1. INTRODUCTION
Today, with the explosion of Internet, an evergrowing number of consumers are shopping online. So, ecommerce has become an essential part of business strategy in the rising of global economy. The model of ecommerce brings many benefits for not only business owners but also consumers, who plays an important role in the survival and growing of product, business and supply chain (Saito and Kusukawa, 2018). Firstly, from ecommerce, a new brand can sell to customers around the world easily. Selling worldwide is a great feat as it helps business build its brand a lot faster and broadens marketplace. Secondly, the business owner can decrease the cost of seting up, marketing and managing inventory through webbased management system. This is a significant benefit that ecommerce brings, compared to traditional model. In addition, ecommerce retailers can easily keep a constant eye on consumers’ buying habits and interests to tailors their offer suit to consumers’ requirements, so, business can build long term relationships with their customers, so that increasing customers’ satisfaction and loyatly. From the consumers’ view, ecommerce allows to shop anywhere and anytime, that save cost, time and will be very suitable for busy consumers in the global market.
Recognizing the benefits of ecommerce, the number of ecommerce web sites are greatly growing recently. With the significant grown of online shopping websites, the competition among thems also extremely increases (Zhang et al., 2008). In order to be successful in the ecommerce marketplace, organizations will need provide high quality web sites that attract and retain customers (Içtenbaş and Rouyendegh, 2012). It is a need to understand and evaluate how effective their ecommerce websites are to achieve high performance (Fei and Yu, 2008). Hence, evaluating competitive advantages of ecommerce becomes a considerable concern for business owners.
There are many methods have been proposed over the years to evaluate web sites. Some studies established evaluation index system to evaluate the influence of website quality factors on success of ecommerce (Barnes and Vidgen, 2001;Agarwal and Venkatesh, 2002;Devaraj et al., 2002). However, evaluation of ecommerce websites can be seen as multiple criteria decision making (MCDM) problem, so it need the integrated analysis of both qualitative and quantitative factors (Liu et al., 2007;Liang et al., 2017). In the past few years, an Analytic Hierarchy Process (AHP) approach has been used to weight the indexes (BanaiKashani, 1989;Kwong and Bai, 2002;Ngai, 2003;Chen, 2006;Lee and Kozar, 2006;Liu et al., 2007;Fei and Yu, 2008;Içtenbaş and Rouyendegh, 2012). The best advantages of AHP is to make comparision among criteria as well as alternatives through pairwise comparision matrixes. Other authors suggested using Technique for Order Preference by Similarity to Ideal Situation (TOPSIS) method to evaluate competitive advantage of ecommerce (Zhang et al., 2008;Sun and Lin, 2009;Alptekin et al., 2015). TOPSIS allows to compare the distance of each alternative to the positive and negative ideal solusion belong to the criteria that helps increasing the accuracy of evaluation process. On the otherhand, in some situations, when decision makers are asked to express their opinions, it is difficult to solve the problem of ambuguity and uncertainty through traditional methods, so those problems can be converted into fuzzy number, which was proposed by Zadeh (1965). Gang (2010) suceessfuly using the Fuzzy Synthesize for ranking shopping website that can be considered as an important contribution for the ecommerce evaluation. To combine the advantages of above mentioned methods, many reseachers have suggested integrated approach based on combination of FuzzyAHP and FuzzyTOPSIS in different fields and got considerable success (Büyüközkan et al., 2007;Muralidhar et al., 2012;Wijayanti et al., 2018).
Following the success of these studies and combining considerable advantages of such methods, this research aims to develop a new approach for ecommerce evaluation based on the combination of Fuzzy Analytic Hierarchy Process (FAHP) and FuzzyTechnique for Order Preference by Similarity to Ideal Situation (FTOPSIS). In which, FAHP is used to calculate criteria weight through pairwise comparision matrix. FTOPSIS is then applied to rank the ecommerces based on their distances to ideal soltions.
The remaining parts of this paper are arranged as follows. The literature review of competitive advantage in ecommerce and related techniques is described in Section 2. An integrated FuzzyAHPTOPSIS framework is presented in Section 3 while Section 4 validates the proposed approach through a case study in Vietnam. Section 5 includes the concluding remarks, suggestions, and limitations of the study.
2. LITERATURE REVIEW
2.1 Competitive Advantage in Ecommerce
According to Porter (1985), competitive advantage is the extent to which a company is able to gain and retain a dominant position over its competitors through creating value for its customers. It comprises capabilities which are the attributes of an organization that allows a company to differentiate itself from its competitors to achieve high levels of customer satisfaction and market performance (Tracey et al., 1999;Li et al., 2006). A company is said to have the competitive advantage when it is implementing a value creating strategy not simultaneously being implemented by any current or potential competitors (Barney, 1991). For example, a company can achieve a cost advantage when the company operates at a lower cost than its competitors but offers a comparable product. In ebusiness, to obtain competitive advantages, business owners need to identify customers’ value, then transform and develop them to be a series of factors, from which customers are willing to spend money on shopping and experience (Kalakota and Robinson, 2001). In other words, business owners need to determine the level of service quality they provide against customers’ expectations.
According to Parasuraman et al. (1985), service quality (SERVQUAL) is measured through ten factors, such as accessibility, communication, capability, courtesy, trustworthiness, reliability, responsiveness, safety, tangibility, and understanding with customers. They are then also reduced into five factors, including tangibility, reliability, responsiveness, assurance, and empathy (Parasuraman et al., 1988). Service quality measures have been applied to evaluate quality of ecommerce. Although customer is the most important factor that affect the survival and growing of a service, consumers’ perceptions of online service quality remain unexplored in these studies. Keeney (1999) developed a model to indicate that customer’s interaction with the eservice is directly related to the service performance. Following the success of this study, Devaraj et al. (2002) used integrated frameworks including technology acceptance model, transaction cost analysis, and SERVQUAL to measure consumer satisfaction in the ecommerce. The study found that technology acceptance model components – perceived ease of use and usefulness – are important in forming consumer attitudes and in strengthening the ecommerce channel. This study found empirical support for the assurance dimension of SERVQUAL as a determinant in ecommerce channel satisfaction. Yang et al. (2009) used four dimensions of SERVQUAL, which include reliability, responsiveness, assurance, and empathy, to measure the users’ cognition of SERVQUAL in online channel. Sun and Lin (2009) mentioned three main groups of criteria in evaluating ecommerce, including technology efficiency (level of ease of use, saving time and effectiveness when customers operate on the website), service quality (communication, confidence level, confidentiality of transactions, reliability of quality, price, delivery), and service evaluation (customer reviews, reputation). Içtenbaş and Rouyendegh (2012) presented four criteria used to evaluate ecommerce as system quality, information quality, service quality and attractiveness.
According to above discussion, this study suggests measuring website quality through four criteria of competitive advantage such as website form, product and reliability, delivery and customer service. Design of website is related to the structure, images, information, user navigation on the website, website speed or simply everything that users see and interact on the website (Keeney, 1999;Devaraj et al., 2002;Sun and Lin, 2009;Içtenbaş and Rouyendegh, 2012). Product and reliability related to price, quality and product diversity, product comments (Parasuraman et al., 1985, 1988;Keeney, 1999;Yang et al., 2009;Sun and Lin, 2009;Içtenbaş and Rouyendegh, 2012). The ability of delivery is related to the form of payment, time and delivery costs (Sun and Lin, 2009). Customer service involves interactions and customer relationship management such as understanding customer preferences, communication, policies, and availability (Parasuraman et al., 1985, 1988; Keeney, 1999;Yang et al., 2009;Sun and Lin, 2009;Içtenbaş and Rouyendegh, 2012).
2.2 Fuzzy Analytic Hierarchy Process (FAHP)
The analytic hierarchy process (AHP) is a quantitative technique proposed by Saaty (1980) for analyzing complex decisions and evaluating alternatives based on their relative performance with respect to suggested criteria (Boroushaki and Malczewski, 2008;Lin et al., 2007). The AHP resolves complex decisions by structuring the problem into a hierarchical framework with the objective at the top, follows by criteria and decision alternatives, respectively. The AHP method prioritizes objects at each level of the hierarchy using pairwise comparisons matrix. This process generally involves six steps (Vahidnia et al., 2009):

1) Defining the unstructured problem, including its objectives, criteria and alternatives;

2) Organizing the problem as a hierarchy (detailed criteria and alternatives);

3) Constructing a pairwise comparison matrix of the relevant contribution or impact of each element on each governing criterion in the next higher level.

4) Using the eigenvalue method (or some other method) to estimate the relative weights of the decision elements;

5) Calculating the consistency properties of the matrices to ensure that the judgments of decisionmakers are consistent;

6) Aggregating the weighted decision elements to obtain an overall rating for the alternatives.
Kwong and Bai (2002) used FAHP method to find out the main factors that determine the success of an ecommerce website. One of the key steps of FAHP is to establish a comparison matrix. During the implementation, a comparison matrix will be given for each main criteria as well as secondary criteria. The relevant priority weights of each primary and secondary criterion will then be calculated. The study provides an important contribution when formulating an objective integrated approach of FAHP to determine the weights of each attribute in an uncertain decisionmaking environment. Lee and Kozar (2006) also investigated website quality factors, their relative importance in selecting the most preferred website, and the relationship between website preference and financial performance. A field study with 156 online customers and 34 managers/designers of ebusiness companies was performed. The study identified different relative importance of each website quality factor and priority of alternative websites across ebusiness domains and between stakeholders. The study also found that the website with the highest quality produced the highest business performance.
Relevant to the ecommerce evaluation, Ngai (2003) presented an application of the AHP in selecting the best website for online advertising. In this study, five criteria were used for evaluating sites, including impression rate, monthly cost, audience fit, content quality, and “look and feel”. Liu et al. (2007) proposed using FAHP approach to construct an evaluation structure for website selection through a case study. Following the success of this study, Içtenbaş and Rouyendegh (2012) presented FAHP method to evaluate three ecommerce web sites in Turkey. Four criteria were selected as system quality, information quality, service quality and attractiveness which were the most suitable for evaluating the webs sites preferred by woman. Wijayanti et al. (2018) suggested using FAHP to weight seven criteria of ESERVQUAL, namely, efficiency, fulfillment, system availability, privacy, responsiveness, compensation, and contacts. The findings indicate that the AHP approach is a useful tool to help support a decision in convention website selection and evaluation.
2.3 Fuzzy Technique for Order Preference by Similarity to Ideal Situation (FTOPSIS)
TOPSIS was firstly presented by Hwang and Yoon and relates to the definition of positive ideal solution and negative ideal solution (Hwang and Yoon, 1981). The positive ideal solution is defined as the solution, in which the benefit criteria are maximized and the cost criteria are minimized, whilst in the negative ideal solution, the benefit criteria are minimized and the cost criteria are maximized. Hence, the best solution is the solution being nearest from the positive ideal solution and farthest from the negative ideal solution.
Formally, the weights of criteria and the ratings of alternatives are measured by crisp values (Behzadian et al., 2012). However, by using crisp values, decision makers are sometime confused with an exact number. Also, in uncertainty environment, using crisp values may not provide reliable solutions. So, the use of fuzzy logic has been introduced by Zadeh (1965) to incorporate the uncertainty, vagueness, and imprecision, resulting in a fuzzy TOPSIS problem (Chen, 2000). Then, a hierarchical TOPSIS in fuzzy environment was suggested for not only measuring uncertainty, but also providing useful steps for weighting criteria accurately with additional objectives (Wang et al., 2009). Also, by proposing Max and Min operations on fuzzy numbers, TOPSIS was also successfully implemented in a fuzzy environment with multicriteria problem (Wang and Lee, 2007).
FTOPSIS is continouslly applied in many fields. In ecommerce, Zhang et al. (2008) proposed FTOPSIS method which can evaluate the ecommerce website more accurate, providing decision support to the site decision makers, users and relevant organization. Sun and Lin (2009) used the FTOPSIS method to determine the weights of each criteria and rank the options in evaluating competitive advantage of ecommerce. The results indicated that security function and transaction reliability are the two most important factors, creating a competitive advantage for online sales websites. Out of the six websites that were proposed, PCHome and Yahoo Taiwan were the two most highly rated websites. Alptekin et al. (2015) applied FTOPSIS method for the evaluation of five Turkish bookstores websites quality. Fifteen subcriteria under four main categories (Service Quality, System Quality, Information Quality, VendorSpecific Quality) were used in the evaluation process.
Kang et al. (2016) presented a FTOPSIS based on the extended version of SERVQUAL (ESERVQUAL). This approach effectively considers the raised issues and preserves the core concept of ESERVQUAL for measurement of electronic service quality in the ecommerce environment. Following the success of using ESERVQUAL for evaluation of above study, Wijayanti et al. (2018) evaluated ecommerce website using seven criteria of ESERVQUAL, namely, efficiency, fulfillment, system availability, privacy, responsiveness, compensation, and contacts. In the study, TOPSIS is used to identify the ranking of all alternatives to be considered.
3. PROPOSED METHODOLOGY
The purpose of this study is to propose a methodology to evaluate online website using FAHP and FTOPSIS. The process includes 9 steps as following:
Step 1: Bulding the unstructured problem
Assum that there are n criteria and m alternatives. The unstructured problem has three levels as in Figure 1.
Step 2: Building pairwise comparisons among decision elements to form comparison matrices
Conversion scale are presented by linguistic variables. The rate of criteria comparision are presented by linguistic variables. Positive triangular fuzzy numbers suggested by Sodhi and Prabhakar (2012) are used to express these linguistic variables (see Table 1).
Assume that the survey is conducted by the group of k members (k decision makers). The relationship score between criterion j and criterion l assigned by k decision makers, called average score ${\tilde{as}}_{jl}$, is defined as follows:
where ${\tilde{as}}_{jl}$ is average score of relationship between criterion j and l from k decision makers
The pairwise comparisons matrix is then built based on the average score of relationships among criteria:
Step 3: Calculating fuzzy weight for each criterion
Using geometric mean method to determine Fuzzy geometric mean and Fuzzy weight for each criterion.
where ${\tilde{g}}_{j}$ is Fuzzy geometric mean of criterion j
${\tilde{w}}_{j}=({w}_{1j},\hspace{0.17em}{w}_{2j},\hspace{0.17em}{w}_{3j})$, is Fuzzy weight of criterion jj = 1...n
Step 4: Building Fuzzy Decision Matrix
Assume that the alternative rating ${\tilde{x}}_{ij}^{k}$ (with i_{th} alternative and j_{th} criterion) of the k_{th} decision maker are define as ${\tilde{x}}_{ij}=({a}_{ij},\hspace{0.17em}{b}_{ij},\hspace{0.17em}{c}_{ij})$ (Sodhi and Prabhakar, 2012). The definition of linguistic variables are shown in Table 1.
The aggregated fuzzy rating ${\tilde{x}}_{ij}$ of alternative (i) according to each criterion (j) are obtained by following equations:
where, ${a}_{ij}={\mathrm{min}}_{k}\left\{{a}_{ij}^{k}\right\},\hspace{0.17em}{b}_{ij}=\text{}\frac{1}{k}{\displaystyle {\sum}_{k=1}^{K}{b}_{ij}^{k}},\hspace{0.17em}{c}_{ij}={\text{max}}_{k}\left\{{c}_{ij}^{k}\right\}$
The fuzzy decision matrix $\tilde{D}$, which shows the relationships between alternatives and criteria, and the criteria weighting set $\tilde{W}$ are obtained as follows:
where $\begin{array}{l}i\hspace{0.17em}=\hspace{0.17em}1,\text{}2,\text{}\dots ,\text{}m\hspace{0.17em}\text{and}\hspace{0.17em}j=1,\text{}\hspace{0.17em}2,\text{}\dots ,\hspace{0.17em}\hspace{0.17em}n\\ {\tilde{x}}_{ij}=({a}_{ij},\hspace{0.17em}\hspace{0.17em}{b}_{ij},\hspace{0.17em}\hspace{0.17em}{c}_{ij})\end{array}$
Step 5: Building Normalized Decision Matrix
To make the fuzzy decision matrix becomes simple, it should be normalized by using linear scale transformation. Hence, the normalized fuzzy decision matrix is obtained as:
where
Step 6: Building the Weighted Normalized Decision Matrix
By combining the weights ${\tilde{w}}_{j}$ and the normalized fuzzy decision matrix ${\tilde{r}}_{ij}$, the weighted normalized fuzzy decision matrix $\tilde{V}$ is obtained:
where ${\tilde{v}}_{ij}={\tilde{r}}_{ij}\text{(}.\text{)}{\tilde{w}}_{j}$
Step 7: Determining the Fuzzy Positive Ideal Solutions (FPIS) and Fuzzy Negative Ideal Solutions (FNIS)
The FPIS ($({A}^{\text{*}})$ and FNIS $({A}^{})$ of the alternatives are defined as:
where ${\tilde{v}}_{j}^{\text{*}}={\mathrm{max}}_{i}\left\{{v}_{ij3}\right\},\hspace{0.17em}\hspace{0.17em}i=1,\hspace{0.17em}2,\hspace{0.17em}\mathrm{...},\hspace{0.17em}m,\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}j=1,\hspace{0.17em}2,\hspace{0.17em}\mathrm{...},\hspace{0.17em}n$
where ${\tilde{v}}_{\overline{j}}={\mathrm{min}}_{i}\left\{{v}_{ij1}\right\},\hspace{0.17em}\hspace{0.17em}i=1,\hspace{0.17em}2,\hspace{0.17em}\mathrm{...},\hspace{0.17em}m,\hspace{0.17em}\hspace{0.17em}j=1,\hspace{0.17em}2,\hspace{0.17em}\mathrm{...},\hspace{0.17em}n$
Step 8: Calculating the Distance from each Alternative to FPIS and FNIS
The distances from each alternative to the FPIS $({d}_{i}^{*})$ and the FNIS $({d}_{i}^{})$ are calculated as:
where
Assume that ${d}_{v}(\tilde{x},\hspace{0.17em}\hspace{0.17em}\tilde{y})$ is the distance between two fuzzy number $\tilde{x}$ and $\tilde{y}$, it is obtained by following equation:
Step 9: Calculating the Closeness Coefficient and Ranking Alternatives
The closeness coefficient CC_{i} is defined to compare the distances from each alternative to fuzzy positive ideal solution ${A}^{*}$ and fuzzy negative ideal solution ${A}^{}$. The closeness coefficient of each alternative is calculated by following formulation:
The formulation mentioned above presents obviously the definition of closeness coefficient. The higher closeness coefficient, the closer distance from alternative to fuzzy positive ideal solution.
4. ILLUSTRATIVE EXAMPLE
In this study, a case study in Vietnam is used to illustrate the proposed methodology. Surveys were conducted around Can Tho City to evaluate criteria weight and rank alternatives. In the first survey, questionaires were sent to 5 experts to idenfify criteria used to evaluate ecommerces website. In the second survey, questionaires were sent to 468 consumers, inorder to evaluate criteria and four websites. The methodology is applied step by step as following:
Step 1: Bulding the unstructured problem
Based on previous research, this study uses 18 criteria for evaluating four ebusiness in Vietnam, including A1, A2, A3 and A4. Criteria are presented in Table 2.
Step 2: Building pairwise comparisons among decision elements to form comparison matrices
Pairwise comparisons among criteria are built using equation (1). The results are shown in Table 3Table 6.
Step 3: Calculating fuzzy weight for each criterion
Using equation (3) and (4) to calculate Fuzzy geometric mean $({\tilde{g}}_{j})$, and Fuzzy weight $({\tilde{w}}_{j})$ of each criterion. The results are shown in Table 7.
Step 46: Building decision matrix, normalized decision matrix and weighted normalized decision matrix
Decision matrix, normalized decision matrix and weighted normalized decision matrix are built based on equations (5) –(9). The result are shown in Table 8, 9 and 10.
Step 7: Calculate Positive ideal solution (A+) and negative ideal solution (A)
Using equations (10) and (11) to calculate Positive ideal solution (A+) and negative ideal solution (A) of each criterion. The results are shown in Table 11.
Step 8: Calculating the Distance from each Alternative to FPIS and FNIS
The distance from each alternative to FPIS and FNIS are calculated using equation (12)(14). The ressults are shown inTable 12.
Step 9: Calculating the Closeness Coefficient and Ranking Alternatives
Using equation (15) to calculate the Closeness Coefficient of each alternative. The results are shown in Table 13.
The results from Table 13 show that the closeness coefficient of alternative A1 is the highest, indicating that alternative A1 is closest to the positive ideal solution. Therefore, option A1 has the highest competitive advantage in the ecommerce market, followed by A3, A4 and A2, respectively. The results provide valuable information to help consumers consider and make appropriate purchase choices. Although the results of evaluating and ranking websites are based on the integration of many criteria and evaluation of many customers, but based on the decision matrix and standardized decision matrix, it is possible to relize that site speed, product quality, delivery speed, reliability, and return policy are five factors that are highly appreciated by customers (much better than A3), giving competitive advantages for A1. Based on the evaluation results, the websites can develop appropriate competitive strategies.
Indeed, businesses can not be successful if they can not build customers trust. Inwhich, quality is one of critical factors that help satisfy customers and build customers loyalty. Quality is defined as fitness for use, including product performance, reliability, and durability and a key element of valuetocustomer (Tracey et al., 1999). With high quality, their product would be increasingly recognized from customers, which in turn, leads to improved performance in terms of sales growth and market share (Forker et al., 1996;Ward and Duray, 2000;Özdemir and Aslan, 2011;Hatani et al., 2013;Laosirihongthong et al., 2013). These findings support that quality is not only an important element in operation but also the key for company success.
Beside quality, one of the most important competitive advantages that increase the customers’ experience is delivery. There are a variety of delivery aspects that create a positive experience for customers, such as speed, accuracy, and reliability (Noble, 1997). In a competitive environment as today, if the service can not satisfy customers, they will search for competitors. Hence, many companies seek to maintain or increase their customer base by focusing on the competitive priorities of development speed, fast and reliable delivery (Li, 2000). Related to the speed, in addition, a website with a long load time tend to inhibit the users. In today’s busy life, everyone does not like to wait and purchase on a long waiting time website. Therefore, optimizing speed is always a top priority and must be continually improved (Sun and Lin, 2009).
On the other hand, online customers do not get to see and hold the physical product before they buy it, so they are worried about problems such as the expected aviation, or no exchange policy, or pay additional costs to return as expected. While the global competitive trend is growing, more and more customers require that the free return and exchange service will be a standard service. Ecommerce sites must ensure that their return policies are fair and attract to their customers and turn these policies into something different for their brands (Yang et al., 2009;Sun and Lin, 2009).
5. CONCLUSION
This study was conducted to develop an integrated of FuzzyAHPTOPSIS to evaluate ecommerce. FAHP was used to calculate criteria weight based on pairwise decision matrix. FTOPSIS is then applied to evaluate four websites. The proposed methodology allows assessing the object on many different criteria, while limiting the subjectivity of the assessors. It provides a comprehensive picture to help consumers consider and make suitable purchasing decisions. Ranking results show that A1 has the highest competitive advantage, followed by A3, A4 and A2, respectively. Among different criteria, site speed, product quality, delivery speed, reliability, and return policy are five factors that are highly appreciated by customers, giving competitive advantages for A1. From these findings, four mentioned websites can develop appropriate competitive strategies to effectively improve their performance and perfectly satisfy their consumers in the future.
Besides the significant contributions, the research also has some limitations. The survey was only conducted on those who are currently studying, working and living around Can Tho City, so it is a local result. In addition, the study has not combined other qualitative analysis methods to build a more comprehensive picture to enhance the reliability of the results. This will be the premise to carry out the next research in website evaluation. Further studies need to expand the scope of the survey to increase the accuracy of the assessment as well as consolidate the results across the board.