Integration of multi-criteria decision-making approaches adapted for quality function deployment: an analytical literature review and future research agenda

1. Introduction

The fast-growing and rapidly changing markets in today's competitive environment have made the product/service quality a key determinant for business success. In general, effective capturing of customer requirements (CRs) is a major advantage for product-oriented firms. In this regard, quality function deployment (QFD) is an efficient customer-oriented design tool that aims to meet customer expectations in a better way and enhance organizational capabilities while maximizing company goals. In 1972 in Japan, in order to develop a product, QFD emerged as a sufficient systematic tool in Mitsubishi heavy industry to translate CRs throughout the design, planning and implementation phases of the product (Li et al., 2014). Basic concept of QFD is to translate the customers' expectations into design requirements. QFD helps a company to make a trade-off between what the customer wants and what the company can afford to produce. The fulfilment of customer needs depends on features of the product/service which can be considered as engineering characteristics (ECs). It is important to determine the requirements that bring more satisfaction to the customer than others. Many industries have employed the QFD technique in various areas including transportation and communication, electronics and electrical utilities, software systems, manufacturing services, education and research, aerospace, agriculture, construction, environment protection, packaging and so on (Chan and Wu, 2002). The QFD in product development consists of four phases: product planning, parts design, process planning and process control planning. The relationship matrix in each stage between CRs and ECs is called the house of quality (HOQ) (Wu et al., 2020). The relationship between CRs and ECs reflects the impact of the fulfilment of the ECs on the satisfaction of the CRs. These relationships should be calculated by QFD team members. The relationship between CRs and ECs [and the relationship between the ECs themselves] are usually determined by linguistic variables. In other words, they are usually interpreted as symbols which should be converted into crisp numbers. The degree of these relationships is usually expressed on a scale system such as 0-1-3-9 or 0-1-3-5, representing linguistic expressions such as “no relationship”, “weak/possible relationship”, “medium/moderate relationship” and “strong relationship”. Table 1 shows the symbols and related weights between CRs and ECs in HOQ.

In order to develop the conventional QFD technique for establishing a more precise ranking process, multi-criteria decision-making (MCDM) methods can be employed. Such methods evaluate a set of alternatives taking different criteria into account under a deterministic or uncertain decision environment. If the data are based on human perception rather than accurate numbers, uncertain analytical tools such as fuzzy methods can be employed (Gündoğdu and Kahraman, 2020). Since 2004, the hybrid QFD-MCDM methods have received increasing attention from both practical and academic perspectives.

The current review investigates subjective categorization and qualitative analysis of the articles during 2004–2021 in QFD-MCDM combinations to achieve a new direction for future agenda and differentiate the review paper from other similar works. The questions which come up differ from previous studies as below answered in this research as reasons for proposing this bibliographic study.

This article contributes to quality management and customer requirement investigations and the various application QFD-MCDM models. This paper addresses the combination and frequency of use of MCDM and other models and tools with QFD and classes the application of the hybrid MCDM-QFD into three categories. The findings present a broader and more accurate focus on the evaluation tables of the hybrid QFD-MCDM models than the reviewed models. The research findings also indicate the geographical distribution and frequency of publication of new integrated methods in terms of the time horizon of the pointed-out topic.

This paper presents a conceptual classification scheme and classifies more than 59 relevant papers in this area published during the interval 2004–2021. The relevant papers are categorized and analyzed under different metrics. Based on the analytical results, some directions for future studies are recommended. The main objective of this paper is to fill the gap and to provide a recent state-of-the-art survey taking the hybrid QFD-MCDM methods into account.

The rest of this paper is organized as follows: In Section 2, first, the integrated QFD and MCDM methods are discussed. Then, in subsection 2.2, the incorporation of other concepts such as fuzzy theory into QFD-MCDM models is addressed. Afterward, the application of QFD-MCDM methods is discussed concerning their practical applications. Discussion remarks are presented in Section 3. Finally, Conclusion and directions for future research are given in Section 4.

2. Proposed advanced QFD models

In this paper, based on the search string, research has been done by the QFD and MCDM keywords on the “Publish or Perish” application considering the Web of Science, Scopus and Google Scholar databases to obtain the relevant articles for bibliographical review analysis. The initial result identified 997 items with a total citation of 38,433 considering the articles published from 2004–2021 in Journal and conference publications. After eliminating the duplicates, 881 articles were retrieved, and after excluding the non-English articles and conferences, 315 items were extracted. After screening the abstracts and whole manuscripts to be sure to not eliminate the relevant articles, the relevancy of the given topic focused on the integration of MCDM and QFD and the high H-index. Finally, 59 articles were referenced as literature review research.

As noted, this paper classifies the relevant literature into the following three categories.

1. QFD-MCDM Models.

2. Hybrid QFD-MCDM models. QFD-MCDM and combination of other approaches such as Fuzzy theory, Kano model and other tools.

3. Hybrid QFD-MCDM in various areas:

   • Healthcare, education, market segment and financial services.

   • Supplier selection.

   • Industries including aerospace, agriculture, construction, automotive, electrical and computer utilities.

Conventional QFD is more qualitative in terms of the parameters' importance to develop the QFD for establishing a more precise ranking process the MCDM methods can be employed. Kim et al. (2000) applied MCDM tools to optimize CRs with target value ranking the ECs in the QFD matrices. Since 2004, the hybrid QFD-MCDM methods have received increasing attention from both practical and academic perspectives because of QFD itself as a more qualitative tool and to improve the importance weight which is significant in QFD process matrices for evaluating the criteria and improving the output of the matrices as a quantitative result of the QFD. Alinezhad and Seif (2020) used MCDM to develop the imprecise CRs' ranking and proposed the decision-making tool to improve supplier selection by prioritizing and comparing them in the QFD technique.

Cui et al. (2021) applied the novel MCDM method namely SWARA (stepwise weight assessment ratio analysis) to the QFD in the manufacturing sector in terms of the large number of criteria which facilitated the data computation rather than traditional MCDM tools and decreased the time-consumption of calculations which if not based on the pairwise comparison doesn't need to have high consistency rate between the CRs or ECs.

It can be discussed that MCDM techniques are the most applied tools in the QFD because of the importance of evaluating the CRs and ECs. In the HOQ, to improve the ranking precision and decrease the computing and quantifying of the model, MCDM tools became strong combination of QFD. Then, it is important to provide a comprehensive literature review of hybrid QFD-MCDM methods. In previous reviews, various types of QFD and Fuzzy-QFD combinations used uncertainty instead of crisp numbers in the HOQ discussed. In the proposed taxonomy, first, we discussed the QFD-MCDM models which are the combination of QFD and MCDM tools. Then, in the second part, the hybrid models consist of QFD-MCDM and sustainability, uncertainty and other supplemental models with hybrid QFD-MCDM in terms of the applied methodology discussed. In the third classification, the hybrid QFD-MCDM methods are described according to how to apply and application area of the methods.

The objective of the three sections was to separate the method adopted in each study and the area of application in section 2.3. For instance, we mean the approach discussed in section 2.2 and the case study explained in section 2.3. Then, some studies discussed in sections 2.1 and 2.2 overlapping in section 2.3 (proposed method in the corresponding case study).

3. Discussion

Nowadays, the integrated MCDM-QFD methods have been extensively used to solve practical problems using functionalities and properties of MCDM methods. In this regard, it is important for decision-makers (DMs) to combine and extend MCDM techniques for certain objectives and requirements. To improve the QFD-MCDM models to a sufficient model, many recent studies have considered some concepts [including uncertainty, sustainability or supplement quality, manufacturing, optimization and statistical tools] which are represented in this paper as a hybrid QFD-MCDM method in various fields of applications in industries and services.

To prioritize customer needs as an important objective of the QFD method, and to form the relationship matrix between CRs and DRs and interrelationships between the DRs themselves, weighting and prioritization methods in the traditional model are not accurate. Consequently, the use of MCDM methods helps to achieve reliable results. The studies mentioned in this literature review during 2004–2021 show the application of QFD-MCDM methods and hybrid models, which include a combination of several methods using different tools in the final ranking of CRs and DRs. Furthermore, case studies that adapted the DEMATEL method on the roof of HOQ, as well as the application of fuzzy theory to increase accuracy and reduce vagueness and expert judgment, had better results in the final evaluation. In some studies, methods such as Markov and Kano model have been used to classify customer needs before entering the QFD process, which is effective in better identifying and classifying customer needs. This study showed, in recent years, that the decision-making tools used to prioritize, and weighting is much more accurate than the traditional ones. In addition, in combination with fuzzy theory, namely in the healthcare area where customers' demands are more qualitative, is more effective, and more accurate. In this study, we categorized the published articles into three main categories, (1) QFD-MCDM models which consists of models adapted QFD and MCDM. (2) hybrid QFD-MCDM includes the use of QFD and MCDM and other tools. (3) The application of hybrid models in different fields, which includes the use of the mentioned methods in the practical field. Table 4 presents papers were discussed the QFD model in four main classifications. The uncertainty adopted in studies to eliminate the vagueness in the voice-of-customer. Sustainability is a broad policy concept in the global public discourse that consists of environmental, economic and social dimensions. The MCDM tools have been discussed, and the supplement models, including mathematical, logical and other quality models can combine with QFD to improve the output of the hybrid model.

Table 4, the decision-making tools shows the most implemented with 68 times, so it can be concluded that this method is introduced as the sufficient method to obtain the optimum results. Also, applying uncertainty in studies (39 times) shows integrated models for translating CRs to DRs, in terms of eliminating the vagueness element, can increase precision.

Figure 3 shows the distribution of the various methods in the studies.

Figure 4 depicts the distribution of different decision-making tools. As seen the AHP method (19%) is the most adapted tool in the studies.

4. Conclusion and directions for future research

According to the broad search of articles in recent years, many studies have been interested in filling the research gap in the QFD model. To this end, due to the identifying strengths and weaknesses of the QFD method, many articles, particularly in the last 4 years (from 2017), have applied hybrid models to improve the traditional QFD model. The application of multi-criteria decision making in the QFD model has improved the traditional methods used in QFD, has also increased the precision and accuracy of evaluation, and ranking the CRs and thus DRs and makes the output of the model provide the design requirements for manufacturing a product or service. In the proposed hybrid models, most of the QFD applications are integrated with fuzzy theory. The entry of customer demands in a fuzzy manner (since the variables are fuzzy-verbal) and qualitative judgments of customers and experts have increased the flexibility and accuracy of the data.

The fuzzy theory has some limitations which two remarkable are the handling of imprecision data and the inherent inference of human thinking. If the data is imprecise in the system a human being cannot infer the knowledge or relation. The application of fuzzy theory requires complex calculations because the combination of fuzzy numbers is used with the decision method, which requires fuzzification and defuzzification. Therefore, creating appropriate software to facilitate the process can be effective in conducting case studies with extensive information. Depending on the type of CRs, the type of fuzzy numbers that may be used differs, but it can make the calculation difficult. For this reason, some research may remain theoretical, or the case study may not be fully implemented.

Investigations in many areas have focused on user-friendliness, sustainability, green manufacturing system and environmental concerns, which are some of the current important needs of customers compared to many years ago. On one hand, these new requirements led to limitations for manufacturers, but with increased customer satisfaction. But, on another hand, enterprises and factories have encountered problems (such as the complexity of the production process and the incompatibility in the interrelationship of DRs), which have led to the use of different machines and increased the final cost of the products or services. Thus, the application of hybrid models by using intelligent multi-criteria decision-making methods can decrease the overlap and conflict in operational processes, especially, in new areas that often seek to reduce fuel consumption and pollution and increase productivity and use new energy sources, such as the use of photovoltaic in automotive process design (Huang et al., 2019).

To implement hybrid models, it is necessary to consider the needs of customers and manufacturers properly. For example, in the health care case, if the target customers are defined as patients, the DRs should be determined differently from the physician and nurse's DRs as the final customers because their needs differ with the patients, but sometimes the experts and physicians can be considered as representative of the patient needs because they have enough knowledge about the tools and treatment process (Neira-Rodado et al., 2020).

In some studies, the Kano model has been used in research that is suitable as a tool to classify customer needs and should be combined with appropriate fuzzy and decision tools to achieve the desired results, which is also very useful in NPD. Kano model is used to show a better understanding of the most important product features from the customers' point of view and divides customers' needs into five main categories: Must-be, Attractive, One-dimensional, Indifferent, or Reverse attribute for a product (Chen et al., 2010). Therefore, this tool can be very useful in riddling CRs to enter the HOQ matrix and make the required classifications on the left wall of the HOQ. Using this tool can also help the manufacturer to avoid wasting extra time and money to meet customer needs.

In many studies, there is no mention of studying the competitors' evaluation section. The creation of this section in the HOQ matrix can be significant and have a difference in the ranking of CRs. The competitor evaluation department can also act as a benchmarking tool in creating business process competition patterns and the best performance from other similar companies, which requires comparing a similar product or service with the desired product and service. The competitive conditions in the market make it necessary to put some factors such as supplier selection, raw materials and transportation preferable to overall customers' preferences, which should provide a model to consider having competitive advantages and faster customer satisfaction. In this case, the Kano model can also integrate into the hybrid model to classify and highlight that customer needs are more basic and show the cause of customer dissatisfaction.

TRIZ can be used as a problem-solving tool to better introduce DRs (Chen et al., 2010). It can aid to engineers in product design, by improving the process to identify technical characteristics in the HOQ matrix. Studies have rarely shown how to use questionnaires and techniques used to assess customer needs. It is recommended to determine the validity and reliability of the questionnaire and sample size determination tools because this has a great impact on the data collection results. Also, establishing a control verification matrix can be effective in prioritizing methods and process control parameters and can provide good support to the prioritized elements of the QFD process design stage by controlling the critical factors in the final stage. After determining the critical processes and operations in the matrix rows, the requirements for controlling them to prevent errors and failures in the matrix columns of the process control planning are determined.

Due to the correlation matrix between engineering characteristics in the roof of the HOQ, many studies have not paid special attention to this sector, which may not cause attention to the impact of ECs on each other, causing inconsistencies in customer needs and conflict in the performance of some processes in the HOQ. Therefore, the use of multi-criteria decision tools including DEMATEL can greatly improve the calculation of the roof of the HOQ and obtain the weights of interrelationships of roof elements. Particularly, the DEMATEL accurately identifies the effect of elements on each other in the roof, and it helps to eliminate the contradiction effects in ECs.

According to the statistics of the research conducted in this study, the use of the hybrid QFD-MCDM methods including QFD, MCDM, and other applied tools is much more than individual QFD-MCDM. It shows the efficiency of hybrid QFD-MCDM in achieving the desired results which are mostly used in manufacturing products and industries, supplier selection and services, respectively. According to recent studies (after 2018), there are hybrid methods for Integrated product and process development and Integrated decision-support mechanisms for all product-related processes which improve the flexibility of the manufacturing system.

Due to the changing customer demands over time, also the advancement of technology, customers who are initially attracted to the new service characteristics take them for granted over time in most cases. Some customer demands from the design process and even the final product change. Therefore, it recommends using a dynamic QFD or other quality tools such as Kano Dynamic (An online platform to meet the updated needs of customers to update customer requirements) or the production of the product without considering customers' need (which leads to customer dissatisfaction and has an extra cost for the company). It is also possible to predict changes in customer demand or use the time series to update CRs.

1. Feature selection methods can efficiently reduce the number of functional requirements and decrease the complexity of a new product/service design by reducing the number of FRs at the customer requirement definition phase. In this regard, unsupervised machine learning techniques in selection of functional requirements during customer requirement definition phase of QFD technique can be an attractive future research topic.

2. An optimal new product design needs high-dimensional information analysis at the early stage of the QFD technique. In this regard, neural network (NN) is a powerful tool for data analysis for product development purposes. Combining NNs with QFD technique for assessing the design alternatives during design phase is suggested as an important direction for future research.

Regarding the limitation, as the current study was a comparative analysis and the frequency of the hybrid QFD-MCDM studies, future study needs to validate empirical research to compare the results obtained from various hybrid models in detail to introduce more efficient models in terms of accuracy and precision of the results. Furthermore, different alternatives in various developed QFD models can be discussed. However, as a limitation in this review has not been addressed, the authors agreed in the future to contribute more by presenting the novel QFD evolution results rather than traditional methods.