the QFD problem are obtained, Phase 2 will calculate the degree of importance i.e., importance weight. of CRs from data collected through pairwise comparisons of CRs. The eigenvector method which has been used as part of AHP can be employed to determine the importance weights of CRs. See Saaty and Alexander, 1989 for details of the eigenvector method.. A computer software package, called Expert Choice Forman et al., 1985., may be used for this eigenvector method. In Phase 3, relationship ratings between CRs and DRs in the relationship matrix should be normalized to generate a more meaningful representation of the DRs. Contribution of each CR to customer satisfaction is represented by the importance weight assigned to the CR. Thus, through the normalization of relationship ratings its contribution should be distributed into the technical importance rating of DRs related to it in proportion to the relationship ratings between the CR and the relevant DRs Lyman, 1990.. Wasserman 1993. presented the following normalization procedure, which can accommodate correlations between DRs:
n

R i kg k j
R norm s ij
ks 1 n n

for i s 1, . . . , m ; j

R i k g jk
js 1 k s 1

s 1, . . . , n ,

2.

where g k j denotes an element of the correlation matrix representing the correlation between DR k and DR j . In Phase 4, absolute importance ratings of DRs should be calculated using the conventional HOQ calculation procedure shown in Eq. 1.. In Phase 5, an integer programming model for maximizing customer satisfaction by selecting appropriate DRs should be formulated as follows:
n

Max f x . s

AIj x j
js 1

s.t. g k x . F 0 for k s 1, . . . , l . xgX.

3.

574

T. Park, K.-J. Kim r Journal of Operations Management 16 (1998) 569581

where AI j s absolute technical importance rating of DR j , x j s 01 decision variable for DR j i.e., if DR j is selected, x j s 1. Otherwise, it is 0., x s a decision variable vector, x j 4 , j s 1, . . . , n, g k x . s k th organizational resource constraint, l s number of organizational resource constraints. The objective function in the above formula is to maximize a total absolute technical importance rating from selected DRs which represents the magnitude of customer satisfaction. When selecting DRs to implement, the conventional QFD does not take into account trade-offs between the amount of customer satisfaction achieved from the selected DRs and the use of organizational resources, such as cost and time, for the DRs. King 1987., however, stressed the need for considering the amount of organizational resources to be used because a targeted selling price of the product in its market should be set. Thus, Wasserman 1993. presented a simple linear cost constraint function for g x . to select the most appropriate DRs under a limitation of a given target cost as follows: g x . s c1 x 1 q . . . qc n x n y B, where c j is cost required to include DR j , and B is a given total target cost. He solved Problem 3. with a constraint of g x . s c1 x 1 q . . . qc n x n y B F 0 using the well-known Knapsack problem approach. DRs should be selected in a decreasing order of the technical importance ratingrcost. ratios until the total cost of selected DRs does not exceed the limited repair budget. In the case where dependencies i.e., correlations. exist among some DRs, some savings in resource consumption are most likely expected when two or more correlated DRs are simultaneously installed into a product or service design. Since most practical QFD problems, as Wasserman 1993. addressed, involve some degree of dependencies among DRs, it is more appropriate to express the g x . function in a quadratic form such that g x . s n js 1 c j x j y n n is 1 j ) i s i j x i x j y B , where s i j is saving of resource e.g., cost. usage associated with simultaneous implementation of i th and j th DRs. Once the QFD problem is formulated as a quadratic integer programming problem shown in 3., in Phase 6, the most appropriate DRs will be determined to satisfy CRs with the limited organizational resources. In this paper, Solver in Microsoft Excel was used to solve the quadratic integer pro-

gramming problem for selecting the DRs. See Salkin, 1975 for the details of quadratic integer programming..

4. Application to building indoor air quality improvement 4.1. Description of the building indoor air quality problem During the last decade, concerns have been expressed regarding indoor air quality for nonindustrial buildings see Banham, 1994; Buckler, 1994; Eaton, 1995.. Such indoor air quality problems, classified as Sick Building Syndrome SBS. or Building Related Illness BRI., have occurred in all types and ages of structures, namely from newly constructed buildings to renovated facilities or old buildings. Today, most commercial buildings are mechanically ventilated with virtually no natural ventilation from the outside. Moreover, designers, owners, and operators have tried to minimize unwanted air infiltration because rising energy costs are the primary cause of the tight building syndrome, along with various chemicals present in building materials and furniture. In addition to other physical factors in the workplace e.g., lack of windows, temperature variations, and poor ergonomic design., poor indoor air quality is viewed as a major building occupant complaint Banham, 1994; Lunau, 1993; Moseley, 1990.. Recently, employees working in the Business Tower BT. building at San Jose State University SJSU. suffered from poor indoor air quality and expressed complaints of being sleepy and tired at work. Stuffiness was a common complaint, along with the presence of flies within the stairwell entrance vestibules; a large accumulation of dust particles in the building was also found I & A Engineering, 1994.. The following is a list of problems identified: 1. stuffiness, 2. temperature, 3. dust particles, 4. ventilation, 5. odors, 6. housekeeping, and 7. flies. Thus, a study was conducted in 1994 to test indoor air quality for gases, volatile organic components, and contents of particulate matters, such as dust. The BT building with a closed environment that is, all windows are sealed. was designed in 1968 and

T. Park, K.-J. Kim r Journal of Operations Management 16 (1998) 569581

575

completed by 19701971. The ten-story buildings Heating, Ventilation, and Air Conditioning HVAC. system consists of a double duct constant volume air distribution system. The air supply system has a floor-mounted double inlet centrifugal fan with a cooling coil and heating coil arranged side by side on the ground floor. The separate heating and cooling ducts are then routed into the shaft to the 10th floor. The ground floor air intake system is composed of a grade level outside air intake louver and pneumatically-controlled motorized intake dampers. The air is then drawn through a prefilter and final bag filters before entering the supply fan room. The return air system employs a non-ducted return air plenum pregnable, which is also typical of HVAC system designs today. Return air from each thermal zone enters the ceiling cavity and is drawn back to the central shaft return air opening. The exhaust air is delivered to a plenum where it is then relieved out of the building through the exhaust louverrdamper combination, or recirculated into the mixed air plenum where it is mixed with outside air before entering the fan system. The indoor air quality study for the BT building recommended the following actions to improve the mechanical BT HVAC operating systems, resulting in a better working and health environment for occupants in the building I & A Engineering, 1994.: 1. Air Plenum: All linear boards should be removed from supply plenum walls and replaced with new 2 in. of 3 lb density matt faced acoustic lining. 2. Door Seals: All fan plenum door seals shall be removed and replaced with new tight fitting weatherstripping to reduce air leakage and uncontrolled infiltration. 3. An eighteen-gage galvanized sheet metal condensation drain should be installed under the chilled water cooling coils. 4. Dampers: The air damper motors should be tested andror replaced as necessary. 5. Plumbing Risers: The shaft should be provided with 2-h fire-rated access panels at the toilet room wall adjacent to the shaft at each door. 6. Air Delivery Systems: Air delivery systems, including air supply fan and air return fan, should be upgraded with more HP motors. 7. Janitor Closet Ventilation: A dedicated roof-

8.

9. 10.

11.

12.

13.

mounted exhaust fan should be installed to provide ventilation of the janitors closet on each floor. Duct Distribution System: A duct distribution system should be improved to prevent air leakage. Duct Plenums and Fans: Duct plenums and fans should be cleaned. Direct Digital Control DDC. System: The existing standard profile control system should be replaced with a DDC system for more efficient air control and delivery. HVAC Louvers: All damper motors, existing insect screens, and the vestibule ventilation exhaust fan should be replaced with new units. Carbon Dioxide Sensor System: A CO 2 monitoring station with sensors should be installed in the return air plenum. Housekeeping: BT janitorial services should be reviewed and improved to ensure the building is clean at all times. In addition, any chemical cleaning detergents should be removed from the building.

4.2. Numerical results To determine the priorities of implementing the above recommendations, the new HOQ prioritization process described in Section 3 was employed. A customer study involved assessing the judgments of department secretaries in their views on the significance of problems caused by poor indoor air quality. The customer study was conducted using a pairwise comparison method in the AHP data collection process. Since a group of eight secretaries working daily in the BT building participated in the survey, a geometric mean which is an 8th root of the product of judgments provided by eight individuals was used to combine group judgments. It should be noted that the secretaries are good surrogates for those who use the building because they spend more time in the building than anyone else. Table 1 shows the results of the customer study done, using a pairwise comparison method in the AHP data collection process, and a geometric mean approach to combining group judgments. Eigenvalues of the judgment matrix in Table 1, which are the importance weights of CRs, are then calculated as

576

T. Park, K.-J. Kim r Journal of Operations Management 16 (1998) 569581

Table 1 An important matrix for CRs obtained using pairwise comparison. The numbers indicate geometric means of judgments of eight secretaries Temperature Dust Ventilation Odors House Flies Stuffiness 1.0 Temperature Dust Particles Ventilation Odors Housekeeping 2.7 3.5 2.2 1.2 0.82 1.1 1.2 0.54 1.8 1.1 1.6 0.63 1.3 2.0 2.4 1.2 1.4 1.8 1.7 2.4

i.e., DR 1 . and replacing all fan plenum door seals with new ones i.e., DR 2 . require US$18,000 and US$12,000, respectively, when each of them is completed separately. When both of them are included in a repair contract, US$4500 out of US$30,000 is discounted because of savings in time. Table 2 illustrates cost savings from simultaneous installation of two DRs. Since there are savings from simultaneous installation of two DRs, the BT building indoor air quality problem should be solved with a quadratic integer programming technique as follows:
16

0.202, 0.187, 0.085, 0.152, 0.157, 0.132, and 0.084, respectively. Since the consistency ratio turns out to be 0.04, judgments of the importance of problems are acceptable. It should be noted that the value of consistency ratio should be around 10% or less to be acceptable Saaty and Kearns, 1985.. Fig. 3 presents an HOQ matrix for the BT building indoor air quality problem, including 1. degrees of importance of CRs as obtained from the AHP analysis, 2. normalized relationship ratings between CRs and DRs obtained using swing method and normalization of relationship ratings by Eq. 2. in Section 3, 3. correlation between DRs, and 4. cost required to install the DRs. It should be noted that two engineers knowledgeable about air quality participated in providing information about relationships between CRs and DRs, correlation between DRs, and costs. According to the results of prioritization of DRs, it is found that upgrading an air delivery system i.e., air supplyrreturn fans. is most important for improving building occupants satisfaction with indoor air quality in the BT building, and the installation of a CO 2 monitoring station with sensors is least important. If a repair budget is enough to complete all recommendations, the problem will become very trivial. As mentioned earlier, however, when available organizational resources e.g., budget and time. are limited, as in our BT indoor air quality problem, a further analysis is necessary to select which DRs should be completed. Each cost shown in Fig. 3 is required to complete an individual DR alone. However, there may be some cost savings when two DRs are installed at the same time. For example, upgrading air plenum walls

Max f x . s

AI j x j
js 1

s.t. c1 x 1 q . . . qc16 x 16 y s1 ,2 x 1 x 2 y s1 ,10 x 1 x 10 y s2 ,9 x 2 x 9 ys6 ,11 x 6 x 11 y s6 ,12 x 6 x 12 y s12 ,15 x 12 x 15 F B xgX,

4.
where AI j and c j s technical importance rating and cost of DR j shown in Fig. 3, x j s 01 decision variable for DR j j s 1, . . . ,16., B s budget available for improving indoor air quality in the BT building. The above quadratic integer programming with a repair budget of US$200,000 is solved using the Solver module in Microsoft Excel, and the following solution is found: 1. Objective function value of the total technical importance rating: 0.8448. 2. Decision variables: DR 1 s . . . s DR 9 s 1, DR 10 s DR 11 s 0, DR 12 s . . . s DR 16 s 1. 3. Total cost required: US$198,700. If the budget is at least US$450,000 i.e., US$513,000 16 . 16 16 .. js 1 c j y US$63,000 is 1 j ) i s i j , all DRs can be installed, resulting in the objective function of 1. Thus, 84.5% customer satisfaction can be achieved with only 44.2% US$198,700rUS$450,000. of total investment required. A sensitivity analysis for the budget constraint is presented in Fig. 4. The achieved level of customer satisfaction increased as a higher budget was allowed, with increments of US$25,000. However, the marginal rate of increase, in general, diminished as the level of the baseline budget became higher. For example, the increase of the budget from US$100,000 to US$125,000 increased the customer satisfaction

T. Park, K.-J. Kim r Journal of Operations Management 16 (1998) 569581

577

Fig. 3. An HOQ matrix for the indoor air quality problem for the BT building at SJSU.

578

T. Park, K.-J. Kim r Journal of Operations Management 16 (1998) 569581

Table 2 Cost savings occurring when two DRs are completed at the same time Pair of DRs DR 1 and DR 2 DR 1 and DR 10 DR 2 and DR 9 DR 6 and DR 11 DR 6 and DR 12 DR 12 and DR 15 Cost saving from simultaneous installation US$4500 US$10,200 US$4050 US$28,500 US$10,500 US$5250

by 9.4% 66.2 y 56.8., while the increase caused by the budget change from US$200,000 to US$225,000 was only 1.4% 85.9 y 84.5.. Likewise, as the customer satisfaction level increases by investment in technology, equipment, and training, more effort and investment are required to achieve the same level of

additional customer satisfaction. Once customer satisfaction is enhanced up to a certain level, there should be impetus to improve it further. In the building indoor air quality problem, the customer satisfaction level will remain at 88.5% without DR 11 , which will replace the existing standard profile control system with a direct digital control DDC. System. To improve the level further, a considerable amount of budget i.e., US$211,500 s US$240,000 cost for DR 11 alone. y US$28,500 savings.. is required. However, the control system conversion will improve the customer satisfaction level by 11.5%. The proposed model is compared with a Knapsack model shown in Wasserman 1993. which does not take cost savings into account. The Knapsack model has a different constraint from that of 4., which is c1 x 1 q . . . qc16 x 16 F B. Fig. 4 shows the results of both models. See Wasserman, 1993 for details of solving the Knapsack model.. Since the

Fig. 4. Sensitivity and performance analyses for customer satisfaction improvement over budget increment.

T. Park, K.-J. Kim r Journal of Operations Management 16 (1998) 569581

579

Knapsack model does not take into account an organizational constraint of cost savings, it cannot allow for installing additional DRs. which might be selected with cost savings. Thus, the Knapsack model results in no greater customer satisfaction than the proposed model.

5. Conclusions A mathematical programming-based approach to determining an optimal set of DRs to be included in a new product or service is presented in this paper by modifying a conventional prioritization process in QFD. The new decision model is to improve the conventional HOQ prioritization process for obtaining importance ratings of DRs by 1. determining the degree of importance of CRs through the AHP multi-attribute decision process, 2. assigning relationship ratings between CRs and DRs using the swing method rather than a conventional relationship rating scale of 139 or 159, and 3. normalizing the relationship ratings between CRs and DRs. Furthermore, it provides an optimal set of DRs by solving a mathematical programming problem formulated with the importance ratings of DRs, correlation between DRs, and information about available resources. The new relationship rating scheme using the swing method measures decision-makers opinions on the relationship between CRs and DRs more systematically and accurately than the conventional relationship rating scale used in HOQ. Since the new relationship rating scheme relies on a simple additive multi-attribute model, it is easy to use; thus, it is a very handy and useful tool for practitioners. In addition, it converts decision-makers thoughts of the relationship between CRs and DRs into a continuum of rating values so that the QFD problem can be formulated into a mathematical programming problem subject to limited resources e.g., budget. in an organization. As a result, the QFD problem could be extended to resource allocation problems in the operations management field. The new integrative decision approach has been applied to a building indoor air quality problem in the College of Business at San Jose State University. In fact, while we were developing the new QFD

method, some recommendations provided by an engineering firm were implemented for improving indoor air quality. Thus, the new QFD method could not be implemented to the final decision process for the air quality problem. However, all necessary stepby-step structured decision processes are included in the paper, which can help practitioners understand the entire decision process easily and implement it to their industrial problems. Although the proposed comprehensive model involves more effort e.g., data collection, mathematical model building, and computation. in comparison to the conventional HOQ process, it is worthwhile to invest such effort during the design stage in order to determine an optimal set of DRs. In other words, the investment will be justified with a better working environment, more customer satisfaction, and more market share resulting from better decision making.

Acknowledgements The authors would like to thank the editor and the reviewers for their helpful comments and suggestions.

Appendix A. Assessment of relationship coefficients using swing method The SMART procedure with swing method embedded can be applied to obtaining the relationship ratings between CRs and DRs in QFD. As an example, consider the HOQ chart in Fig. 1. A detailed step-by-step procedure for assessing the relationship ratings between CR 2 and DRs using the swing method is illustrated below. It is presumed that DR 1 , DR 2 , and DR 4 have important effects on the customer satisfaction of CR 2 , whereas DR 3 is not related to CR 2 , as manifested by the symbols recorded in the second row of the chart. Step 1: Show the design team two alternatives: one leads to the worst consequence with respect to CR 2 i.e., DR 1 s DR 2 s DR 4 s 0., and the other one leads to the best design condition i.e., DR 1 s DR 2 s DR 4 s 1.. Step 2: Ask the design team to imagine the worst design condition and choose a DR that would best

580

T. Park, K.-J. Kim r Journal of Operations Management 16 (1998) 569581 Banham, R., 1994. The growing concern over indoor air quality. Risk Manage. 41 6., 6062. Bounds, G., Yorks, L., 1994. In: Adams, M., Ranney, G. Eds.., Beyond Total Quality Management. McGraw-Hill, New York. Buckler, G.F., 1994. The effects of indoor air quality on health. NSNArImprint 41 3., 6093. Camp, R.C., 1989. Benchmarking. ASQC Quality Press, Milwaukee, WI. Cohen, L., 1995. Quality Function Deployment. Addison-Wesley, Reading, MA. Eaton, M., 1995. Indoor air quality requires integrated approach. Building Design Construction 36 4., 5256. Edwards, W., 1977. How to use multi-attribute utility measurement for social decision making. IEEE Transactions on System, Man, and Cybernetics, SMC-7, pp. 326340. Eppel, T., 1990. Eliciting and Reconciling Multi-attribute Utility Weights, Ph.D. Thesis, University of Southern California. Forman, E.H., Saaty, T.L., Selly, M.A., Waldron, R., 1985. Expert choice, decision support software. Pittsburgh. Glushkovsky, E.A., Florescu, R.A., Hershkovits, A., Sipper, D., 1995. Avoid a flop: use QFD with questionnaires. Qual. Prog. 28 6., 5762. Griffin, A., 1992. Evaluating QFDs use in US firms as a process for developing products. J. Prod. Innov. Manage. 9 3., 171 187. Griffin, A., Hauser, J.R., 1993. The voice of the customer. Marketing Sci. 12 1., 127. Hauser, J.R., Clausing, D., 1988. The house of quality, Harv. Business Rev., MayJune, 6373. I&A Engineering, 1994. Final report on indoor air quality of business tower building at San Jose State University, San Jose. King, B., 1987. Better designs in half the time: Implementing QFD in America, GoalrQPC, Methuen, MA. Lockamy, A. III, Khurana, A., 1995. Quality Function Deployment: a case study. Prod. Inv. Manage. J. 36 2., 5659. Lu, M., Madu, C.N., Kuei, C., Winokur, D., 1994. Integrating QFD, AHP, and benchmarking in strategic marketing. J. Business Ind. Marketing 9 1., 4150. Lunau, F.W., 1993. Air quality standards in offices: should they be health- or comfort-based? Indoor Environ. 2 4., 213216. Lyman, D., 1990. Deployment normalization, transactions from a second symposium on Quality Function Deployment, a conference co-sponsored by the Automotive Division of the American Society for Quality Control, the American Supplier Institute, Dearborn, MI, and GOALrQPC, Methuen, MA, pp. 307315. Maddux, G.A., Amos, R.W., Wyskida, A.R., 1991. Organizations can apply Quality Function Deployment as strategic planning tool. Ind. Eng. 23 9., 3337. Mallon, J.C., Mulligan, D.E., 1993. Quality Function Deployment a system for meeting customers needs. J. Construction Eng. Manage. 119 3., 516531. McElroy, J., 1989. QFD: building the house of quality, Automotive Industries, pp. 3032. Moseley, C., 1990. Indoor air quality problems. J. Environ. Health 53 3., 1922. Saaty, T.L., 1980. The analytic hierarchy process, McGraw-Hill, New York.

improve the design condition if its level changes from 0 to 1 that is called a swing.. Suppose the design team answers that they would swing DR 4 first because it is believed to have the most significant impact on CR 2 . Step 3: Assign 100 to DR 4 , which was chosen in Step 2. Rate all other DR swings on a scale of 0100. Suppose the design team rates the contribution of changing the levels of DR 2 and DR 1 from 0 to 1 to be 60 and 40, respectively, with regard to CR 2 . The rating for DR 3 should remain zero because it is irrelevant to improving CR 2 . Step 4: Normalize the ratings so that they add up to one. The normalized ratings can be used as the relationship ratings in the HOQ chart. The relationship ratings R 2 j s. associated with CR 2 are normalized as follows: R 21 s 40r40 q 60 q 0 q 100. s 0.2, R 22 s 60r40 q 60 q 0 q 100. s 0.3, R 23 s 0r40 q 60 q 0 q 100. s 0.0, R 24 s 100r40 q 60 q 0 q 100. s 0.5. The same procedure can be employed to assess the relationship ratings of other cells in the relationship rating matrix of HOQ. The intermediate relationship ratings which are output of Steps 2 and 3. and the normalized ones are summarized in a table shown below:
CRs DRs Relationship ratings DR 1 100 40 0 0 50 DR 2 0 60 0 60 70 DR 3 50 0 100 100 0 DR 4 0 100 0 0 100 Normalized relationship ratings DR 1 DR 2 DR 3 DR 4 0.67 0.00 0.33 0.00 0.20 0.30 0.00 0.50 0.00 0.00 1.00 0.00 0.00 0.38 0.62 0.00 0.23 0.32 0.00 0.45

CR 1 CR 2 CR 3 CR 4 CR 5

References
Akao, Y., 1990. Quality Function Deployment: Integrating Customer Requirements into Product Design. Productivity Press, Cambridge, MA. Ansari, A., Modarress, B., 1994. Quality Function Deployment: the role of suppliers. Int. J. Purchasing Mater. Manage. 30 4., 2835. Armacost, R.L., Componation, P.J., Mullens, M.A., Swart, W.W., 1994. An AHP framework for prioritizing customer requirements in QFD: an industrialized housing application. IIE Trans. 26 4., 7279.

T. Park, K.-J. Kim r Journal of Operations Management 16 (1998) 569581 Saaty, T.L., Alexander, J.M., 1989. Conflict Resolution: the Analytic Hierarchy Approach. Praeger Publishers, New York. Saaty, T.L., Kearns, 1985. Analytical Planning. Pergamon, Oxford. Salkin, H.M., 1975. Integer programming, Addison-Wesley, Massachusetts. Sullivan, L.P., 1987. Quality Function Deployment QFD.: the

581

beginning, the end, and the problem in-between, A collection of presentations and QFD case studies, American Supplier Institute. von Winterfeldt, D., Edwards, W., 1986. Decision analysis and behavioral research, Cambridge University Press, New York. Wasserman, G.S., 1993. On how to prioritize design requirements during the QFD planning process. IIE Trans. 25 3., 5965.