Improving the elicitation of critical customer requirements through an understanding of their sensitivity

Abstarct

Customer requirements (CRs) are the essential driven forces of product development. Constrained by the rigid budget and time allocated to product development, much attentions and resources should be paid on critical customer requirements (CCRs). Product design occurs with an increasingly frenetic pace of change in today’s competitive market, and the changes of external environment will lead to the changes of CRs. Thus, involving the sensitivity of CRs toward influence factors to identify CCRs is of great significance to grasp the directions of product evolution and enhance market competitiveness. To fill this gap, this study proposes a CCRs identification method integrated Kano model and structural equation model (SEM). First, the Kano model is adopted to determine the category of each CR. Second, based on CRs’ categorization, an SEM model is established to measure the sensitivity of CRs toward the turbulence of influence factors. Then the importance of each CR is calculated, and by integrating the sensitivity and importance, a four-quadrant diagram is constructed to identify the CCRs. Finally, the identification of CCRs for smartphone is implemented as an example to demonstrate the feasibility and additional value of the proposed method.

Introduction

In the last decade, product design has already moved from innovation/technology oriented to customer oriented (Hsiao et al. 2010; Borgianni et al. 2015). Understanding and identifying customer requirements (CRs) serve as vital steps in customer-driven product and service development (Reich and Levy 2007; Lai et al. 2008). However, constrained by the rigid budget and resource allocated in product development projects, companies are pressured to focus on satisfying critical customer requirements (CCRs) (Li et al. 2021). To this end, considerable research has been conducted regarding the weighting (prioritization/ranking) of CRs. By capturing and integrating CCRs or customer preferences accurately into product development, enterprises can develop a high-quality marketable product to meet or even exceed customer expectations with low cost (Li et al. 2018; Liu et al. 2020).

Product development occurs with an increasingly frenetic pace of change in today’s rapidly evolving market (Borgianni and Rotini 2015; Long and Ferguson 2021). Most products are designed and used under an uncertain environment (such as market price, natural resources, regulations, technologies), in which the most important factor is the uncertainty of CRs (Chong and Chen 2010; Cardin et al. 2017). The change of natural and social environment will lead to the change of CRs, and the frequently changing CRs will reduce product stability, shorten product life cycle, and increase the cost of design change or even redesign (Allen et al. 2016, 2017; Min et al. 2018). For this reason, many researchers paid attention to the change and evolution of CRs (Sun et al. 2020; Zhang et al. 2021). Nevertheless, the change trend of each CR may be different under the same influence environment (Li et al. 2021), i.e., some CRs may be more sensitive to the changes of internal and external influence factors than other CRs. In the product research and development process, designer and practitioners can embed design margins to the sensitive CRs (which must be obtained by sensitivity analysis), through which, the design change or redesign of the products can be effectively limited and controlled. Thus, much design effort would be saved for designers and practitioners, and the life cycle of the products are prolonged. Therefore, involving the sensitivity of CRs toward influence factors to identify CCRs is of great significance to grasp the direction of product evolution and enhance market competitiveness.

According to different impacts on customer satisfaction and different sensitivities to influence factors, CCRs with high importance and sensitivity can be identified. However, to the best of our knowledge, research on the identification of CCRs involving the sensitivity of CRs is still limited. In prior research, there are lots of common and useful methods in CR analysis for product design, such as Kansei engineering (Levy 2013; Dong et al. 2021), quality function deployment (QFD) (Reich and Paz 2008; Franceschini et al. 2015; Kirgizov and Kwak 2021; Karasan et al. 2022), and conjoint analysis (Le et al. 2019). These methods mainly focus aligning product technical attributes in relation to CRs and provide guideline for designers and industrial practitioners. However, they involved an inherent deficiency that only emphasize static CRs and ignore the rapidly changing environment where CRs are generated in. The present work is an attempt to bridge this gap by analyzing the sensitivity of CRs toward influence factors.

Motivated by the foregoing discussions and analysis, a CCR identification method is proposed by involving the sensitivity of CRs in a changing environment. One challenge involved is that the influence factors and relationships are complex and difficult to calculate directly. To this end, structural equation model (SEM) is employed to quantify the sensitivity of CRs toward the influence factors. First, the Kano model is adopted to determine the category of each CR, based on which, a survey is designed and the SEM is established to measure the sensitivity of CRs toward the turbulence of influence factors. In the field of CCRs identification, Kano model only focuses on the static CRs while ignoring the dynamic changing environment. Fortunately, the advantages of SEM allow for mining the relationships between dynamic CRs and their driven forces. With the combination of Kano model and SEM, we explored different sensitive characteristics among CRs in different Kano categories to provide supplement information on dynamic CRs and help designers’ decision making. Second, the importance of CRs is calculated through analyzing customer satisfaction/dissatisfaction degree, which is measured by Kano model. By integrating the sensitivity and importance, a four-quadrant diagram is developed to identify the CCRs. According to different impacts on customer satisfaction and different sensitivities to influence factors, CCRs with high importance and sensitivity are identified. Finally, the CCRs identification of smartphone is carried out as an example to demonstrate the feasibility and additional value of the proposed method. This study mainly makes the following two contributions:

1. According to the impacts of influence factors, the sensitivity of CRs is excavated. Combing the Kano model and SEM, the sensitivity of CRs in different Kano categories is quantified by the factor loading through SEM. It explores the relationships between mutative influence factors and the change of related CRs, through which, the CRs that are more agile to the fluctuation of influence factors can be detected and provides directions for designing evolution-adapted products.

2. Since the sensitivity and importance can represent the relative fluctuation degree toward influence factors and the absolute significance degree of CRs, they are integrated to propose a novel CCRs identification framework. A four-quadrant diagram is established to identify the CCRs with high sensitivity and importance. This is performed to ensure the identification results are not discounted due to the flaw in CRs change consideration that existed in prior studies.

The rest of this study is organized as follows. Section 2 reviews related literature. Section 3 details the proposed methodology. Section 4 illustrates an empirical example to demonstrate the feasibility and additional value of the proposed method. Result analysis and discussion are presented in Sect. 5, and Sect. 6 contains conclusion and future works.

Literature review

Identification of critical customer requirements

Traditionally, the identification of CCRs is realized through prioritization or ranking of CRs. We define it as the indirect CCRs identification method in this study. A better understanding of the importance ranking of CRs can help designers identify the CCRs and effectively optimize the design scheme. In the field of CRs prioritization, the importance of CRs is generally determined through three ways: direct assignment, pairwise comparison, and preference ranking (Zheng et al. 2016). As a widely used method in marketing analysis, direct assignment straightly invites customers to evaluate the importance of a CR in terms of point scale, interval number or linguistic variables, etc. (Wang et al. 2015; Du and Liu 2021; Geng et al. 2021). Considering customers are easier to evaluate a comparative value rather than an absolute one, pairwise comparison analysis is investigated (Song et al. 2013; Neira-Rodado et al. 2020). The above two ways of CRs’ importance ranking result in a need of elaborate input efforts from customers. To mitigate this issue, preference ordering has been developed with the capability of dealing with incomplete information. Nahm et al. (2012) developed a customer preference rating method and customer satisfaction rating method to rank CRs, capturing the incomplete or uncertain perception on the importance of CRs according to customers’ preferences. Franceschini et al. (2015) proposed a generalized Yager’s algorithm to merge the preference ranking of multiple interviewed customers for different CRs into consensus fusion ranking. Besides, several methods applied online reviews as a new data source to measure the customer preferences (Chen et al. 2019). Li et al. (2020) determined a CCR index to identify the CCRs with high customer satisfaction by text mining and sentiment analysis of online reviews. Cai et al. (2021) studied the relationships between products and users for requirement analysis, and they ranked CRs based on sentiment analysis, which further revealed the requirement differences among users. With the knowledge of importance–performance analysis, Joung and Kim (2021) prioritized CRs extracted form online reviews based on a Shapley additive explanation method through deep neural network. Prior studies mainly assess the importance of CRs by analyzing customer satisfaction/dissatisfaction degree. However, they involved an inherent deficiency that ignore the rapidly changing environment. Enterprises that only emphasize static CRs will make products fail to meet customer expectations, thus affecting product performance.

In addition to customer satisfaction, scholars have also identified CCRs with other considerations in the past decades. Dealing with multiform customers’ assessments, Wang et al. (2018) introduced a hybrid competitive priority rating to identify CCRs for customer collaborative production innovation. Liu et al. (2018) developed a low-carbon CCRs identification approach with carbon footprint analysis to clarify the CCRs which have dominant impacts on carbon emissions. As decision makers may hesitate in multiple attribute decision-making problems, Du et al. (2022) combined the evaluation information and position information in online reviews to prioritize CRs of smart-connected products with frequency, position, and the value of rejoicing and regret. Considering the heterogeneity in CRs and customers, Fang et al. (2022) improved a bi-cluster algorithm with the detection of a subset of customers who respond to the same subset of CRs to identify heterogeneous CRs. In product design process, distinct analysis of CRs characteristic not only guides designers in determining the function and structure of a product, but also increases customer satisfaction by adjusting related technical attributes. Although these former studies demonstrated the identification of CCRs with ordinary concerns, analysis on the dynamical impact of influence factors to CRs is still inadequate.

According to the axiomatic design, developed by Suh (2001), the change of CRs will cause a cascade of changes in functional requirements, design parameters, and process variables. The corresponding modified strategies of a product contributed by changing CRs are usually laborious and time consuming, especially when applied to complex products and systems. It is worth exploring what CRs are more significant and more likely to change when identifying CCRs in product design process. Hence, this study proposed a CCR identification method involving the dynamical impact of influence factors to different CRs to genuinely portray the characteristic of dynamic CRs for designers as realistically as possible.

Change of customer requirements

The increasing need to be sustainable pressures enterprises for designing evolution-adapted products to dynamic CRs. To better understand the change of CRs, many scholars focus on the dynamic CRs identification/mining/capturing/extracting. Dou et al. (2018) used the fuzzy Kano model and benchmarking theory to measure satisfaction improvement of each product attribute from the aspects of customer perception and competitor performance, so as to obtain new CRs, and employed optimized grey model to identify dynamic CRs. Believing that the interactions between customers will lead to serious changes in CRs, Li et al. (2019) proposed a fuzzy Delphi method to capture and evaluate dynamic CRs from customers’ perspective in the environment of open design. Based on online reviews, Zhao et al. (2022) mined dynamic CRs over multi-generation products by combining customer satisfaction and its dynamic change trends for continuous product improvement. Distinctly, these methods of identifying dynamic CRs only focus on the change of outward manifestations (i.e., change of customer satisfaction, importance of CRs), the driven forces that promote changes of CRs are not well investigated.

The change of natural and social environment will lead to the change of CRs, and analysis on CRs’ change and evolution is critical for enterprises to achieve long-term competitiveness (Zhang et al. 2021). Chong and Chen (2010) found that products cannot continuously meet customer expectations without considering the changes of CRs, so a method based on artificial immunity and nervous system is proposed to actively analyze and predict CRs. Due to the initiative of customers, many CRs vary proactively instead of passively adapting changes in the environment. According to Kano dynamics, Min et al. (2018) suggested a review-based framework to analyze the dynamic changes in CRs. Using this framework, firms can check how the characteristics of certain functions have changed over time. Additionally, considering the ever-changes of CRs, the discrepancy of CRs’ importance over time is also an effective way that reflects the change of CRs. Sun et al. (2020) classified CRs into user attributes, manufacture attributes, and common attributes and calculated changes of CRs’ importance in these three categories based on opinion mining and natural language processing to propose improvement strategy for next-generation product design. Since CRs often change dynamically on multi-generation products, it is worth exploring the mechanism by which CRs are impacted under turbulent influence factors. Although these studies effectively considered and described the dynamics of CRs, the reasons for change of CRs and sensitivity of CRs toward the turbulence of influence factors have not well been explored in depth.

The driven force of CRs fluctuation across product development process could be mainly divided into two categories. For one thing, CRs are isolated and there are connected relationships between them. The initial change of one CR may propagate to others due to functional and non-functional dependencies (Hein et al. 2021). For another, the changes of natural and social environment (such as market price, natural resources, regulations, technologies) will lead to the changes of CRs. Kim et al. (2022) presented empirical research on dynamic changes in customer responses to product features by comparing sentiment changes in positive/negative online reviews before and during COVID-19. As the relationships among CRs are mostly studied in the field of CRs propagation management, the later driven force is considered in detail in this study to analyze the change sensitivity of CRs and identify CCRs for evolution-adapted products. Table 1 shows a comparison of existing CCR identification methods with the proposed method.

Table 1.

Comparison of typical types of CCR identification methods with the proposed method

Reference	Weighting approach	Importance/satisfaction of CRs	Dynamics of CRs	Sensitivity of CRs	Driven force of CRs dynamics
Franceschini et al. (2015); Zheng et al. 2016	QFD-based weighting	Importance
Song et al. (2013); Neira-Rodado et al. (2020)	Group AHP	Importance
Geng et al. (2021)	Kano-based weighing	Importance and satisfaction
Li et al. (2020); Cai et al. (2021)	Sentiment analysis	Satisfaction
Kilroy et al. (2022)	Sentiment analysis	Satisfaction	√
Yakubu and Kwong (2021)	Statistical measure of a z-score	Importance	√
This study	Kano-based weighing	Importance and satisfaction	√	√	√

Methodology

The proposed CCRs identification methodology mainly comprises three stages: (1) categorization of CRs using Kano model; (2) sensitivity assessment of CRs toward influence factors based on SEM; (3) construction of a four-quadrant diagram to determine the CCRs. The framework of proposed methodology is depicted in Fig. 1.

Fig. 1.

Framework of the proposed CCRs identification method

The procedure of CRs classification is described in Sect. 3.1; the sensitivity of CRs toward influence factors is analyzed in Sect. 3.2; and the four-quadrant model for the identification of CCRs is constructed in Sect. 3.2.1.

Customer requirement classification

We assume t customers participate in this survey and CR_i (i = 1, 2, …, m) represents the i^th CR. Customers have different attitudes toward different CRs, and different CRs contribute to customer satisfaction diversely (Kano et al. 1984). With the evolution of CRs, the content and intensity of different types of CRs will change greatly. With the purpose of exploring the relationship between the classification of CRs and their sensitivity, a structured Kano questionnaire (listed in Table 14 in Appendix) is employed to categorize CRs. According to the classification standards in Table 2, customer’s answers and related CRs are classified into two categories, i.e., attractive CRs (A), one-dimensional CRs (O), indifferent CRs (I), must-be CRs (M), questioned CRs (Q), and reverse CRs (R).

Table 14.

Kano questionnaire

Kano question	Answer
If the customer requirement is fulfilled, how do you feel?	I like it
	It must be so as
	I don’t mind it
	I can live with it
	I hate it
If the customer requirement is not fulfilled, how do you feel?	I like it
	It must be so as
	I don’t mind it
	I can live with it
	I hate it

Table 2.

Kano assessment classification

Functional question	Dysfunctional question
	Like	Take for granted	Neutral	Can endure	Hate
Like	Q	A	A	A	O
Take for granted	R	I	I	I	M
Neutral	R	I	I	I	M
Can endure	R	I	I	I	M
Hate	R	R	R	R	Q

To reflect the feedback of customer satisfaction clearly and accurately, this study quantified customer satisfaction contribution of each CR by employing the Better–Worse coefficient analysis method (Wang et al. 2019). As shown in Table 2, considering the practical meaning of Kano categories, questioned CRs and reverse CRs are not considered. According to Eqs. (1) and (2), the satisfaction index (SI_i) and dissatisfaction index (DI_i) of CR_i are obtained.

$$S{I}_i=(A_i+O_i)/(A_i+O_i+M_i+I_i)$$ 1

$$D{I}_i=(O_i+M_i)/(A_i+O_i+M_i+I_i)$$ 2

where A_i, O_i, M_i, and I_i are the calculated quantity that CR_i is categorized into A, O, M, I, and Q, respectively, among t customers’ evaluations. SI_i is a positive number with the meaning of the degree of customer satisfaction improvement when providing a function. The greater the SI_i, the more significant improvement of customer satisfaction. DI_i represents the change degree of customer dissatisfaction when a function is not provided. The thresholds to classify CRs into Kano categories are the mean value of SI_i and DI_i, denoted as $\stackrel{-}{\mathit{SI}}$ and $\stackrel{-}{\mathit{DI}}$, respectively (Chen et al. 2021; Kirgizov and Kwak 2021). The objective is to identify CRs that contribute a relatively large degree of customer satisfaction or dissatisfaction. By establishing the vertical and horizontal axis based on $\stackrel{-}{\mathit{SI}}$ and $\stackrel{-}{\mathit{DI}}$, a quadrant graph is built to classify the CRs, in which each CR in quadrant I to IV is determined into O, A, I, and M, respectively.

Sensitivity assessment of customer requirement

SEM is widely applied in the behavior and social sciences to explain the relationships between cause-and-effect variables (Kang and Ann 2021). In recent years, some studies have tried to use it to analyze product and service performance and customer satisfaction in different fields (Ge et al. 2012; Gao et al 2013; Yu et al. 2019; Miao et al. 2019; Sun and Lau 2019). For the variables that cannot be clearly and directly measured (latent variables), SEM can use some observed variables to verify the hypothetical relationships between latent variables indirectly. For example, Miao et al. (2017) employed confirmatory factor analysis (CFA) and standard path coefficient in SEM to assess key influence factors of customer preferences on mobile health adoption. In this study, the SEM approach is applied to establish and quantify the causal relationships between CRs and related influence factors, and the CFA and standard path coefficient are employed to verify the predefined causal model, and then measure the sensitivity of CRs.

In the construction of SEM, researchers need to determine the latent variables and corresponding observed variables first. Each measured model is constructed by connecting latent variables with relevant observed variables. Then the influence relationships between latent variables are hypothesized. Surveys are conducted to verify the hypotheses and obtain conclusions from the path and coefficient of the model.

Hypotheses and SEM construction

As the market competition intensified and institutional policies changed, enterprises are driven to timely update their products and services (Dell’Anna et al. 2019). To investigate the relationships between environmental factors and enterprise factors, H₁ hypothesis is proposed. By managing the changes of CRs, enterprises continuously and imperceptibly guide the purchase behavior of customers in a competitive market environment (Min et al. 2018; Le et al. 2019; Akbar et al. 2020). To investigate the relationships between enterprise factors and customer factors, H₂ hypothesis is proposed. The advancement of social situations will also lead to changes in customer preferences (Dell’Anna et al. 2019). To investigate the relationships between environment factors and customer factors, H₃ hypothesis is proposed. In addition, the purchase behavior is dominated by the customer himself/herself. The changes of CRs can be reflected through customer satisfaction/preference which can be indirectly monitored by customer purchase behavior (Dou er al. 2018; Sun et al. 2022; Zhao et al. 2022; Du et al. 2022). In other words, customer factors have impacts on CRs of a product. To describe the relationships between customer factors and CRs minutely, we divide CRs into four categories using the Kano model, i.e., attractive CRs, one-dimensional CRs, indifferent CRs, and must-be CRs. Based on this, H₄–H₇ are constructed. In summary, we hypothesize that there are structural relationships among the influence factors (environmental factors, enterprise factors, and customer factors) and different Kano categories of CRs. The relationships are illustrated in the path diagram, as shown in Fig. 2. The hypotheses are as follows:

Fig. 2.

Structured model of CRs sensitivity assessment

• H₁: Environmental factors have positive impacts on enterprise factors.

• H₂: Environmental factors have positive impacts on customer factors.

• H₃: Enterprise factors have positive impacts on customer factors.

• H₄: Customer factors have positive impacts on attractive CRs.

• H₅: Customer factors have positive impacts on one-dimensional CRs.

• H₆: Customer factors have positive impacts on indifferent CRs.

• H₇: Customer factors have positive impacts on must-be CRs.

Using the conventional SEM notations, we assume that enterprise factors, customer factors, attractive CRs, one-dimensional CRs, indifferent CRs, and must-be CRs are endogenous latent variables, denoted as η_k (k = 1, …, 6), respectively. The ‘environmental factors’ is an exogenous latent variable denoted as ξ. Cause-and-effect variables are connected by one-way arrows pointing to the affected variable.

Survey design

A detailed process diagram of SEM estimation and evaluation is shown in Fig. 3. Experts with domain knowledge on CRs of selected products and customers of these products are invited and interviewed to answer prepared questions (the interview questions about influence factors are supplemented in the Appendix). Based on the grounded theory and relevant research, the influence factors (sub-factors) can be extracted. A Standardized SEM questionnaire is designed to support the CFA that tests whether the hypothetical model fits the survey data, thereby exploring the relationships between CRs and influence factors based on the standard path coefficient. In the questionnaire, self-identification questions are used to collect the demographics (such as age and occupation) of survey respondents. The assessment questions are the main part of the SEM questionnaire that collects the perceptions of survey respondents to each observed variable. The five-level Likert scale is used in the main part of the questionnaire, where 1–5 points represent strongly disagree to strongly agree.

Fig. 3.

Process of SEM estimation and evaluation

By testing the data with Bartlett test (Tobias and Carlson 1969) and Cronbach’s α in SPSS software, we obtain the Kaiser–Meyer–Olkin (KMO) value and the Cronbach’s α to verify the internal consistency and adequate reliability and validity of samples. On the basis of reliability and validity test, CFA is used to evaluate the fitness between the survey data and the hypothetical model, i.e., the structured model in SEM that consists of hypotheses between latent variables. The vector form of the SEM is:

$$y={\mathrm{\Lambda }}_y\eta +\epsilon$$ 3

$$x={\mathrm{\Lambda }}_x\xi +\delta$$ 4

$$\eta =B_{\eta }+\mathrm{\Gamma }\xi +\zeta$$ 5

where Eqs. (3–5) are measured models of endogenous latent variables, exogenous latent variables, and structured model, respectively. In Eq. (3), y is a p × 1 vector composed of p endogenous indices, η is a k × 1 vector composed of k endogenous latent variables, Λ_y is the p × k factor loading matrix of y on η, and ε is a p × 1 vector composed of p measurement errors. In Eq. (4), x is a q × 1 vector composed of q exogenous indices, ξ is an n × 1 vector composed of n exogenous latent variables, Λ_x is the q × n factor loading matrix of x on ξ, and δ is a q × 1 vector composed of q measurement errors. In Eq. (5), B_η is the k × k coefficient matrix, which describes the mutual influences among the endogenous latent variables η; Γ is the k × n coefficient matrix, which describes the influence of the exogenous latent variable ξ on the endogenous latent variable η; and ζ is the k × 1 residual vector. The maximum likelihood estimation (MLE) is applied to obtain the values of Λ_y, Λ_x, B_η, and Γ for influencing paths analysis and CRs sensitivity analysis.

The change sensitivity of CRs is proportional to the intensity of impact which is reflected by factor loading. The more significant a CR is affected, the more likely it is to change and the more sensitive it is, and vice versa. Based on this, the factor loading of Λ_x in measured model of η₃–η₆ is regarded as CRs sensitivity and denoted as s_i (i = 1, 2, …, m). The impact of influence factors on CRs is obtained as Λ_y and Λ_x in measured model of ξ, η₁ and η₂, respectively.

Identification of critical customer requirement

In this study, Kano model is used to mine the weight information of CRs. By adopting the weighting method proposed by Liu et al. (2015), the importance degree of CR_i, i.e., w_i (i = 1, 2, ..., m), is obtained by Eq. (6).

$$w_i=max\left\{\frac{S{I}_i}{{\sum }_{i=1}^{m}S{I}_i},\frac{D{I}_i}{{\sum }_{i=1}^{m}D{I}_i}\right\}$$ 6

Accordingly, each CR can be represented as CR_i = (w_i, s_i). By calculating the mean value $\stackrel{-}{w}$ and $\stackrel{-}{s}$ and plotting the value pair (w_i, s_i) into a four-quadrant diagram, as shown in Fig. 4, the CCRs with high importance and sensitivity can be identified. If w_i < $\stackrel{-}{w}$, CR_i falls into quadrant III or IV; otherwise, it belongs to quadrant I or II. Similarly, if s_i < $\stackrel{-}{s}$, CR_i falls into quadrant II or III; otherwise, it belongs to quadrant I or IV. The CCR area includes the CRs of significant importance and high sensitivity, toward which, more design resource should be paid.

Fig. 4.

An illustration of the four-quadrant model for identification of CCRs

Case study

Huge market demand not only provides development opportunities for smartphones, but also puts forward potential challenges. The research and development of smartphones should consider the change of CRs as much as possible. Thus, in this section, the identification of CCRs of smartphones is implemented as an example to demonstrate the feasibility and additional value of the proposed method, and seeks to provide suggestions for the development of smartphones.

Customer requirements classification

Since the heterogeneity of customers who may come from different fields with diversified backgrounds and knowledge, 30 customers are invited and interviewed to answer prepared questions about smartphones (the interview questions are listed in Appendix). After distilling CRs based on grounded theory and supplemented with relevant research (Li et al. 2020; Charmaz and Thornberg 2021; Kim et al. 2022), 31 CRs for smartphones are extracted as shown in Table 3. The Kano questionnaire for the 31 CRs are designed and randomly distributed to customers. Finally, 655 questionnaires are collected, of which, 545 ones are effective, with an effective rate of 83.21%. By testing reliability and validity, the KMO value and the Cronbach’s α are obtained and listed in Table 4. The results indicate that the samples have a high level of internal consistency and adequate reliability and validity.

Table 3.

Extracted customer requirements for smartphones

	CRs		CRs		CRs
CR1	Water resistant	CR12	Wireless charging	CR23	User interface
CR2	Heat dissipation	CR13	Battery capacity	CR24	Voice control
CR3	Privacy protection	CR14	Screen style	CR25	Vibration
CR4	Storage	CR15	Screen resolution and refresh rate	CR26	Price
CR5	Chip	CR16	Care of eyes	CR27	Sound quality
CR6	Dual SIM	CR17	Weight	CR28	Volume
CR7	Shooting effect	CR18	Size	CR29	5G network
CR8	Pixel	CR19	Appearance	CR30	System Apps
CR9	Responding and running speed	CR20	Material	CR31	Universal control
CR10	Battery life	CR21	Iris recognition
CR11	Fast charging	CR22	Fingerprint recognition

Table 4.

Reliability and validity test of the Kano questionnaire

	Cronbach’s α	KMO
All questions	0.905	0.955
Functional questions	0.962	0.967
Dysfunctional questions	0.964	0.969

According to Eqs. (1) and (2), the Better–Worse value of each CR is calculated and listed in Table 5. The $\stackrel{-}{\mathit{SI}}$ (0.680548) and $\stackrel{-}{\mathit{DI}}$ (0.514416) of all CRs are crossed as the origin. The classification results are shown in Fig. 5.

Table 5.

Kano classification number and Better–Worse coefficient statistics of the 31 CRs

CRs	M	O	A	R	Q	I	SI	DI	Category
CR1	113	185	92	6	24	125	0.5379	− 0.5786	M
CR2	96	223	98	4	22	102	0.6185	− 0.6146	M
CR3	90	347	38	2	19	49	0.7347	− 0.834	O
CR4	59	304	98	2	17	65	0.7643	− 0.6901	O
CR5	63	260	133	2	15	72	0.7443	− 0.6117	O
CR6	77	162	122	2	13	169	0.5358	− 0.4509	I
CR7	53	255	122	2	15	98	0.714	− 0.5833	O
CR8	41	239	154	4	15	92	0.7471	− 0.5323	O
CR9	68	368	51	3	15	40	0.7951	− 0.8273	O
CR10	58	307	99	3	11	67	0.7646	− 0.6874	O
CR11	43	206	185	1	9	101	0.7308	− 0.4654	A
CR12	20	86	241	2	6	190	0.6089	− 0.1974	I
CR13	47	257	144	1	10	86	0.7509	− 0.5693	O
CR14	40	174	177	2	7	145	0.6549	− 0.3993	I
CR15	38	221	163	1	7	115	0.7151	− 0.4823	A
CR16	42	169	182	2	8	142	0.6561	− 0.3944	I
CR17	43	154	182	2	7	157	0.6269	− 0.3675	I
CR18	48	168	184	2	7	136	0.6567	− 0.403	I
CR19	42	199	174	2	6	122	0.6946	− 0.4488	A
CR20	48	199	166	3	8	121	0.6835	− 0.4625	A
CR21	63	180	159	2	8	133	0.6336	− 0.4542	I
CR22	80	243	125	1	9	87	0.6879	− 0.6037	O
CR23	52	207	167	1	8	110	0.6978	− 0.4832	A
CR24	53	175	151	1	9	156	0.6093	− 0.4262	I
CR25	51	138	164	3	8	181	0.5655	− 0.3539	I
CR26	67	324	89	1	8	56	0.7705	− 0.7295	O
CR27	65	276	104	1	9	90	0.7103	− 0.6374	O
CR28	57	243	121	1	9	114	0.6804	− 0.5607	M
CR29	34	141	197	1	8	164	0.6306	− 0.3265	I
CR30	30	167	206	2	9	131	0.6985	− 0.3689	A
CR31	35	180	182	2	9	137	0.6779	− 0.4026	I

Fig. 5.

Results of the Kano categorization for the 31 CRs

Construction and fitting of the structural equation model

Five experts with domain knowledge on smartphone CRs and 30 customers are invited and interviewed to answer prepared questions (the interview questions about influence factors are listed in Appendix). Based on grounded theory and relevant research (Fernandes et al. 2015; Charmaz and Thornberg 2021), the influence factors (sub-factors) for the 31 CRs are extracted, as shown in Table 6. To keep consistent with the CRs notation in Sect. 4.1, the endogenous indices y of attractive CRs (η₃), one-dimensional CRs (η₄), indifferent CRs (η₅), and must-be CRs (η₆) in Eq. (3) is replaced by the symbol CR_i in Table 3. The fixed parameters of x₁ − x₃, y₁ − y₁₄, and CR₁ − CR₃₁ are denoted as ε₁ − ε₃, δ₁ − δ₁₄, and δ₁₅ − δ₄₅, respectively.

Table 6.

Influence factors of the 31 CRs for smartphone

Latent variables	Observed variables
Environmental factors (ξ)	Market environment (x1), social economy (x2), natural condition (x3)
Enterprise factors (η1)	Product upgrade (y1), technological innovation (y2), system update (y3), brand image (y4), marketing strategy (y5), service (y6), quality (y7)
Customer factors (η2)	Purpose (y8), personal preference (y9), psychological factors (y10), use experience (y11), consumption ability (y12), interaction with others (y13), knowledge reserve about smartphone (y14)

In the SEM survey, customers who have bought and used smartphones are randomly selected. Through offline and online surveys (an overview of the questionnaire is listed in Tables 15 and 16 in Appendix), 605 questionnaires are collected, of which, 461 ones are effective, with an effective rate of 76.19%. The respondents include 206 males and 255 females. Among the respondents, 90.02% age among 18–40, and 88.93% have used more than two smartphones. People in this age group have a relative mature cognition of smartphones and understand their own requirements. The individual information of surveyed customers is shown in Table 7. The reliability and validity of the questionnaire are tested as shown in Table 8. The results indicate that the scale has high reliability, and the correlations among variables are significant.

Table 15.

Self-identification questions in SEM

Gender	Age /Year(s)	Used smartphone amount /Piece(s)
Male	< 18	1
Female	18 − 25	2 − 4
	25 − 40	> 4
	> 40

Table 16.

Assessment questions in SEM

Latent variable	Measurement	Detailed questions
Environmental factors	x 1	My requirements for smartphone will change due to competition in phone industry and political factors such as the trade war
	x 2	Different social economy conditions (such as new fashion trends, environmental protection, etc.) will influence my requirements for smartphone
	x 3	Different natural conditions will make my requirements for smartphone change. (For example, my smartphone is not waterproof now. When the smartphone is wet in rainy and snowy days, I will think that the waterproof of the smartphone needs to be improved.)
Enterprise factors	y 1	Smartphones are upgrading rapidly. New and advanced smartphones are coming out all the time, which will change my requirements
	y 2	With technological innovations (such as Always on Display), I feel that smartphone needs to be improved on that function
	y 3	As the ecosystem of smartphone changes (such as system update, maintenance), some of my requirements for smartphone will change
	y 4	As enterprise’s brand image changes (such as reputation damage), some of my requirements for smartphone will change
	y 5	As enterprise’s marketing strategies change (such as gifts, celebrity endorsements), some of my requirements for smartphone will change
	y 6	The service of enterprises is uneven. When there are some forms of high -quality service, I think my smartphone service should be improved accordingly
	y 7	Defective products will occur in the production process. Good quality control of smartphone will make me feel more secure. As the quality control level of smartphones in this enterprise changes, some of my requirements will change
Customer factors	y 8	When the function of smartphone cannot satisfy my requirement (For example, I am a game enthusiast, and the smartphone does not run smoothly when playing games), I think the function of the smartphone needs to be improved
	y 9	When my preferences change (such as esthetics, hobbies, and the pursuit of cost performance), some of my requirements for smartphone will change
	y 10	When my psychological factors change (such as consumption view, brand loyalty), some of my requirements for smartphone will change
	y 11	When the use experience of existing smartphone changes (such as not running smoothly, running out of memory), some of my requirements for smartphone will change
	y 12	As my consumption ability improves, some functions of my smartphone need to be improved
	y 13	Due to the interactions with others (such as friends’ recommendation, celebrities’ recommendation), some of my requirements for smartphone will change
	y 14	As my knowledge and understanding about smartphone become more insightful, some of my requirements for smartphone will change
Attractive CRs	CR 11	My requirement for fast charging of smartphones varies easily
	CR 15	My requirement for screen resolution and refresh rate of smartphones varies easily
	CR 19	My requirement for appearance of smartphones varies easily
	CR 20	My requirement for material of smartphones varies easily
	CR 23	My requirement for user interface of smartphones varies easily
	CR 30	My requirement for system Apps of smartphones varies easily
One-dimensional CRs	CR 3	My requirement for privacy protection of smartphones varies easily
	CR 4	My requirement for storage of smartphones varies easily
	CR 5	My requirement for chip of smartphones varies easily
	CR 7	My requirement for shooting effect of smartphones varies easily
	CR 8	My requirement for pixel of smartphones varies easily
	CR 9	My requirement for responding and running speed of smartphones varies easily
	CR 10	My requirement for battery life of smartphones varies easily
	CR 13	My requirement for battery capacity of smartphones varies easily
	CR 22	My requirement for fingerprint recognition of smartphones varies easily
	CR 26	My requirement for price of smartphones varies easily
	CR 27	My requirement for sound quality of smartphones varies easily
Indifferent CRs	CR 6	My requirement for dual SIM of smartphones varies easily
	CR 12	My requirement for wireless charging of smartphones varies easily
	CR 14	My requirement for screen style of smartphones varies easily
	CR 16	My requirement for care of eyes of smartphones varies easily
	CR 17	My requirement for weight of smartphones varies easily
	CR 18	My requirement for size of smartphones varies easily
	CR 21	My requirement for iris recognition of smartphones varies easily
	CR 24	My requirement for voice control of smartphones varies easily
	CR 29	My requirement for 5G network of smartphones varies easily
Must-be CRs	CR 1	My requirement for water resistant of smartphones varies easily
	CR 2	My requirement for heat dissipation of smartphones varies easily
	CR 28	My requirement for volume of smartphones varies easily

The answers to all questions are five-level Likert scale ranging “totally agree, agree, general, disagree and totally disagree”

Table 7.

Demographics of the surveyed customers

Gender		Age /Year(s)		Used smartphone amount /Piece(s)
Male	44.69%	< 18	2.82%	1	11.07%
Female	55.31%	18 − 25	74.84%	2 − 4	63.77%
		25 − 40	15.18%	> 4	25.16%
		> 40	7.16%

Table 8.

Reliability and validity test of the SEM questionnaire results

Latent variables	Cronbach’s alpha	Items	KMO
Totality	0.940	48	0.915
ξ	0.738	3	0.667
η1	0.818	7	0.848
η2	0.771	7	0.786
η3	0.799	6	0.840
η4	0.886	11	0.895
η5	0.890	11	0.897
η6	0.693	3	0.587

On the basis of reliability and validity test, CFA is used to evaluate the fitness performance between the survey data and predefined causal model. A statistical software is employed to fit the SEM. The χ²/df, incremental fit index (IFI), comparative fit index (CFI), parsimony-adjusted normed fit index (PNFI), and root mean square error of approximation (RMSEA) are chosen as indices to evaluate the fitness performance (Sun and Lau 2019). Unfortunately, χ²/df and IFI do not pass the fitting test; therefore, the initial model needs to be modified. Free estimating a restricted parameter enables the model to be improved, and the reduced value of χ² in the model is called the modification indices (M.I.) of this parameter. When fitting indices do not reach the ideal standard, the predefined causal model can be modified appropriately according to M.I., significance probability (P), and practical meaning. To this end, eight paths in hypothetical model are picked and modified as shown in Table 9. The fixed parameters δ₅, δ₆, δ₉, δ₁₀, δ₁₃, δ₁₈, δ₂₁, δ₂₂, δ₃₁, δ₃₂, δ₄₀, δ₄₁, δ₄₂, δ₄₄, and δ₄₅ are released as free parameters. The chosen indices of the modified model shown in Table 10 all reach the fitting standard.

Table 9.

Modification paths

			M.I	P				M.I	P
δ21	↔	δ22	106.492	0.283	δ18	↔	δ40	42.084	0.150
δ31	↔	δ32	105.335	0.230	δ5	↔	δ13	40.857	0.308
δ9	↔	δ10	59.063	0.182	δ5	↔	δ6	39.028	0.267
δ44	↔	δ45	56.438	0.252	δ41	↔	δ42	30.492	0.178

Table 10.

Test result of modified structural equation fitness

Index	χ2/df	IFI	CFI	PNFI	RMSEA
Default model	2.799	0.808	0.807	0.689	0.063
Standard	< 3	> 0.8	> 0.8	> 0.5	< 0.08

Through MLE, the fitting result of revised hypothetical model is obtained and the standardized path coefficient (factor loading) is shown in Table 11. As all P values of the path coefficient are lower than 0.1, we believe that all the hypotheses should be accepted. The coefficient between variables in SEM shows the degree of influence, which has mostly been assumed to be linear, i.e., the higher the factor loading, the higher the degree of influence and vice versa. Specially, the sensitivity of the CRs of smartphones is quantified based on the values of factor loading.

Table 11.

The standardized path coefficient in SEM

	Estimate	S.E	C.R	P		Estimate	S.E	C.R	P
η1 ← ξ	0.610	0.062	9.489	***	CR23 ← η3	0.288	0.133	10.871	***
η2 ← ξ	0.114	0.041	1.658	*	CR30 ← η3	0.667	0.096	5.691	***
η2 ← η1	0.582	0.050	7.203	***	CR3 ←η4	0.697
η3 ← η2	0.696	0.094	7.960	***	CR4 ← η4	0.643	0.080	13.142	***
η4 ← η2	0.683	0.098	8.893	***	CR5 ← η4	0.576	0.085	12.267	***
η5 ← η2	0.633	0.103	7.965	***	CR7 ← η4	0.627	0.087	11.105	***
η6 ← η2	0.468	0.126	7.037	***	CR8 ← η4	0.733	0.086	11.997	***
x1 ← ξ	0.714				CR9 ← η4	0.732	0.076	13.739	***
x2 ← ξ	0.796	0.087	12.206	***	CR10 ← η4	0.741	0.077	13.736	***
x3 ← ξ	0.698	0.074	10.779	***	CR13 ← η4	0.568	0.086	13.874	***
y1← η1	0.794				CR22 ← η4	0.682	0.087	7.980	***
y2 ← η1	0.610	0.055	16.456	***	CR26 ← η4	0.629	0.085	11.718	***
y3 ← η1	0.776	0.055	17.237	***	CR27 ← η4	0.660	0.080	11.631	***
y4 ← η1	0.764	0.058	13.451	***	CR6 ← η5	0.719
y5 ← η1	0.798	0.073	6.522	***	CR12 ← η5	0.650	0.110	11.111	***
y6 ← η1	0.636	0.065	11.266	***	CR14 ← η5	0.703	0.101	10.538	***
y7 ← η1	0.317	0.059	12.268	***	CR16 ← η5	0.705	0.102	10.885	***
y8 ← η2	0.541				CR17 ← η5	0.670	0.103	11.489	***
y9 ← η2	0.585	0.108	8.918	***	CR18 ← η5	0.638	0.098	10.744	***
y10 ← η2	0.528	0.118	8.801	***	CR21 ← η5	0.408	0.104	11.337	***
y11 ← η2	0.431	0.096	7.562	***	CR24 ← η5	0.815	0.106	11.357	***
y12 ← η2	0.578	0.122	9.394	***	CR29 ← η5	0.790	0.099	10.985	***
y13 ← η2	0.424	0.129	7.551	***	CR1 ← η6	0.682	0.102	10.633	***
y14 ← η2	0.515	0.117	8.666	***	CR2 ← η6	0.629	0.075	7.910	***
CR11 ← η3	0.536				CR28 ← η6	0.660
CR15 ← η3	0.621	0.130	9.770	***			0.088	11.244	***
CR19 ← η3	0.800	0.144	11.206	***			0.057	7.899	***
CR20 ← η3	0.750	0.145	11.122	***

Standard error (S.E.); critical ratio (C.R.)

Identification of the CCRs

The importance of CR_i in Table 12 is obtained based on the data of Kano questionnaire by Eqs. (1−2). The value pair (w_i, s_i) of CR_i is plotted into a four-quadrant diagram, as shown in Fig. 6. With the mean value $\stackrel{-}{w}$ (0.0323) and $\stackrel{-}{s}$ (0.6449) of all CRs, CR₃, CR₈, CR₉, CR₁₀, CR₁₉, CR₂₀, CR₂₂, CR₂₇, and CR₃₀ are identified as the CCRs.

Table 12.

Critical value of CRs

	Wi	Si		Wi	Si		Wi	Si
CR1	0.025	0.682	CR12	0.029	0.650	CR23	0.033	0.288
CR2	0.029	0.629	CR13	0.036	0.568	CR24	0.029	0.815
CR3	0.035	0.697	CR14	0.031	0.703	CR25	0.027	0.790
CR4	0.036	0.643	CR15	0.034	0.621	CR26	0.037	0.629
CR5	0.035	0.576	CR16	0.031	0.705	CR27	0.034	0.660
CR6	0.025	0.719	CR17	0.030	0.670	CR28	0.032	0.660
CR7	0.034	0.627	CR18	0.031	0.638	CR29	0.030	0.405
CR8	0.035	0.733	CR19	0.033	0.800	CR30	0.033	0.667
CR9	0.038	0.732	CR20	0.032	0.750	CR31	0.032	0.568
CR10	0.036	0.741	CR21	0.030	0.408
CR11	0.035	0.536	CR22	0.033	0.682

Fig. 6.

Results of CCRs identification

Result analysis and discussion

Result analysis

To simplify the visualization of SEM and help to generate managerial insights, three measured models of latent variable ξ, η₁, and η₂ in the SEM are extracted and visualized in Fig. 7. According to Table 11 and Fig. 7, several findings of influence factors are highlighted as follows. First, social economy (x₂), marketing strategy (y₅), and product upgrade (y₁) are the most dominant influence factors of smartphone CRs. In other words, the changes of CRs are mainly caused by social economy development, enterprises’ product promotion and product upgrade. Second, in addition to the above three factors, market environment (x₁) contributes more in the influence of environmental factors, while system update (y₃) and brand image (y₄) are more prominent in enterprise factors. This is because political factors and industry competition closely exist in customers’ daily life, which affect customers’ behavior, thought and purchase intention. Third, use experience (y₁₁), interactions with others (y₁₃), and quality (y₇) are not key drivers of CRs change. Thus, enterprises should not regard these aspects as the core influence factors of CRs’ change, but should pay more attention to product innovation and marketing promotion to guide the evolution of CRs.

Fig. 7.

Path coefficient in measured model of ξ, η₁, and η₂

Table 13 documents the affect value of the paths between latent variables in the structured model with accompanying Fig. 8 providing a visualization. Several discoveries can be found. First, the CRs of smartphones will be affected by environment, enterprise, and customer factors and the customers themselves have the greatest influence on CRs in any category. Second, in the direct influence path, customers have more influence on attractive CRs, where the factor loading is 0.696, followed by the one-dimensional CRs (0.683) and indifferent CRs (0.633). Third, in the indirect influence path (such as ξ → η₃ in Fig. 8), the affect value of enterprise factor to attractive CRs is the largest among others. Fourth, there are no significant influences on must-be CRs from all influence factors.

Table 13.

Total affect value

	ξ	η1	η2
η1	0.610	0.000	0.000
η2	0.114	0.582	0.000
η3	0.326	0.405	0.696
η4	0.320	0.398	0.683
η5	0.297	0.368	0.633
η6	0.220	0.272	0.468

Fig. 8.

Affect value of the path in structured model

The sensitivity assessment results in Table 12 are extracted, combined with which three conclusions from Fig. 9 are attained. First, voice control (CR₂₄), appearance (CR₁₉), vibration (CR₂₅), and material (CR₂₀) are the most sensitive CRs among others. According to Avikal et al. (2018), esthetic design are the most censorious factors for the product accomplishment in industry, and the visual aspect plays an important role in customer satisfaction enhancement. This is in line with the results of this study. Second, the sensitivity of CRs differs from Kano categories, among which the sensitivity of attractive CRs and indifferent CRs have larger variation. Designers should be concentrated on these two types of CRs as they are more susceptible, easy to change, and even evolve into other types of CRs. Third, user interface (CR₂₃) is the least sensitive CR. This is because the main functions of smartphones are relative mature and stable. There are no significant differences in user interaction functions among different brands, and many smartphones even support customers to customize the interface. Moreover, the sensitivity of iris recognition (CR₂₁), 5G network (CR₂₉), and fast charging (CR₁₁) is also at a lower level among others.

Fig. 9.

Ranking of CRs sensitivity in different Kano categories

After acquiring the sensitivity of CRs, the CCRs have been identified with the combination of the importance to help enterprises improve the product. Two results can be obtained as follows. For one thing, privacy protection (CR₃), pixel (CR₈), responding and running speed (CR₉), battery life (CR₁₀), appearance (CR₁₉), material (CR₂₀), fingerprint recognition (CR₂₂), sound quality (CR₂₇), and system Apps (CR₃₀) are identified as CCRs with higher importance and sensitivity. Moreover, it should be noted that these CCRs are all attractive CRs or one-dimensional CRs, which is in line with the research results of Li et al. (2020). While different from the above research, this study does not identify all attractive and one-dimensional CRs as CCRs. For example, user interface (CR₂₃) is classified into attractive CRs, while it is not identified as CCRs in this study for its least sensitivity among CRs. In one-dimensional CRs, battery capacity (CR₁₃) is also failed to be identified as CCRs in this study because of its relatively low change sensitivity. What’s more, driven from the perspective of user experience, customers are more curious about battery life rather than battery capacity today. High battery capacity does not guarantee long battery life. As a huge competitive brand, iPhone has rarely been strong on battery capacity, but it serves good work in terms of battery life for its total control of the IOS ecosystem and targeted optimization of processors.

Comparison and discussion

To verify the effectiveness and rationality of the proposed CCRs identification method, as the comparison, the traditional method that does not consider CRs sensitivity is adopted to identify the CCRs. By the traditional identification method, the final CCRs are obtained and the comparison results are shown in Fig. 10. The storage (CR₄), chip (CR₅), battery capacity (CR₁₃), price (CR₂₆), and fast charging (CR₁₁) are not identified as CCRs in this study, although they have high importance. This is because these requirements, even though contributing significantly to customer satisfaction, are not susceptible and, thus, have little impact on product design change. In contrast, sound quality (CR₂₇), system Apps (CR₃₀), fingerprint recognition (CR₂₂), material (CR₂₀), and appearance (CR₁₉) are identified as CCRs in this study which are highly sensitive to environment change. Compared to the traditional methods for CCR identification, this study injects the sensitivity of CRs toward influence factors, in which the added steps include influence factors elicitation, SEM construction, and CRs’ sensitivity assessment. Among these procedures, the surveyed time is mainly related to the question number of questionnaires (in this study, it is closely related to the specific number of CRs and influence factors), the number and professional level of respondents. In this study, we invited 30 interviewers on influence factors elicitation, and collected a total of 605 SEM questionnaires, which only take several days more than the traditional methods. The assessment of the sensitivity of CRs is obtained through the fitting process between hypothetical model and survey data, and can be achieved easily by the SEM software. The consideration of change sensitivity of CRs constitutes an incredibly attractive opportunity for reducing design change costs and prolonging product life cycle, while not compromising product stability. Above all, according to the results of this study, the critically of CRs in different Kano categories is quite various, enterprise should pay full attention to these CCRs in product development.

Fig. 10.

Comparison results between traditional method and proposed method

There are three pre -assumptions in this study to identify CCRs, which are (1) CRs are not stable and will change as time progresses; (2) CRs are inevitably affected and changed by the changing influence factors; (3) the sensitivity of CRs toward the turbulence of influence factors is different. From prior research, the former two assumptions have been well verified (Lam and Shankararaman 1999; Fernandes et al. 2015; Kim et al 2022). The third assumption is verified in this study. From the case study, we find that CRs have different sensitivities toward the changing influence factors. What’s more, the sensitivity of CRs differs from Kano categories, among which the attractive CRs and indifferent CRs are more sensitive. Although differences in sensitivity among some CRs in Fig. 9 are not significant, design strategies for their related product technical attributes may be extremely diverse. In the case study, for example, voice control (CR₂₄) and appearance (CR₁₉) both have high sensitivity, while the modification of appearance to meet changed CRs is freer and more varied compared to the voice control that focuses more on technology and system. The conclusion in Li et al. (2022) showed that the core components in product structure are always important even if the product evolves rapidly. Accordingly, although a requirement may be of a high volatility, its actual significance may be minimal. Real CCRs are those of a high significance and high sensitivity, and should be highlighted for special attention. The proposed method is an attempt to practice SEM in exploring the relationships between mutative influence factors and the change of related CRs. Therefore, the CRs that are more significant and more likely to change is identified as CCRs in product design process. In product design process, enterprises and designers can utilize this method with mapping methods (such as QFD) between CRs and technical attributes to help agile manufacturing or carry out modular design, etc., thereby prolonging the product life cycle and reducing the cost of product design change or redesign.

There are also some limitations in the methodology. For example, the relationships between CRs and influence factors in SEM are assumed linear, are there nonlinear relationships between them? Is there a systematic and generalized approach to extract the influence factors of CRs? In addition, this study is grounded on the assumption that all CRs do not interact with each other, while the truth is that CRs are not isolated and there might be interactions among them.

Conclusion and future work

As a vital step in customer-driven product development, satisfying CCRs has been increasingly concerned by enterprises to meet the rigid budget and resource allocation in product development projects. The change of influence factors including external environment and personal experiences of customers will lead to the changes of CRs. Involving the sensitivity of CRs toward influence factors to identify CCRs enables enterprises effectively grasp the direction of product evolution. However, research on the identification of CCRs considering the sensitivity of CRs is still limited. To bridge this gap, this study proposes a CCRs identification method that integrates Kano model and SEM, involving the sensitivity of CRs and exploring the change impact of influence factors on CRs from the perspective of environment, enterprises, and customers. A case study of smartphones is provided to illustrate the feasibility and additional value of the proposed CCRs identification method. This study is concluded as follows.

1. To explore the sensitivity of CRs toward influence factors, the sensitive characteristics of CRs are excavated according to the impact of external environment. The results from Kano model and SEM show that voice control and appearance, with the sensitivity of 0.815 and 0.8, respectively, are more sensitive to the change of influence factors. Among the four Kano categories, attractive CRs are more susceptible, easy to change, and even evolve into other types of CRs.

2. Involving the sensitivity of CRs, a novel CCRs identification framework is proposed. Based on the sensitivity and importance of each CR, the pairwise criticality value is constructed and nine CCRs are identified whose sensitivity and importance are both higher than the mean values $\stackrel{-}{w}$ (0.0323) and $\stackrel{-}{s}$ (0.6449).

3. The elicited nine CCRs are all attractive CRs and one -dimensional CRs. In addition, among all Kano categories, the attractive CRs and indifferent CRs have larger variation. Compared to the traditional CCRs identification method, some CRs with high importance are not identified as in this study, and only CRs that contribute significantly to customer satisfaction and are highly sensitive to the environment changes are identified as CCRs.

The proposed model can help enterprises grasp the causes and potential change quantities of CR changes and then reduce the cost of product change and prolong the product life cycle. It is beneficial to the product deformation design in the future, and has certain practical significance to the product manufacturing of the enterprise. However, this study has some limitations that can be further explored in the future research. The first limitation is the extraction of influence factors. Although the SEM performs well within the case study, CRs of each specific product focus on different influence factors. It is suggested to shape a structured and standardized influence factor elicitation system for CRs. This will both help to segment product markets and allow for the applicability against heterogeneous products. The second limitation is the lack of market segmentation. As a core marketing concept, market segmentation is vital for successful customer relationship management that aims at understanding and measuring the true value of customers. The consideration of market segmentation is not emphasized in this study, but the integration of CR sensitivity and market segmentation is an interesting topic for future research.

Appendix

Survey questions in the case study

Customer requirements-related interview

1. What smartphone are you using now?

2. How long have you used this smartphone?

3. How long have you used your last smartphone?

4. Why did you replace your last smartphone?

5. Why did you choose the smartphone you are using now?

6. What makes you satisfied/dissatisfied about the smartphone you are using now?

7. What do you usually use your smartphone for?

8. What function do you want your smartphone to have? Why?

9. What are the main aspects that customers pay attention to when choosing a smartphone?

10. What is the competition point between different brands of smartphones?

11. What are the upgrades, improvements or innovations of the latest smartphone generation compared with previous ones?

12. What breakthroughs do you hope in the future smartphone development?

*Questions (9–12) are especially for experts.

Influence factors-related interview

1. What will affect your choice of smartphone?

2. What motivates you changed your last smartphone?

3. Do you feel like your needs have changed since your last phone? (For example, memory needs to be larger, pixels need to be higher)

4. What will you pay attention to if changing your smartphone again? Why?

5. What do you think is the reason for the constant upgrading of smartphones?

Kano questionnaire

See Table 14.

SEM questionnaire

See Tables 15, 16

Author contributions

Yupeng Li and Kaixin Sha contributed to the study conception and design. Material preparation, data collection and analysis were performed by Kaixin Sha, Haoran Li, Yu Wang, Jianhua Feng, Shuang Zhang and Yijiang Chen. The first draft and revised version of the manuscript was written by Kaixin Sha. Yupeng Li and Kaixin Sha commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Data availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Conflict of interest

The authors have no relevant financial or non-financial interests to disclose.