Sensory evaluation of customer satisfaction using type-2 fuzzy logic

Abstract

Sensory experiences that include vision, hearing, touching, smelling and tasting are important parameters that enable people to trade effectively in retail stores. In this study, based on multisensory attributes the evaluation of customer satisfaction is considered using fuzzy set theory and conjoint analysis. Fuzzy set theory is one of the best methodologies for describing the meaning of linguistic values that express customer preferences. However, there may be different customer and expert opinions in the evaluation of preferences by expressing linguistic values. In the paper, a type-2 fuzzy set is used to handle these uncertainties. This paper proposes the combination of type-2 fuzzy sets and conjoint analysis in order to evaluate customer satisfaction using customer opinions about sensory variables such as sight, sound, taste, touch and smell when purchasing goods in retail stores. For this purpose, using statistical survey results and type-2 fuzzy sets the customer satisfaction degrees were determined. The methodology used for the determination of customer satisfaction is based on conjoint analysis that uses the similarity measure to determine the closest opinions of the customers and experts for the evaluation of customer satisfaction degrees. The obtained experimental results indicate the efficiency of the presented approach in the determination of customer satisfaction in retail markets.

Keywords

Sensory attributes customer satisfaction fuzzy sensory evaluation conjoint analysis

1 Introduction

In retail marketing, retailers develop and build transaction methods regarding price, convenience, timesaving, and trading effectiveness to run their trade efficiently [1]. To be more profitable the retail markets need to have more customers. When a customer enters a retail store, the entire purchasing process can be based on the result of a sensory experience such as seeing, hearing, touching, smelling, tasting and also service quality. Based on sensory evaluation and subjective feeling, customers make a purchasing decision. Customers will try to keep future relationships with the marketplace for repurchasing the product if they were satisfied with the products. This increases the profit of the marketplace. People use their five senses of smell, hearing, seeing, tasting, and touching in their daily experiences to understand and feel the environment and the outer world. The five senses help to create multi-sensory attributes for consumers and they are used for personal brand perceptions of goods, products and services. The reference [2] studied the role of sensory marketing and also its influence on customer satisfaction.

Customer satisfaction is the evaluation of the customer experience about the products or services. When the perceived performance of products or services meets the expectation of customers then customer satisfaction is achieved. Customer satisfaction is characterised by the degree of feeling or degree of happiness that can range from satisfaction level to dissatisfaction. Satisfaction is a critical success factor in any business as the satisfied customer try to repurchase the same product in the future, at the same time recommend it to others. Therefore, the satisfaction and dissatisfaction of customers affect the performance of the marketplace. When customers are satisfied with the quality of the products, they become the best and most practical source for the advertisement of these marketplaces where they buy the products. Customers who are satisfied with the products purchased are more likely to purchase the same products again. These customers will most probably talk positively about their purchases. This will lead to the positive advertisement of the products. Otherwise, the dissatisfied customers will most probably switch to other products. Which lead to the negative advertisement of the products.

Customer satisfaction is the most important factor for the development of today’s competitive business world. It gives information that is more valuable for the company. For example, dissatisfied customers provide information related to the reason for which the product and the service have not fulfilled their expectations by answering surveys. Customer satisfaction helps the company to address issues related to customer requirements, to understand and address the customer expectations. For this purpose, the evaluation of customer satisfaction is very important in the functioning of any company, for increasing profitability and therefore increasing the financial situation of the company in future. In the paper, based on sensory attributes the evaluation of customer satisfaction is considered.

Recently, a number of research studies were done for the analysis of customer satisfaction. Quantitative approaches based on multi regression analysis, descriptive statistics, factor analysis, discriminant analysis, probability approach, ideal point approach and consumer behavioural analysis are the most important approaches among others used for the evaluation of customer satisfaction [3]. Consumer behavioural analysis is the well-distributed approach used for the evaluation of customer satisfaction.

The evaluation of customer satisfaction using statistical methods is basically based on the number of customers who survived. The constructed surveys are typically in the form of linguistic terms. The estimation of customer satisfaction is based on several subjective criteria, each of which includes several variables. The evaluation of each criterion is based on the responses of respondents to the survey in the form of preference agreements. The values of these linguistic terms are measured by numeric values. Percentage and statistical mean are widely used for the analysis of data. However, the mean value is not an adequate approach for evaluation. The values between ordinal values are not interpretable. For example, a mean value of 3.6 in five points Likert scale does not mean that the output is “satisfied”. Ordinal values are not the best way to represent linguistic values. The linguistic values “satisfied” or “not satisfied” are preferences’ degrees, which in fact are fuzzy values. Therefore, the use of these numbers does not offer the best way of representing linguistic terms. Fuzzy sets can be efficiently used for the description of the linguistic terms that present preference level or satisfaction degrees of customers [4]. The membership degree used in the fuzzy set is very useful in representing the degree of preferences of respondents. Fuzzy sets theory can adequately present the imprecise meaning of preference levels and their subjective nature [5]. In the literature, a set of research studies related to the measurement of customer satisfaction using fuzzy set theory are presented. The reference [6] used a fuzzy weighted rule mining approach and extract the relationship between customer satisfaction and product form features. The specific parameter guidelines are extracted for the enterprise’s business decision. The paper [7] presented the method for the evaluation of customer satisfaction using a combination of fuzzy triangular numbers, linguistic variables and fuzzy entropy. The paper [3] used a fuzzy inference system for evaluation of the individual customer satisfaction. The system uses some linguistic statements and knowledge from an informant and experienced people. This knowledge defines the computational structure of this method and can simulate the behaviour of each customer. [8] presented an approach that uses perception and individual satisfaction for evaluating customer satisfaction. In [9] the application of Fuzzy Cognitive map as a decision-making tool in the banking industry is considered. [10] presented fuzzy rule-based systems used for rating customer satisfaction. The system can present the case of the existence of a nonlinear relationship between input and outputs. The reference [11] presented integrated fuzzy regression –data envelopment analysis method for evaluation of customer satisfaction. The model describes the relationship between customer satisfaction and new product design. [12] investigated the different combinations of service quality after customer satisfaction. The study demonstrated that the combination of service qualities comfort, connection and convenience leads to customer satisfaction. [13] used a fuzzy set qualitative comparative analysis to investigate the relationship between environmental characteristics such as restaurant location, individual characteristics of customers and the psychological influences regarding a customer’s emphasis on service attitudes. Analysis showed that this relationship offered the critical antecedents to customer satisfaction and stickiness. The study is used for the analysis of customer satisfaction for restaurants. Based on fuzzy set theory [14] presented a model of customer satisfaction index in e-commerce. [15] used fuzzy set theory in order to demonstrate the effect of environmental and emotional characteristics on customer satisfaction and stickiness. The reference [16] used fuzzy set theory to determine customer satisfaction through sensory evaluation and applied the presented approach to milk barberry drinks. [17] used fuzzy conjoint analysis for measuring the level of customer satisfaction in cleanliness on the food premises.

Conjoint analysis was developed by Luce and Tukey [18], while the detailed consumer-oriented approach was given by Green and Rao [19]. Conjoint analysis is a survey-based technique that can be used to measure the consumers’ preference of a product in the market. Conjoint analysis using fuzzy set theory was proposed by Turksen and Wilson in [20] for analysis of consumer preferences. Fuzzy conjoint analysis is used for solving various problems, such as job satisfaction [5, 21], customer satisfaction [17]. In these researches, the Likert scale is applied to represent linguistic terms for measuring satisfaction degrees. [20] demonstrated that the fuzzy conjoint method that uses linguistic terms is a more appropriate tool for the analysis of data and determining the satisfaction of respondents. The fuzzy conjoint model gives the flexibility to deal with linguistic attributes. Using conjoint analysis the preferences of different individuals are measured and then integrated for decision making.

The constructed surveys used for measuring satisfaction levels in the above research papers typically use linguistic terms to measure customer preferences. Fuzzy sets are used to represent the opinions of respondents presented in linguistic forms. A set of research papers used fuzzy set theory for the evaluation of information, for decision making. The reference [22] integrated fuzzy set theory into decision making to enrich the value of information and model the human thinking process. By integrating fuzzy logic with the different artificial intelligence methodologies decision-making systems were designed [23 –25] and used in control. Sometimes, type-1 fuzzy systems cannot handle the uncertainties associated with information. For example, different expert opinions may be used in expressing linguistic values and evaluating preferences. Different experts may specify the numeric transformation of the linguistic values differently. Also, when respondents answer questions they can use different numeric transformations for expressing linguistic values. In such cases, type-1 fuzzy sets cannot handle these uncertainties. Type-2 fuzzy sets that are an extension of type-1 fuzzy sets can be an excellent framework for handling these uncertainties. Mendel [26, 27] specified the sources of uncertainties in type-2 fuzzy sets. These are: (1) the answers of experts to the same question in a different manner (different people different words), (2) the estimation of the membership function of the same linguistic value by different experts, (3) the noise of measurements that activate type-1 fuzzy system, and (4) noisy data used to turn parameters of type-1 fuzzy systems [26]. Type 2 fuzzy sets were proposed by Zadeh [27] and later improved by Mendel and his students [26, 28]. Because of handling of increased uncertainties and fuzziness in information type-2 fuzzy sets are gaining more popularity. A number of research studies have used type-2 fuzzy sets to solve different scientific problems. These are decision making [29, 30], identification and control of dynamic plants [31 –33], channel equalization [34], time-series applications [35, 36], estimation of energy performance of buildings [36]. The research works demonstrated the advantages of using type-2 fuzzy sets in handling uncertainties, imprecision. This paper proposes the type-2 fuzzy system that employs conjoint analysis for the determination of the satisfaction degree of customers. In the paper, type-2 fuzzy set theory is integrated with conjoint analysis for the multisensory evaluation of customer satisfaction. The contributions of the paper are: Type-2 fuzzy sets and conjoint analysis are proposed to determine the degrees of satisfaction of respondents using multisensory attributes; Type-2 fuzzy similarity is proposed for estimating the similarity degrees between customer opinions and opinions of experts for ranking and evaluation of customer satisfaction; The algorithm that combines conjoint analysis and type-2 fuzzy similarity degree is designed for the evaluation of customer satisfaction; The type-2 fuzzy system that integrated type-2 fuzzy set theory and conjoint analysis is designed for sensory evaluation of customer satisfaction.

The paper is organised as follows. Section 2 describes the structure and development of the proposed type-2 fuzzy model for sensory evaluation. Section 3 describes experimental results obtained. Section 4 depicts the conclusions of the paper.

2 Type-2 fuzzy model for evaluating sensory marketing

In real life, there are lots of concerns that can be identified and measured using linguistic values. The numerical meaning of these linguistic values can be described by fuzzy sets. In the paper, fuzzy sets are used to represent linguistic data obtained by sensory evaluation of preferences. The measurement is based on a Liker-type scale, which uses linguistic values. These linguistic values are ordered and used to represent the preference levels of respondents. The use of fuzzy sets on a Likert-type scale allows a more flexible evaluation of considered attributes. The level of the agreement will not be crisp numbers, it will accept any value between 0 and 1. This will make a more flexible evaluation of attributes. Here based on expert beliefs, a different number of fuzzy terms may be used for evaluation. This allows a distinct evaluation of attributes by respondents.

Fuzzy set theory has the potential to explain concepts sets of human language that cannot be presented by traditional set theory. Fuzzy sets can be efficiently used to represent linguistic data obtained by sensory evaluation. Fuzzy sets can also provide a good framework for describing the imprecise meaning of preferences of respondents. Fuzzy sets use membership functions that provide a gradual (progressive) transition from membership to non-membership. That is, the linguistic value in fuzzy sets is defined as degrees in an interval of 0 and 1. It has been shown that fuzzy sets can be employed to induce models from trendy and subjective concepts for evaluation purposes. There may be different expert opinions in the expression of linguistic values and evaluation of preferences. For example, different experts may specify the numeric transformation of the linguistic value “Normal” differently. (Membership functions of linguistic value “Normal” can be formulated differently with the different experts). In such cases, type-1 fuzzy sets cannot handle the uncertainties. Type-2 fuzzy sets that extension of type-1 fuzzy sets can be an excellent framework for handling these uncertainties [26]. Uncertainties in type-2 fuzzy sets can arise from different sources. These uncertainties are described in [26, 28]. In the paper, the type-2 fuzzy set is employed to analyze the impact of sensory marketing on customer satisfaction. Because of membership functions of the general type-2 fuzzy set is tree-dimensional it can directly model uncertainties. The increase of dimension complicates computations using type-2 fuzzy sets. To simplify the calculation, the interval type-2 fuzzy membership function is applied to describe these uncertainties. The membership functions can be represented using triangle, trapezoid, bell-shape etc. forms. One of the widely used ones is the triangle type membership function. These are appropriate for describing vague information for most decision-making problems. Triangular membership functions have less number of parameters and they are computationally efficient. Therefore, due to its simplicity, the triangular type-2 membership functions are used to represent the customer opinions on the paper. Figure 1 (a,b) depicts triangular type interval type-2 fuzzy membership function. In figure (a) uncertainty is associated with the center of membership function, in figure (b) uncertainty is associated with the width (left and right parts) of the membership function.

Fig. 1

Type-2 membership function.

Each point in interval type-2 fuzzy sets is characterised by two membership functions (MFs) called lower and upper membership functions as ${\underline{μ}}_{F} (x)$ and ${\bar{μ}}_{F} (x)$ (or F^L and F^U) (see Fig. 1(a)). Using lower and upper MFs we can do different operations on type-2 fuzzy sets.

The measurement of customer satisfaction is based on the evaluation of expert-defined parameters. The parameters are evaluated by the preference levels of respondents. Type-2 fuzzy sets are used to represent the customer’s preference levels described by linguistic values. In this research, type-2 fuzzy sets is integrated with conjoint analysis to determine the satisfaction levels of respondents.

As mentioned above, conjoint analysis was proposed by Luce and Tukey [18]. Conjoint analysis is a survey-based technique that can be used to measure consumers’ preferences for products in the market. The conjoint measurement is an overall preference for a product or service that can be decomposed into a combination of preferences for its constituent parts, which are combined with appropriate combination function [18 –20]. This combination function appropriately uses linguistic variables and produces an overall linguistic preference. Conjoint models represent linguistic preferences as a ratio (or interval scaled numbers), use numeric product attributes and require aggregation of individuals for the estimation process. The fuzzy conjoint model proposed by Turksen and Wilson [20], provides flexibility for dealing with linguistic attributes. Conjoint analysis is a survey-based technique that can be used to measure the consumers’ preference of a product in the market. The advantages of conjoint analysis are as follows: Conjoint analysis is performed using multiple physical variables (in given case vision, hearing, touching, smelling and tasting) unlike other tools that use one variable at a time; Unlike classical approach that uses “yes” and “no”, the conjoint analysis provides preference measure allowing more clearly the understanding of consumer behaviour and preferences; The fuzzy conjoint model gives the flexibility to deal with linguistic attributes; Conjoint analysis uses the survey results to calculate the numerical value that measures how much each attribute influenced the consumer’s preference; Using conjoint analysis the preferences of different individuals are measured and then integrated for decision making.

In this paper, we present a type-2 fuzzy version of conjoint analysis for evaluation of customer satisfaction. The type-2 membership function of each element y_j for the linguistic label A in fuzzy set R is defined as ${\tilde{μ}}_{R} (y_{j}, A) = \sum_{i = 1}^{n} [\frac{w_{i}}{\sum w_{i}}] \cdot {\tilde{μ}}_{F_{i}} (x_{j}, A)$ (1)

Where ${\tilde{μ}}_{F_{i}} (x_{j}, A)$ are type-2 membership functions characterised by lower ${\underline{μ}}_{F_{i}} (x_{j}, A)$ and upper ${\bar{μ}}_{F_{i}} (x_{j}, A)$ membership degrees (Fig. 1(a)) for respondent j for item A according to linguistic label x_i = 1,2, ... ,n. A is an item/a question, n is a number of linguistic terms. w_i/∑w_i is the weight representing the level of satisfaction of i-th respondent, w_i is a score of linguistic values of i-th respondent.

The evaluation is based on the finding similarity between expert preferences and respondents’ preferences presented by type-2 fuzzy sets. The preferences of respondents are based on statistics collected from customers. The fuzzy sets characterising membership degrees of the responses of respondents R is compared with the expert opinions represented as type-2 membership function F. Wu and Mendel proposed the similarity measure (SM) for type-2 fuzzy sets which was an extension of Jaccard’s similarity measure used for type 1 fuzzy sets. The similarity degrees S_j(R,F) between type-2 fuzzy sets F and R are calculated as [37] $\begin{matrix} S_{j} (R, F) = \frac{p (R \cap F)}{p (R \cup F)} = \\ \frac{\int_{X}^{*} min ({\bar{μ}}_{R} (x), {\bar{μ}}_{F} (x)) dx + \int_{X}^{*} min ({\underline{μ}}_{R} (x), {\underline{μ}}_{F} (x)) dx}{\int_{X}^{*} max ({\bar{μ}}_{R} (x), {\bar{μ}}_{F} (x)) dx + \int_{X}^{*} max ({\underline{μ}}_{R} (x), {\underline{μ}}_{F} (x)) dx} \end{matrix}$ (2) where F is the standard fuzzy sets determined for the expert linguistic label, R is the fuzzy sets that were determined using the responses of respondents. ${\underline{μ}}_{R} (x), {\underline{μ}}_{F} (x) and {\bar{μ}}_{R} (x), {\bar{μ}}_{F} (x)$ are lower and upper MFs of type-2 fuzzy sets R and F, respectively. S_j(R,F) is similarity degree.

The similarity rate ranges from 0 to 1 and it is calculated for each attribute. Here, the fuzzy sets F contains type-2 membership functions of linguistic variables which have a subjective nature. The triangular forms are used to represent the membership functions. These membership functions can be defined according to experts’ opinions (see Fig. 1). Fuzzy set R is the membership functions of respondents that contain all the states to be ranked. Formula (2) demonstrates how experts’ opinions correspond to the customers’ satisfaction.

3 Experimental results

The proposed approach is used to measure the satisfaction of the customers of retail markets. For this purpose, a questionnaire that includes five sensory attributes is used. A 7 point Likert scale is used for the evaluation of each attribute. The survey was collected through an online survey and social media. A total of 193 responses were collected in the time frame. Cronbach’s alpha was measured to evaluate the reliability of the designed questionnaires for Likert-scale questions and verify the internal accuracy of these scale questions. Cronbach alpha coefficient was above 0.9, which means a high level of internal reliability.

Sensory marketing variables (Sight, Touch, Taste, Smell, and Sound) are used to evaluate customer satisfaction in supermarkets. Each of these variables has different questions related to satisfaction feelings in supermarkets. Totally 27 sensory marketing questions were used (Appendix 1). As mentioned before, 193 questionnaires were distributed over 4 big supermarkets in Erbil using a 7 point Likert Scale to understand the relationships between sensory marketing and customer satisfaction in these supermarkets. The linguistic values are employed for the evaluations of each question. They are presented as follows: Strongly Disagree (SD), Disagree (D), Somewhat Disagree (SWD), Neutral (N), Somewhat Agree (SWA), Agree (A), and Strongly Agree (SA). The triangular type-2 MFs are utilised to represent the linguistic values. Figure 2 depicts the type-2 MFs used for the estimation of satisfaction levels.

Fig. 2

Type-2 fuzzy sets used for estimation of satisfaction levels.

The respondents have answered all 27 questions with agreement degrees. The shoppers’ level of agreement for each question in terms of 7 point Likert scale is given in Table 2. The table includes the numbers and percentages of shoppers. For example, for item 1 which is “I feel the store inside is bright”: 0% of shoppers Strongly Disagreed, 0.5% - Disagreed, 5.2% - Somewhat Disagreed, 19.7% - Neutral, 27.5% - Somewhat Agreed, 22.3% - Agreed, and 24.9% - Strongly Agreed.

The following steps were used for the sensory evaluation of customer satisfaction.

The flowchart of the algorithm is given in Fig. 3. In the first step, the design of the survey instrument is performed. Here, basic criteria and variables are defined, and linguistic terms used for evaluations of the variables are specified. Using defined criteria the linguistic questionnaire is created. The 27 questions that were suitable for customers are selected and included in the survey. In the questionnaire, for evaluation of each question, the seven points-Likert scale is used. The questionnaires are distributed among customers. In the second step, the results of the survey are collected and entered into the system. After the survey was conducted, a Cronbach’s alpha was calculated to check the reliability of the designed questionnaires for Likert-scale questions and verify the internal accuracy of these scale questions. In the third step, the type-2 membership functions are defined for expressing linguistic values. In the fourth step based on expert opinions, statistics and using formula (1) membership degrees of customer opinions for each question is calculated. In the fifth step, using formula (2), the similarity degrees between customers’ opinions and experts’ opinions are determined. In the sixth step, the ranking is performed by ordering the similarity degrees. In the seventh step, the evaluation of customer satisfaction using the ranking results is carried out.

Fig. 3

The basic steps for the sensory evaluation of customer satisfaction.

At first, all attributes, preference levels, and all states are formulated. The questionnaire for sensory marketing and customer satisfaction was developed using attributes and preference levels. As mentioned above, 27 questions are applied. Agreement levels are determined using linguistic terms “Strongly Disagree”, “Disagree”, “Somewhat Disagree”, “Neutral”, “Somewhat Agree”, “Agree”, and “Strongly Agree”. Seven membership functions defining experts’ opinions are presented using Table 1. As shown each linguistic value is represented by type-2 membership functions.

Table 1

Membership degrees of linguistic terms

Linguistic Terms	Membership degrees
Strongly Agree	F1={(0.9,1)/1, (0.6,0.7)/2, (0.3,0.4)/3, (0,0.1)/4, (0,0)/5, (0,0)/6, (0,0)/7}
Agree	F2={(0.6,0.7)/1, (0.9,1)/2, (0.6,0.7)/3, (0.3,0.4)/4, (0,0.1)/5, (0,0)/6, (0,0)/7}
Somewhat Agree	F3={(0.3,0.4)/1, (0.5,0.6)/2, (0.9,1)/3, (0.5,0.6)/4, (0.3,0.4)/5, (0,0)/6, (0,0)/7}
Neutral	F4={ (0,0)/1, (0.2,0.3)/2, (0.6,0.7)/3, (0.9,1)/4, (0.6,0.7)/5, (0.2,0.3)/6, (0,0)/7}
Somewhat Disagree	F5={ (0,0)/1, (0.1,0.2)/2, (0.3,0.4)/3, (0.5,0.6)/4, (0.9,1)/5, (0.5,0.6)/6, (0.3,0.4)/7}
Disagree	F6={ (0,0)/1, (0,0)/2, (0,0)/3, (0.3,0.4)/4, (0.5,0.6)/5, (0.9,1)/6, (0.6,0.7)/7}
Strongly Disagree	F7={ (0,0)/1, (0,0)/2, (0,0)/3, (0.1,0.2)/4, (0.4,0.5)/5, (0.7,0.8)/6, (0.9,1)/7};

In the second step, the customer satisfaction questionnaire is constructed using attributes and preference levels and distributed among supermarket customers. The opinions of supermarket customers for each attribute in the questionnaire were collected from the customers of different retail markets. The responses were analysed based on customers’ opinions regarding a selected linguistic variable. A total of 193 customers participated in the review. Table 2 describes the results of evaluations. In the table, for the first item, 48 customers Strongly Agree (24.9% of customers), 43 (22.3%) Agree, 53(27.5%) somewhat agree, 38 (19.7%) neutral, 10 (5.2%) somewhat disagree and 1 (0.5%) Disagree. The numerical meaning of linguistic values is presented in Table 1.

Table 2

Customer sensory evaluation results

	SA		DA		SDA		N		SWD		A		SA
Questions	Fr.	%	Fr.	%	Fr.	%	Fr.	%	Fr.	%	Fr.	%	Fr.	%
Item 1	0	0.0%	1	0.5%	10	5.2%	38	19.7%	53	27.5%	43	22.3%	48	24.9%
Item 2	8	4.1%	2	1.0%	20	10.4%	44	22.8%	40	20.7%	36	18.7%	43	22.3%
Item 3	0	0.0%	10	5.2%	17	8.8%	28	14.5%	31	16.1%	39	20.2%	68	35.2%
Item 4	0	0.0%	6	3.1%	8	4.1%	26	13.5%	30	15.5%	41	21.2%	82	42.5%
Item 5	3	1.6%	6	3.1%	10	5.2%	32	16.6%	34	17.6%	44	22.8%	64	33.2%
Item 6	6	3.1%	5	2.6%	8	4.1%	38	19.7%	42	21.8%	38	19.7%	56	29.0%
Item 7	1	0.5%	1	0.5%	3	1.6%	19	9.8%	31	16.1%	47	24.4%	91	47.2%
Item 8	0	0.0%	0	0.0%	9	4.7%	17	8.8%	16	8.3%	29	15.0%	122	63.2%
Item 9	7	3.6%	13	6.7%	18	9.3%	26	13.5%	22	11.4%	35	18.1%	72	37.3%
Item 10	15	7.8%	17	8.8%	15	7.8%	32	16.6%	25	13.0%	17	8.8%	72	37.3%
Item 11	9	4.7%	5	2.6%	6	3.1%	26	13.5%	18	9.3%	30	15.5%	99	51.3%
Item 12	12	6.2%	9	4.7%	6	3.1%	31	16.1%	25	13.0%	44	22.8%	66	34.2%
Item 13	6	3.1%	0	0.0%	7	3.6%	52	26.9%	50	25.9%	32	16.6%	46	23.8%
Item 14	2	1.0%	4	2.1%	17	8.8%	32	16.6%	39	20.2%	49	25.4%	50	25.9%
Item 15	3	1.6%	12	6.2%	17	8.8%	24	12.4%	36	18.7%	35	18.1%	66	34.2%
Item 16	22	11.4%	17	8.8%	12	6.2%	23	11.9%	26	13.5%	37	19.2%	56	29.0%
Item 17	17	8.8%	11	5.7%	9	4.7%	23	11.9%	18	9.3%	44	22.8%	71	36.8%
Item 18	5	2.6%	1	0.5%	7	3.6%	29	15.0%	45	23.3%	39	20.2%	67	34.7%
Item 19	2	1.0%	4	2.1%	7	3.6%	37	19.2%	50	25.9%	27	14.0%	66	34.2%
Item 20	5	2.6%	2	1.0%	14	7.3%	42	21.8%	49	25.4%	21	10.9%	60	31.1%
Item 21	9	4.7%	3	1.6%	13	6.7%	33	17.1%	41	21.2%	32	16.6%	62	32.1%
Item 22	13	6.7%	18	9.3%	17	8.8%	40	20.7%	22	11.4%	32	16.6%	51	26.4%
Item 23	19	9.8%	11	5.7%	23	11.9%	34	17.6%	30	15.5%	23	11.9%	53	27.5%
Item 24	9	4.7%	7	3.6%	8	4.1%	27	14.0%	25	13.0%	30	15.5%	87	45.1%
Item 25	13	6.7%	3	1.6%	7	3.6%	34	17.6%	30	15.5%	22	11.4%	84	43.5%
Item 26	15	7.8%	7	3.6%	8	4.1%	22	11.4%	15	7.8%	23	11.9%	103	53.4%
Item 27	94	48.7%	20	10.4%	18	9.3%	24	12.4%	12	6.2%	9	4.7%	16	8.3%

Each attribute in the questionnaire was analysed. In the third step, the degree of satisfaction of each attribute is determined. The operations were carried out by measuring the weight and correspondingly by the degree of membership μ of respondent j for item C according to the linguistic term. The values of lower ${\underline{μ}}_{F_{i}} (x_{j}, A)$ and upper ${\bar{μ}}_{F_{i}} (x_{j}, A)$ membership degrees were determined in Table 3. In the fourth step, the values of the similarity degree between customers’ opinions and the opinions of experts are determined. This operation is carried out by calculating the degree of similarity between R and F fuzzy sets using the formula (2). The exact number of similarity degrees for each state is calculated in the last fifth step. The extent of correlation reflects the maximum closeness of the viewpoints of experts and consumers to each other. The degrees of similarity between the F and R fuzzy sets are seen in Table 4.

Table 3

The Values of lower ${\underline{μ}}_{F_{i}} (x_{j}, A)$ and upper ${\bar{μ}}_{F_{i}} (x_{j}, A)$ membership degrees of fuzzy sets R

	${\underline{μ}}_{R} (x)$							${\bar{μ}}_{R} (x)$
	F1	F2	F3	F4	F5	F6	F7	F1	F2	F3	F4	F5	F6	F7
C1	0.0187	0.0948	0.2503	0.4415	0.4477	0.5736	0.4399	0.0440	0.1477	0.3254	0.5415	0.6891	0.6731	0.5342
C2	0.0746	0.1648	0.3109	0.4508	0.4104	0.4943	0.3746	0.1130	0.2238	0.3886	0.5508	0.6290	0.5891	0.4591
C3	0.0575	0.1430	0.2456	0.3560	0.3187	0.5332	0.4865	0.0860	0.1876	0.3104	0.4560	0.5850	0.6280	0.5725
C4	0.0311	0.0933	0.1834	0.3124	0.2969	0.5798	0.5565	0.0518	0.1295	0.2409	0.4124	0.5881	0.6767	0.6492
C5	0.0482	0.1181	0.2223	0.3637	0.3420	0.5596	0.4881	0.0746	0.1622	0.2891	0.4637	0.6083	0.6549	0.5782
C6	0.0560	0.1259	0.2456	0.3995	0.3855	0.5409	0.4446	0.0855	0.1772	0.3166	0.4995	0.6285	0.6352	0.5347
C7	0.0124	0.0466	0.1259	0.2689	0.2813	0.6280	0.6187	0.0249	0.0751	0.1788	0.3689	0.5953	0.7269	0.7161
C8	0.0140	0.0544	0.1197	0.2021	0.1865	0.5907	0.6839	0.0275	0.0762	0.1565	0.3021	0.5212	0.6907	0.7793
C9	0.1010	0.1788	0.2503	0.3202	0.2658	0.4959	0.4788	0.1342	0.2233	0.3130	0.4202	0.5285	0.5855	0.5591
C10	0.1461	0.2223	0.2845	0.3264	0.2658	0.4306	0.4275	0.1870	0.2762	0.3472	0.4264	0.4964	0.5140	0.5031
C11	0.0668	0.1104	0.1663	0.2503	0.2207	0.5440	0.5829	0.0907	0.1435	0.2150	0.3503	0.5166	0.6368	0.6725
C12	0.0933	0.1461	0.2098	0.3233	0.2907	0.5363	0.4834	0.1233	0.1891	0.2756	0.4233	0.5554	0.6254	0.5694
C13	0.0389	0.1212	0.2813	0.4694	0.4554	0.5285	0.3917	0.0725	0.1808	0.3575	0.5694	0.6736	0.6254	0.4850
C14	0.0482	0.1275	0.2549	0.4057	0.3839	0.5549	0.4461	0.0767	0.1762	0.3290	0.5057	0.6368	0.6518	0.5342
C15	0.0777	0.1554	0.2518	0.3497	0.3233	0.5176	0.4725	0.1067	0.2031	0.3176	0.4497	0.5788	0.6098	0.5560
C16	0.1741	0.2207	0.2549	0.3093	0.2689	0.4632	0.4166	0.2124	0.2725	0.3259	0.4093	0.5021	0.5430	0.4902
C17	0.1275	0.1679	0.2021	0.2767	0.2378	0.5176	0.4959	0.1585	0.2083	0.2653	0.3767	0.5078	0.6031	0.5767
C18	0.0373	0.0870	0.2036	0.3591	0.3715	0.5751	0.5036	0.0591	0.1321	0.2689	0.4591	0.6337	0.6720	0.5969
C19	0.0326	0.1041	0.2409	0.3979	0.4010	0.5440	0.4694	0.0585	0.1560	0.3067	0.4979	0.6446	0.6409	0.5627
C20	0.0513	0.1337	0.2860	0.4275	0.4135	0.5021	0.4212	0.0839	0.1917	0.3549	0.5275	0.6368	0.5984	0.5104
C21	0.0715	0.1337	0.2503	0.3762	0.3637	0.5207	0.4523	0.1016	0.1850	0.3181	0.4762	0.6052	0.6145	0.5394
C22	0.1430	0.2394	0.3140	0.3855	0.3031	0.4383	0.3715	0.1886	0.2964	0.3876	0.4855	0.5254	0.5223	0.4466
C23	0.1585	0.2347	0.3233	0.3762	0.3171	0.4181	0.3653	0.2036	0.2953	0.3959	0.4762	0.5254	0.5026	0.4378
C24	0.0762	0.1275	0.1959	0.2860	0.2596	0.5301	0.5378	0.1026	0.1668	0.2508	0.3860	0.5368	0.6218	0.6254
C25	0.0808	0.1290	0.2145	0.3124	0.2907	0.5098	0.5067	0.1104	0.1741	0.2710	0.4124	0.5487	0.6016	0.5948
C26	0.1041	0.1383	0.1741	0.2207	0.1865	0.5083	0.5751	0.1311	0.1731	0.2207	0.3207	0.4746	0.5969	0.6596
C27	0.5285	0.4788	0.3855	0.2503	0.1725	0.1663	0.1212	0.6093	0.5658	0.4772	0.3503	0.2627	0.2073	0.1528

Table 4

The values of similarity degrees between fuzzy sets F and R (Wu-Mendel SM)

	The similarity degrees between fuzzy sets F and C							Maximum similarity degree for each state	Ran-king	Linguistic values
	F1	F2	F3	F4	F5	F6	F7
C1	0.1190	0.1901	0.3113	0.5016	0.6460	0.6533	0.4872	0.6533	22	Agree
C2	0.1734	0.2454	0.3773	0.5516	0.6322	0.5543	0.4170	0.6322	17	Somewhat Agree
C3	0.1442	0.2169	0.3174	0.4603	0.5550	0.6187	0.5080	0.6187	12	Agree
C4	0.1041	0.1785	0.2634	0.3989	0.5247	0.6964	0.5880	0.6964	26	Agree
C5	0.1275	0.2003	0.3015	0.4514	0.5648	0.6458	0.5229	0.6458	20	Agree
C6	0.1389	0.2109	0.3233	0.4878	0.5957	0.6197	0.4829	0.6197	13	Agree
C7	0.0696	0.1371	0.2143	0.3456	0.4863	0.7492	0.6505	0.7492	27	Agree
C8	0.0698	0.1216	0.1870	0.3035	0.4314	0.6699	0.6899	0.6899	25	Strongly Agree
C9	0.1722	0.2453	0.3322	0.4501	0.5145	0.5765	0.4833	0.5765	5	Agree
C10	0.2144	0.2868	0.3788	0.4785	0.5095	0.5118	0.4251	0.5118	1	Agree
C11	0.1163	0.1783	0.2501	0.3670	0.4634	0.6465	0.5836	0.6465	21	Agree
C12	0.1476	0.2212	0.3113	0.4342	0.5198	0.6144	0.5146	0.6144	11	Agree
C13	0.1423	0.2129	0.3459	0.5403	0.6593	0.6007	0.4409	0.6593	23	Somewhat Agree
C14	0.1387	0.2103	0.3240	0.4905	0.6045	0.6239	0.4861	0.6239	16	Agree
C15	0.1563	0.2288	0.3284	0.4681	0.5503	0.6017	0.4915	0.6017	9	Agree
C16	0.2137	0.2861	0.3725	0.4619	0.4983	0.5216	0.4339	0.5216	2	Agree
C17	0.1641	0.2313	0.3074	0.4092	0.4771	0.5833	0.5062	0.5833	6	Agree
C18	0.1100	0.1831	0.2813	0.4391	0.5717	0.6801	0.5416	0.6801	24	Agree
C19	0.1240	0.1962	0.3066	0.4779	0.6021	0.6454	0.4995	0.6454	19	Agree
C20	0.1514	0.2227	0.3449	0.5222	0.6228	0.5921	0.4491	0.6228	15	Somewhat Agree
C21	0.1478	0.2202	0.3270	0.4811	0.5760	0.6075	0.4813	0.6075	10	Agree
C22	0.2239	0.2969	0.4150	0.5248	0.5505	0.4910	0.3987	0.5505	4	Somewhat Agree
C23	0.2299	0.3045	0.4202	0.5276	0.5482	0.4834	0.3879	0.5482	3	Somewhat Agree
C24	0.1327	0.2035	0.2815	0.4041	0.4931	0.6346	0.5463	0.6346	18	Agree
C25	0.1406	0.2150	0.3020	0.4278	0.5147	0.6209	0.5221	0.6209	14	Agree
C26	0.1399	0.1930	0.2609	0.3613	0.4346	0.5931	0.5545	0.5931	7	Agree
C27	0.5190	0.5962	0.5358	0.4143	0.3042	0.2163	0.1568	0.5962	8	Disagree

The maximum value of similarity degrees for each attribute was chosen after the similarity degrees were measured. The ranking is based on the maximum degree of similarity between all states. The right column in Table 4 displays the highest similarity degrees and the ranking results. As shown in the table, the results obtained are basically “Agree” that have the best rank. There are one “Strongly Agree”, five “Somewhat Agree” and one “Disagree” linguistic values. The seventh state has the best rank. The state 7 is obtained as 0.5% strongly disagree, 0.5% disagree, 1.6% somewhat disagree, 9.8% neutral, 16.1% somewhat agree, 24.4% agree and 47.2% strongly agree. States 4 and 8 are the next states that have the best ranks. The worst state is 27 which has 48.7% strongly disagree, 10.4% disagree, 9.3% somewhat disagree, 12.4% neutral, 6.2% somewhat agree, 4.7% agree, and 8.3% strongly agree. States 10 and 16 have very low ranks. Since the model’s input parameters are fuzzy variables, the proposed model’s output (rankings) would be fuzzy.

Next, for comparative purposes, we applied Zeng-Li [30] similarity measure (SM) to determine customer satisfaction degrees also. Zeng-Li SM is presented as $S_{z} (F, R) = 1 - \frac{1}{2} \sum_{i = 1}^{n} | {\underline{μ}}_{F} (x_{i}) - {\underline{μ}}_{R} (x_{i}) | + | {\bar{μ}}_{F} (x_{i}) - {\bar{μ}}_{R} (x_{i}) |$ (3) where F is the standard fuzzy sets determined using the expert opinions, R is the fuzzy sets determined using the responses of the respondents. ${\underline{μ}}_{R} (x), {\underline{μ}}_{F} (x) and {\bar{μ}}_{R} (x), {\bar{μ}}_{F} (x)$ are lower and upper membership functions, respectively.

The similarity is computed for each of the n possible linguistic terms. Analogously to the previous method given above, triangular forms are used to represent type-2 membership functions of F. Formula (3) demonstrates how experts’ opinions correspond to the customers’ satisfaction. Simulation has been done using statistical data and Zheg-Li SM. Table 5 depicts the similarity degrees between fuzzy sets F and R. The right column in Table 5 shows the highest similarity degrees and the ranking results. As shown in the table, the obtained results are basically “Agree”. There are two “Strongly Agree”, five “Somewhat Agree”. The only last state is “Disagree”. The best ranking states are C7 and C8. The worst ranking states are C10 and C16.

Table 5

The values of similarity degrees between fuzzy sets F and R (Zheng-Li SM)

	The similarity degrees between fuzzy sets F and C							Maximum similarity degree for each state	Ranking	Linguistic values
	F1	F2	F3	F4	F5	F6	F7
C1	0.4815	0.4886	0.5790	0.7316	0.8368	0.8424	0.7729	0.8424	21	Agree
C2	0.5354	0.5441	0.6372	0.7660	0.8288	0.7842	0.7286	0.8288	16	Somewhat Agree
C3	0.5210	0.5282	0.5942	0.7079	0.7881	0.8273	0.7910	0.8273	14	Agree
C4	0.4899	0.4970	0.5501	0.6654	0.7728	0.8709	0.8369	0.8709	25	Agree
C5	0.5040	0.5111	0.5793	0.7011	0.7939	0.8421	0.7992	0.8421	20	Agree
C6	0.5099	0.5170	0.5954	0.7252	0.8104	0.8260	0.7740	0.8260	11	Agree
C7	0.4614	0.4597	0.5070	0.6260	0.7515	0.8979	0.8689	0.8979	27	Agree
C8	0.4780	0.4571	0.4909	0.5991	0.7218	0.8630	0.8898	0.8898	26	Strongly Agree
C9	0.5533	0.5605	0.6114	0.7033	0.7652	0.8052	0.7797	0.8052	5	Agree
C10	0.5909	0.5981	0.6507	0.7241	0.7620	0.7658	0.7449	0.7658	1	Agree
C11	0.5156	0.5085	0.5472	0.6472	0.7389	0.8487	0.8391	0.8487	23	Agree
C12	0.5307	0.5379	0.5932	0.6916	0.7688	0.8269	0.7976	0.8269	13	Agree
C13	0.5038	0.5110	0.6094	0.7580	0.8438	0.8120	0.7436	0.8438	22	Somewhat Agree
C14	0.5067	0.5138	0.5940	0.7258	0.8146	0.8274	0.7745	0.8274	15	Agree
C15	0.5325	0.5396	0.6038	0.7135	0.7851	0.8176	0.7816	0.8176	7	Agree
C16	0.5899	0.5970	0.6452	0.7117	0.7546	0.7718	0.7500	0.7718	2	Agree
C17	0.5527	0.5531	0.5943	0.6759	0.7441	0.8116	0.7957	0.8116	6	Agree
C18	0.4869	0.4941	0.5610	0.6921	0.7983	0.8605	0.8097	0.8605	24	Agree
C19	0.4957	0.5029	0.5809	0.7184	0.8145	0.8406	0.7840	0.8406	19	Agree
C20	0.5189	0.5260	0.6124	0.7480	0.8250	0.8090	0.7518	0.8250	10	Somewhat Agree
C21	0.5223	0.5294	0.6012	0.7220	0.8000	0.8202	0.7747	0.8202	9	Agree
C22	0.5902	0.5993	0.6738	0.7519	0.7837	0.7477	0.7222	0.7837	4	Somewhat Agree
C23	0.5962	0.6066	0.6784	0.7541	0.7826	0.7430	0.7155	0.7826	3	Somewhat Agree
C24	0.5240	0.5272	0.5714	0.6725	0.7550	0.8403	0.8176	0.8403	18	Agree
C25	0.5288	0.5359	0.5881	0.6892	0.7675	0.8320	0.8036	0.8320	17	Agree
C26	0.5428	0.5273	0.5611	0.6453	0.7218	0.8215	0.8264	0.8264	12	Strongly Agree
C27	0.8026	0.8188	0.7684	0.6797	0.6140	0.5385	0.5455	0.8188	8	Disagree

We used Sperman’s rank correlation test to evaluate the statistical relations of the ranks of the alternatives. The Sperman’s rank correlation test can be used to determine the degree of dependence between two variables [38 –41]. Sperman’s rank correlation coefficient r_s can be determined from the ranks provided by each method [39]. $r_{s} = 1 - \frac{6 \sum_{i = 1}^{n} d_{i}^{2}}{n (n^{2} - 1)}, Z = r_{s} \sqrt{n - 1}$ (4)

Using (4), Z statistical test value is determined and compared with the specific predefined Z value [40]. Initially, the null hypothesis is stated as “There is no similarity between two rankings”. The alternative hypothesis would be “Two ranks are similar”. If the determined Z statistical test is more than 1.645, the null hypothesis is rejected.

The Z value for the Mendel-Wu and Zheng-Li method was determined as 4.0359. This value is greater than the predefined value of 1.645. This indicates that the null hypothesis is rejected and the difference between these two approaches is insignificant.

Table 6 depicts the comparisons of the results obtained by traditional “mean”, Mendel-Wu and Zheng-Li similarity measures. As shown in Table 6 the results obtained by Zeng-Li SM are almost the same as the results of Mendel-Wu SM. The satisfaction levels of parameters are the same for these two methods. There is only one difference in state C26 (“Loud music in-store annoying me”). As seen using Mendel-Wu SM the satisfaction level is obtained as “Agree”, but by Zheng-Li SM the satisfaction level is obtained as “Strongly Agree”. There are differences between the traditional “mean” method and the Mendel-Wu method. The satisfaction levels of states C1, C3, C5, C6, C9, C10, C12, C14, C15, C16, C17, C19, C21 and C25 for the “mean” method are “Somewhat Agree” but for the Mendel-Wu method they are “Agree”. Also, satisfaction levels of states C8 and C26 for the “mean” method are “Agree” for the “mean” method, but they are “Somewhat Agree” in the Mendel-Wu method. The satisfaction level of state C27 for the “mean” method is “Somewhat Agree”, but for the Mendel-Wu and Zheng-Li methods its value is “Disagree”. As shown, based on the mean value, the most items (C1, C3, C5, C6, C9, C10, C12, C14, C15, C16, C17, C19, C21 and C25) have satisfied as “Somewhat Agree” compared to conjoint analysis that used Mendel-Wu and Zheng-Li SMs, which are “Agree”. The results obtained using conjoint analysis are based on customer opinions about each attributes that were presented by linguistic values. One of the advantages of this approach is the integration of fuzzy theory with the Likert scale which allows more accurate evaluations of preference levels. As known, the conventional Likert scale uses integer numbers that don’t exactly describe human opinions. This reflects the actual differences between output categories [17]. The results demonstrated that the conjoint analysis provided more accurate results for customer satisfaction than the traditional “mean” based approach. The simulation results obtained show the effectiveness of using the proposed model in the evaluation of customer satisfaction.

Table 6

Comparative results

	Traditional			Mendel-Wu SM		Zheng-Li SM
R	Mean	Standard Deviation	Label of satisfaction	Ranking	Label of satisfaction	Ranking	Label of satisfaction
C1	5.4041	2.6395	Somewhat Agree	22	Agree	21	Agree
C2	5.0000	2.4152	Somewhat Agree	17	Somewhat Agree	16	Somewhat Agree
C3	5.4301	2.6557	Somewhat Agree	12	Agree	14	Agree
C4	5.7513	2.8714	Agree	26	Agree	25	Agree
C5	5.4663	2.6786	Somewhat Agree	20	Agree	20	Agree
C6	5.2953	2.5738	Somewhat Agree	13	Agree	11	Agree
C7	6.0207	3.0709	Agree	27	Agree	27	Agree
C8	6.2332	3.2380	Agree	25	Strongly Agree	26	Strongly Agree
C9	5.2591	2.5527	Somewhat Agree	5	Agree	5	Agree
C10	4.9378	2.3860	Somewhat Agree	1	Agree	1	Agree
C11	5.7202	2.8494	Agree	21	Agree	23	Agree
C12	5.3005	2.5768	Somewhat Agree	11	Agree	13	Agree
C13	5.1762	2.5061	Somewhat Agree	23	Somewhat Agree	22	Somewhat Agree
C14	5.3264	2.5922	Somewhat Agree	16	Agree	15	Agree
C15	5.3161	2.5860	Somewhat Agree	9	Agree	7	Agree
C16	4.8083	2.3300	Somewhat Agree	2	Agree	2	Agree
C17	5.2280	2.5349	Somewhat Agree	6	Agree	6	Agree
C18	5.5544	2.7360	Agree	24	Agree	24	Agree
C19	5.4560	2.6720	Somewhat Agree	19	Agree	19	Agree
C20	5.2332	2.5379	Somewhat Agree	15	Somewhat Agree	10	Somewhat Agree
C21	5.2694	2.5587	Somewhat Agree	10	Agree	9	Agree
C22	4.7617	2.3116	Somewhat Agree	4	Somewhat Agree	4	Somewhat Agree
C23	4.6891	2.2849	Somewhat Agree	3	Somewhat Agree	3	Somewhat Agree
C24	5.5389	2.7257	Agree	18	Agree	18	Agree
C25	5.4197	2.6492	Somewhat Agree	14	Agree	17	Agree
C26	5.5699	2.7463	Agree	7	Agree	12	Strongly Agree
C27	2.6425	2.6109	Somewhat Disagree	8	Disagree	8	Disagree

The paper has some limitations also. In the paper, multisensory attributes are evaluated for the determination of satisfaction levels. Service quality is another attribute that also affects customer satisfaction. Next, the questionnaires were collected at different times of the day. Therefore, different rates were given by each respondent on the same day.

5 Conclusions

In the paper, the evaluation of customer satisfaction is considered using multisensory attributes, such as vision, hearing, touching, smelling and tasting. To achieve the study aim, the authors proposed the integration of type-2 fuzzy sets and conjoint analysis to investigate the impact of sensory marketing on customer satisfaction. For this purpose, an online questionnaire was used to collect statistical data from supermarket shoppers. Based on the statistical data of the multisensory attributes, a type-2 fuzzy conjoint analysis is proposed to determine the satisfaction degrees of the respondents. Afterwards, using type-2 fuzzy sets, the similarity degrees between customer opinions and opinions of experts are determined. For this purpose, Mendel-Wu and Zeng-Li similarity measures are employed for each attribute. Using type-2 fuzzy similarity degrees, the ranking method is applied for the evaluation of customer satisfaction. The experimental results were done for 27 sensory marketing questions to understand customer opinions in supermarkets. The findings reflect significantly positive customer satisfaction. Future research includes analysing the effect of service quality on customer satisfaction.

Footnotes

Appendix

The Questionnaire used for multisensory evaluation

N	Questions	Facets
Item 1	I feel the store inside is bright	Sight
Item 2	I feel the store inside is colourful	Sight
Item 3	I feel the store inside is interesting	Sight
Item 4	I feel the store inside is organized	Sight
Item 5	I feel the store inside is comfortable	Sight
Item 6	I feel the store inside is attractive	Sight
Item 7	I can touch the products	Touch
Item 8	I feel more comfortable purchasing a product after physically examining it	Touch
Item 9	It is important for me to touch all kinds of products	Touch
Item 10	I am afraid to buy the product because I can not touch it before I buy it	Touch
Item 11	I feel more comfortable buying a product after touching it	Touch
Item 12	There are other products I would purchase only if I could handle them before purchasing	Touch
Item 13	The test products they offer taste good	Taste
Item 14	i like their products taste	Taste
Item 15	I feel comfortable to taste their products before purchasing it	Taste
Item 16	I am afraid to buy the product because I can not taste it before I buy it	Taste
Item 17	There are other products I would purchase only if I could taste them before purchasing	Taste
Item 18	The smell of store is fresh	Smell
Item 19	the store has a pleasant scent (Smell)	Smell
Item 20	I like the fragrance they use in this store	Smell
Item 21	I can smell their fresh products	Smell
Item 22	I like the music playing in the store	Sound
Item 23	Music that plays in-store is important to me	Sound
Item 24	Pleasant music creates a favourable atmosphere	Sound
Item 25	Music is an important factor that influences my shopping experience in-store	Sound
Item 26	Loud music in-store annoying me	Sound
Item 27	I like loud music in-store as it creates a pleasant in-store experience	Sound

References

Hulten

, Sensory Marketing: an introduction, London, UK, 2020, Sage Publication.

Satti

Z.W.

, Babar

S.F.

and Ahmad

H.M.

, Exploring mediating role of service quality in the association between sensory marketing and customer satisfaction, Total Quality Management 2019. https://doi.org/10.1080/14783363.2019.1632185

Darestani

A.Y.

and Jahromi

A.E.

, Measuring customer satisfaction using fuzzy inference system, Journal of Applied Sciences 9(3) (2009), 469–478.

Zadeh

L.A.

, Fuzzy sets, Information and Control 8 (1965), 338–353.

Abiyev

, Saner

, Eyupoglu

and Sadikoglu

, Measurement of job satisfaction using fuzzy sets, Procedia Computer Science 2016. doi:102.294-301.10.1016/j.procs.2016.09.404

Kang

, Porter

C.S.

and Bohemia

, Using the fuzzy weighted association rule mining approach to develop a customer satisfaction product form, Journal of Intelligent & Fuzzy Systems 38 (2020), 4343–4357. DOI: 10.3233/JIFS-190957 IOS Press.

Fasanghari

and Roudsari

F.H.

, The fuzzy evaluation of e-commerce customer satisfaction, World Applied Sciences Journal 4(2) (2008), 164–168.

Tsenov

, Ivanov

, Toparova

, Neykov

, Ivanova

and Gadjeva

, Fuzzy evaluation of customer satisfaction with mobile services, 2011 10th International Conference on Telecommunication in Modern Satellite Cable and Broadcasting Services (TELSIKS) (2011), 665–668, doi: 10.1109/TELSKS.2011.6143199.

Reza Nasserzadeh

S.M.

, Hamed Jafarzadeh

, Mansouri

and Sohrabi

, Customer satisfaction fuzzy cognitive map in banking industry, Communications of the IBIMA 2 (2008).

10.

Ammar

, Moore

and Wright

, Analysing customer satisfaction surveys using a fuzzy rule-based decision support system: Enhancing customer relationship management, Database Marketing & Customer Strategy Management 15(2) (2008), 91–105. Palgrave Macmillan

11.

Nazari-Shirkouhi

and Keramat

, Modeling customer satisfaction with new product design using a flexible fuzzy regression-data envelopment analysis algorithm, Applied Mathematical Modelling 50 (2017), 755–771.

12.

Miranda

, Tavares

and Queir

, Perceived service quality and customer satisfaction: A fuzzy set QCA approach in the railway sector, Journal of Business Research 89 (2018), 371–377.

13.

Hsiao

Y.-H.

, Chen

L.-F.

, Chang

C.-C.

and Chiu

F.-H.

, Configurational path to customer satisfaction and stickiness for a restaurant chain using fuzzy set qualitative comparative analysis, Journal of Business Research 69 (2016), 2939–2949.

14.

Liu

, Zeng

, Xu

and Koehl

, A fuzzy model of customer satisfaction index in e-commerce, Mathematics and Computers in Simulation 77 (2008), 512–521.

15.

Hsiao

Y.-H.

, Chen

L.-F.

and Chang

C.-C.

, Configurational path to customer satisfaction and stickiness for a restaurant chain using fuzzy set qualitative comparative analysis, Journal of Business Research 69(8) (2016), 2939–2949.

16.

Tahsiri

, Niakousari

, Khoshnoudi-Nia

and Hosseini

S.M.H.

, Sensory evaluation of selected formulated milk barberry drinks using the fuzzy approach, Food Sci Nutr. 5 (2017), 739–749.

17.

Lim

S.Y.

, Harun

U.B.

, Gobil

A.R.

, Mustafa

N.A.

, Zahid

N.A.

, Amin-Nordini

, Ariffin Md. , Kalsom

, Tengku Zetty Maztura Tengku Jamaluddin , Nurul Azmawati Mohamed , Nor Afiah Mohd

and Shohaimi

, Measuring customer satisfaction on the cleanliness of food premises using fuzzy conjoint analysis: A pilot test, Plos One 16(9) (2021), e0256896. https://doi.org/10.1371/journal.pone.0256896

18.

Luce

R.D.

and Tukey

J.W.

, Simultaneous conjoint measurement: a new type of fundamental measurement, Journal of Mathematical Psychology 1 (1964), 1–27.

19.

Green

P.E.

and Rao

V.R.

, Conjoint measurement- for quantifying judgmental data, Journal of Marketing Research (1971).

20.

Turksen

I.B.

and Wilson

I.A.

, A fuzzy preference model for consumer choice, Fuzzy Sets and Systems 68(3) (1994), 253–266.

21.

Abiyev

, Sadikoglu

and Abiyeva

, Fuzzy evaluation of job satisfaction of hotel employees, Int’l Conf. Artificial Intelligence | ICAI’15 (2015), 175–180.

22.

Vilela

, Oluyemi

and Petrovski

, A fuzzy inference system applied to value of information assessment for oil and gas industry, Decision Making: Applications in Management and Engineering 2(2), 1–18.

23.

BoŽanić

, Tešić

, Marinkovic

and Milić

, Modeling of neuro-fuzzy system as a support in decision-making processes, Reports in Mechanical Engineering 2(1), 222–234. https://doi.org/10.31181/rme2001021222b

24.

Precup

R.-E.

, Preitl

, Petriu

, Bojan-Dragos

C.-A.

, Szedlak-Stinean

A.-I.

, Roman

R.-C.

and Hedrea

E.-L.

, Model-based fuzzy control results for networked control systems, Reports in Mechanical Engineering 1(1), 10–25. https://doi.org/10.31181/rme200101010p;

25.

Kuzmanović

and Vukić

, Incorporating heterogeneity of travellers’ preferences into the overall hostel performance rating, Decision Making: Applications in Management and Engineering 4(2), 200–224.

26.

Mendel

J.M.

, Uncertain Rule-Based Fuzzy Logic System: Introduction and NewDirections, Prentice Hall, Upper Saddle River, NJ, 2001.

27.

Zadeh

L.A.

, The concept of linguistic variable and its application to approximate reasoning, Information Sciences 8 (1975).

28.

Mendel

J.M.

, John

R.I.

and Liu

, Interval type-2 fuzzy logic systems made simple, IEEE Trans. Fuzzy Syst. 14(6) (2006), 808–821.

29.

Abdullah

, Adawiyah

C.W.R.

and Kamal

C.W.

, A decision making method based on interval type-2 fuzzy sets: An approach for ambulance location preference, Applied Computing and Informatics 14(1) (2018), 65–72.

30.

Runkler

, Coupland

and John

, Interval type-2 fuzzy decision making, International Journal of Approximate Reasoning 80 (2017), 217–224.

31.

Castillo

and Melin

, Intelligent systems with interval type-2 fuzzy logic, Int. Journal of Innovative Computing, Information and Control 4(2) (2008), 771–783.

32.

Abiyev

R.H.

and Kaynak

, Type-2 fuzzy neural structure for identification and control of time-varying plants, IEEE Transactions on Industrial Electronics 57(12) (2010), 4147–4159.

33.

Lin

C.T.

, Pal

N.R.

, Wu

S.L.

, Liu

Y.T.

and Lin

Y.Y.

, An interval type-2 neural fuzzy system for online system identification and feature elimination, IEEE Trans. on Neural Networks and Learning Systems 26(7) (2015), 1442–1455. https://doi.org/10.1109/TNNLS.2014.2346537

34.

Abiyev

R.H.

, Kaynak

, Alshanableh

and Mamedov

, A type-2 neuro-fuzzy system based on clustering and gradient techniques applied to system identification and channel equalization, Applied Soft Computing 11(1) (2011), 1396–1406. https://doi.org/10.1016/j.asoc.2010.04.011

35.

Baklouti

, Abraham

and Alimi

A.M.

, A beta basis function interval type-2 fuzzy neural network for time series applications, Eng. Appl. of AI 71 (2018), 259–274. https://doi.org/10.1016/j.engappai.2018.03.006

36.

Abiyev

and Abizada

, Type-2 fuzzy wavelet neural network for estimation energy performance of residential buildings, Soft Computing 25(16) (2021), 11175–11190. https://doi.org/10.1007/s00500-021-05873-4;

37.

and Mendel

J.M.

, A comparative study of ranking methods, similarity measures and uncertainty measures for interval type-2 fuzzy sets, Information Sciences 179(8) (2009), 1169–1192.

38.

Zeng

and Li

, Relationship between similarity measure and entropy of interval valued fuzzy sets, Fuzzy Sets and Systems 157 (2006), 1477–1484.

39.

İç

Y.T.

and Yurdakul

, Development of a quick credibility scoring decision support system using fuzzy topsis, Expert Systems with Applications 37(1) (2010).

40.

Karagul

A.A.

and Özdemir

, Evaluation of financial information quality attributes: a comparison from Turkey, International Journal of Business and Social Science 3(23) (2012).

41.

Srinivasa

R.K.

and Pillai

C.R.S.

, Multicriterion decision making in river basin planning and development, European Journal of Operational Research 112(2) (1999).