Reliability and validity However rigorous the above NGT procedure was, some participants might change
their attributes classification due to conformity after hearing the explanations of other participants on their SIVA-Need attributes classification matrix. The possible lack of confidence out of unlikeness might led to changes of grades when they were asked to provide final evaluations at stage 5. In order to avoid the collective bias caused ty conformity, which would affect the reliability and validity of SIVA-Need attributes classification matrix (Table 3), this study conducted content analysis on SIVA-Need attributes classification matrix. The reliability was derived from the gradings of 7 participants and 4 authors, which could help to verify the decision in NGT. The content analysis on SIVA-Need attributes classification matrix contained to steps:
Step 1 Grading the degree of consent on attributes.
Step 2 The resulting grades then went into the function. After calculations, reliability was generated.
Detailed descriptions of the above two steps are as follows.
Step 1 After NGT, the authors distributed an evaluation sheet to each participant. Participants were asked to give scores for the degree of consent on attributes in SIVA-Need attributes classification table; meanwhile, 4 authors also graded the attributes. Without discussions or interferences, participants and authors completed his or her own evaluation. After sorting the results, the content analysis result (Table 4) on SIVA-Need attributes classification matrix was then derived.
46 International Journal of Electronic Commerce Studies
Table 4. The evaluation results on SIVA-Need attributes classification matrix
SIVA dimensions
SIVA-Need Attributes
Grade distribution based on the evaluations of 7 expert participants (E) and 4 authors (R)
Strongly disagree disagree
Neither agree nor disagree
agree Strongly agree 50% and
more identical -2 -1 0 +1 +2
E R E R E R E R E R E R
customer solutions
Amenity 1 1 6 3 +2 +2 Commodity safety
2 1 5 3 +2 +2
sociability 1 1 2 2 4 1 +2 +1 Exclusiveness 1 2 1 5 2 +2 +2 Personal style 1 1 6 3 +2 +2
Customer information
Transparency 1 1 6 3 +2 +2 Information safety
1 7 3 +2 +2
Online consultation
1 1 2 4 3 +2 +2
Respect for diversity
1 1 1 5 3 +2 +2
Completeness 2 1 5 3 +2 +2
Customer value
Product benefits
7 4 +2 +2
Psychological benefits
7 4 +2 +2
Brand community value
7 4 +2 +2
Sense of honor
1 1 2 6 1 +2 +1
Personal value 1 1 6 3 +2 +2
Customer access
Efficiency 4 3 3 1 +1 +1 Consumption safety
1 1 6 3 +2 +2
Belongingness 1 7 3 +2 +2 Courteous reception
1 1 1 5 3 +2 +2
Autonomous consumption
1 2 1 4 3 +2 +2
Step 2 Based on Table 4, the reliability on SIVA-Need attributes classification matrix was calculated. The calculation method on content analysis was taken from Holsti [89]. The threshold of 50% and more similarity was applied on both the grading results of 7 participants (group 1 evaluators) and 4 authors (group 2 evaluators). The degree of agreement of the results that met the threshold was then calculated. Based on that, the reliability of SIVA-Need attributes classification matrix was derived. The calculation functions of the degree of agreement and reliability was as follows:
Function 1
Tsuen-Ho Hsu, Sen-Tien Her, Yung-Han Chang and Jia-Jeng Hou 47
A = 2 × X ÷ (Y1 + Y2) A = degree of mutual agreement X = number of identical grades of two groups of evaluators Y1 = the supposed number of number for group1 Y2 = the supposed number of agreements for group 2
Function 2 Reliability = n × A ÷〔1+(n-1) × A〕 n = number of evaluator groups
According to Table 4, there were 18 attributes that met the threshold of 50% and more similarity in group 1 and 2. They then went into the functions above, which generated the following results: Function 1: A = 2 × 18 ÷ (20+20) = 0.9 Function 2: Reliability = 2 × 0.9 ÷〔1+(2-1)× 0.9〕= 0.95
Based on the above two steps, the derived reliability of SIVA-Need attributes classification matrix was 0.95, which showed a high consistency among the grading results on the degree of consent on SIVA-Need attributes classification matrix (Table 3) evaluated by 7 participants and 4 authors. In respect of validity, all participants in NGT had a doctorate degree with the experiences of 9-20 years of teaching and worked as assistant, associate professors or professors in the university. Hence, this study inferred a high expert validity on SIVA-Need attributes classification matrix. After conceptualization analysis and NGT, the 20 attributes in the SIVA-Need attributes classification matrix were determined. Following the reliability and validity tests, this study verified that 20 attributes in the matrix had high reliability and validity. This study then incorporated the definitions on marketing strategies discussed in the literature review. That is, systematic analysis, resources allocation, action planning, and performance evaluation, the four core principles, conducted by the corporations by corporations to gain profits. After incorporating the marketing strategy definition into SIVA-Need attributes classification matrix, it was transformed into SIVA-Need model (Figure 2).
48 International Journal of Electronic Commerce Studies
Objectives Dimensions Attributes Strategic marketing evaluation
Figure 2. SIVA-Need Model
SIVA-Need
customer solutions
amenity
commodity safety
sociability
exclusiveness
personal style
customer information
transparency
information safety
online consultation
respect for diversity
completeness
customer value
product benefits
psychological benefits brand
community value
sense of honor
personal value
customer access
efficiency
consumption safety
belongingness
courteous reception
autonomous consumption
create satisfied repeat-
purchase customers
Weight analysis
resources allocation
action planning
performance evaluation
Tsuen-Ho Hsu, Sen-Tien Her, Yung-Han Chang and Jia-Jeng Hou 49
4. SIVA-NEED MODEL VALIDATION—AN EMPIRICAL STUDY ON APPLE COMPANY
“Turn the device you have into the one you want” was the open declaration in the Apple Trade In online (Figure 3). This shows that Apple company emphasizes on customers dominant marketing thinking, which corresponds with the customer-focused perspectives of SIVA marketing mix. This study thus chose Apple company for an empirical case study.
Figure 3. Apple Trade In Online
Source: Apple Store Website (2020. July. 30) https://www.apple.com/shop/trade-in
4.1 Fuzzy numbers and linguistic variables Zadeh [90] proposed the Fuzzy Sets Theory by arguing that there is a certain
degree of fuzziness existing in human beings’ subjective perceptions and understandings of things around us. An adoption of the fuzzy logic on human perceptions of things may compensate for the blind spots created by traditional binary logics (0 and 1) which swings between propositional answers of true or false. Fuzzy Sets are sets of objects without clear boundaries or definite characteristics. Membership Function describes the degree that a certain attribute in a set belongs to a sub-set, which ranges from 0 to 1.
According to Hsu and Lin [91, p.6], “All studies of marketing and consumer behavior contain qualitative variables. Wu and Chen [92] suggested that, based on social science meaurement principlas, most data or incidents are of fuzzy or uncertain nature. Given this, handling the data with hypothetical accuracy might lead to model misconstruction and increase the bias between research findings and reality. McCauley- Bell and Badiru [93] applauded Fuzzy Sets theory as an appropriate and efficient method, especially when the study pertains to explanations of risk attributes and evaluations of risk levels. Krcmar, et al. [94] found that the application of Fuzzy Sets theory had higher flexibility than statistical analysis when dealing with fuzzy and uncertain decisions. Chen [95] argued that given the qualitative elements of evaluation process and the subjectivities of decision-makers, accurate decision model was unfit to
50 International Journal of Electronic Commerce Studies
explain the actual decision-making scenarios. Instead, drawing on the linguistic variables to represent the subjective evaluations of decision makers, and incorporating the fuzzy evaluation values of multiple decision-makers, may generate better results. ”
4.2 Using Fuzzy Linguistic Preference Relations Analysis to generate the weight values of SIVA-Need attributes
The marketing thinking of SIVA marketing mix was customer-focused, which emphasized on customer needs. The was developed based on human needs. With the combination of the two, the SIVA-Need hierarchical model should embody, more completely, the customer-focused spirit with emphases on customer needs. Therefore, to obtain the weight values of each dimension and attributes of the model, it is only fit to collect data from customers as they can provide direct responses. This study developed a questionnaire based on the SIVA-Need model and targeted the heavy users2 of Apple products as survey participants. With a total 326 valid questionnaires collected (Appendix 1), this study used the following functions for subsequent analyses.
2Heavy users were defined as those who had purchased and used at least 3 Apple products with the user experience of minimum 6 years. As these users had accumulated a certain degree of understanding of Apple services based on their multiple years of experiences of Apple products, they were treated as the experts and representatives of customers of Apple company in this study.
This study followed the Fuzzy Linguistic Preference Relations proposed by Wang and Chen [96] and calculated the weights of SIVA-Need dimensions and attributes. By combining fuzzy analytic hierarchy process and fuzzy linguistic variables, the Consistent Fuzzy Linguistic Preference Relations method was applied to establish the Consistent Fuzzy Linguistic Preference Relations (CFLPR) matrix. This method could simplify the computation procedure and reduce the pairwise comparisons with only n- 1 questions required in the questionnaires. This method also improved the response inconsistencies among experts and increased the efficiency and accuracy of the overall decision-making process as it met actual needs better in operations with higher consistency [96].
To assess the subjective preferences of consumers during various decision-making process, linguistic variables were adopted to indicate the values of each dimension and attributes. Linguistic variables are linguistics terms by nature, which are often used in complex or ambiguous situations [97]. For example, phrases such as “equal importance”, “moderate importance”, “strong importance”, “fairly importance”, and “very strong importance” could be used to describe the degree or value of importance of items. By applying fuzzy linguistic variables, participants could express their preliminary opinions plainly and respond adequately to the vagueness of questions, which helps to increase the feasibility of analysis results [98]. Among the commonly- used fuzzy numbers—triangular, trapezoidal, and normal fuzzy number, triangular fuzzy number is most widely adopted [99-101]. The computation procedure of Consistent Fuzzy Linguistic Preference Relation is as follows.
1. Establishing fuzzy linguistic variables Based on the Triangular fuzzy importance scale (Figure 4) developed by Tolga and colleagues (Tolga, et al. [102]) and 9-point linguistic scale, a fuzzy linguistic preference relation matrix was established. The linguistic value set was defined as ???? =﹛equal importance, moderate importance, strong importance, very strong
Tsuen-Ho Hsu, Sen-Tien Her, Yung-Han Chang and Jia-Jeng Hou 51
importance, demonstrated importance﹜(K=1, 2, …, 5), which was then provided for the participants. Participants could assign values to each dimension and attributes in the SIVA-Need model, as listed in Table 5.
Figure 4. Triangular fuzzy importance scale
Source: Büyüközkan [103]
Table 5. Fuzzy number definitions Linguistic variables
Designation Triangular fuzzy number
Triangular fuzzy reciprocal scale
Demonstrated importance
DI (2, 5/2, 3) (1/3, 2/5, 1/2)
Very strong importance
VSI (3/2, 2, 5/2) (2/5, 1/2, 2/3)
Strong importance
SI (1, 3/2, 2) (1/2, 2/3, 1)
Moderate importance
MI (1/2, 1, 3/2) (2/3, 1, 2)
Equal importance
EI (1, 1, 1) (1, 1, 1)
Source: Tolga, et al. [102]
2. Fuzzy linguistic variable weight The selected set was defined as C=﹛??1, ??2, …, ????﹜, which was then transformed into the fuzzy positive reciprocal matrix �̃�? = a~ ???? , a
~ ???? ∈ [
1 9
, 9]. Let triangular fuzzy number a~ ???? represent the results of pairwise comparisons of attributes (fuzzy positive reciprocal matrix �̃�?), which was used to develop the Consistent Fuzzy Linguistic Preference Relations matrix ( )( )
nn k
ijk PP ×= ~~ (k=1, 2, 3, …, m) with
n-1 assessments {??12 (??), ??23
(??), ??34 (??), … , ??(??−1)??
(??) .
=
1 ~
… ~~
…………
~ …1
~~ ~
… ~
1 ~
~
21
221
112
nn
n
n
CC
CC CC
C
=
−−
−
1 ~
… ~~
…………
~ …1
~~ ~
… ~
1 ~
1 2
1 1
2 1
12
112
nn
n
n
CC
CC CC
52 International Journal of Electronic Commerce Studies
11111 9 ~
,7 ~
,9 ~
,5 ~
,7 ~
,3 ~
,5 ~
,3 ~
1 ~
,1 ~
,1 ~
~
−−−−−
=
= jiijC
Expert evaluation value ??� = �??�????� = �??????
?? , ?????? ??, ??????
??�, ????���� = (?????? (??)������)??×??(k=1, 2,
3, …, m) Where L is the number on the left side of the triangular fuzzy number, M is the central number in the triangular fuzzy number, and R is the number on the right side of the triangular fuzzy number. Following functions 4-1 to 4-7, by deriving the fuzzy linguistic variable preference values embedded in the matrix, a complete Consistent Fuzzy Linguistic Preference Relations matrix was established.
( ) ( ),~log1 2 1~~
9 ijijij aagP ⋅+⋅== (4-1)
Formulas (4-2)~(4-4) are now used to obtain the triangular fuzzy number in each field of the upper triangle in the matrix.
,1=+ Rji L
ij PP { },,…,1,, nkji ∈∀ (4-2) ,1=+ Mji
M ij PP { },,…,1,, nkji ∈∀ (4-3)
,1=+ Lji R
ij PP { },,…,1,, nkji ∈∀ (4-4) Formulas (4-5)~(4-7) are now used to obtain the triangular fuzzy number in each field of the lower triangle in the matrix.
( ) ( )( ) ( ) R
jj R
ii R
i L ji PPP
ij P 12111 …2
1 −+++ −−−
+− = (4-5)
( ) ( )( ) ( ) M
jj M
ii M
i M ji PPP
ij P 12111 …2
1 −+++ −−−
+− = (4-6)
( ) ( )( ) ( ) L
jj L
ii L
i R ji PPP
ij P 12111 …2
1 −+++ −−−
+− = (4-7)
By applying the functions 4-8, 4-9, and 4-10, all the fuzzy linguistic variable preference values ??�???? in the Consistent Fuzzy Linguistic Preference Relations matrix were within the range between 0 and 1, and the fuzzy linguistic preference matrix obtained using conversion function corresponding to the fuzzy set was uniformly within a certain scope, which maintained the consistency of addition and positive reciprocal numbers (c denotes the minimum value in the Consistent Fuzzy Linguistic Preference Relations matrix).
Function 4-11 was adopted to calculate all participants’ opinions by averaging participants’ ratings of each attribute.
( ) c cx
xf L
L
21 + +
= , [ ]ccc +−∈ 1, (4-8)
( ) c cx
xf M
M
21+ +
= , [ ]ccc +−∈ 1, (4-9)
( ) , 21 c
cx xf
R R
+ +
= [ ]ccc +−∈ 1, (4-10)
By comparison with dimension j, i is more important.
By comparison with dimension j, i is less important.
Tsuen-Ho Hsu, Sen-Tien Her, Yung-Han Chang and Jia-Jeng Hou 53
Function 4-12 calculated the mean of ????��, the averages of item i (where n is the number of attributes).
Weights normalization, the weight vector of attribute i, was obtained through Function 4-13.
Weight of each attribute was generated through Function 4-14. Defuzzied weights ( )niDi ,…,3,2,1= were derived based on each element ( )nix …, ,3 ,2 ,1= , and then
ranked in order.
After cleaning and organizing the data of 326 valid questionnaires from heavy users, the above functions were used to calculate the defuzzied weights of each dimension and attributes. The result indicated that the crucial attribute of Apple company’s marketing strategy was brand community value with a relative weight of 0.0631 (see Table 6). In other words, brand community value3 is the crucial factor which affects customers’ purchase of Apple products or services.
3Muniz and O’guinn [104, P412]:「A brand community is a specialized, non- geographically bound community, based on a structured set of social relationships among admirers of a brand. It is specialized because at its center is a branded good or service. Like other communities, it is marked by a shared consciousness, rituals and traditions, and a sense of moral responsibility. Each of these qualities is, however, situated within a commercial and mass-mediated ethos, and has its own particular expression. Brand communities are participants in the brand’s larger social construction and play a vital role in the brand’s ultimate legacy.」Zeithaml [105] defined “customer value” as the total benefits that customers obtained from the product or service after weighing against the costs paid. In a similar vein, this study conceptualized “brand community value” as the total benefits obtained by customers of a brand community who participated in the activities of the structures social relations after weighing against the total cost paid.
( )
,
~ ~ 1
m
P P
m
k
k ij
ij
∑ == ,, ji∀
(4-11)
,
~ ~ 1
n
P P
n
j ij
i
∑ == ,i∀
(4-12)
( ) 1
,ii n i
j
P W
P =
=
∑
(4-13)
( )1 3
L M R i i iD w w w= + + (4-14)
54 International Journal of Electronic Commerce Studies
Table 6. Apple company SIVA-Need model weights Dimension Attributes Relative
weight Rank Name weight rank name weight rank
Solutions 0.2490 2
Amenity 0.2115 2 0.0526 5 Commodity safety 0.2218 1 0.0552 3 Sociability 0.1859 5 0.0463 19 Exclusiveness 0.1947 3 0.0485 12 Personal style 0.1861 4 0.0463 18
Information 0.2391 3
Functionality 0.1944 5 0.0465 16 Information safety 0.2054 1 0.0491 9 Realtime interaction 0.2048 2 0.0490 10
Respect for diversity 0.2004 3 0.0479 13
completeness 0.1949 4 0.0466 15
Value 0.2770 1
Product benefits 0.2201 2 0.0610 2 Psychological benefits 0.1824 4 0.0505 6
Brand community value 0.2276 1 0.0631 1
Sense of honor 0.1938 3 0.0537 4 Personal value 0.1761 5 0.0488 11
Access 0.2349 4
Efficiency 0.2092 2 0.0491 8 Consumption safety 0.2118 1 0.0498 7
Belongingness 0.1788 5 0.0420 20 Courteous reception 0.2025 3 0.0476 14
Autonomous consumption 0.1977 4 0.0465 17
4.3 Apple company’s performance rating of each attribute in SIVA-Need model
Based on the resulting Apple company’s relative weights of attributes in SIVA- Need model, this study conducted a performance evaluation. A rating scale ranges from 1 to 5 was used to assess the “expected performance” and “actual performance” in terms of each attribute.
The relative weights of SIVA-Need attributes were denoted by (??????), the actual performance rated by participants was denoted by (????), and the expected performance was ( ???? ). Multiplying the relative weights by actual performance, the actual performance score (????) of SIVA-Need attributes was derived (Function 4-15). Dividing expected performance (????) by actual performance (????), the improvement score (????) of SIVA-Need attributes was derived (Function 4-16). The analysis results could be
Tsuen-Ho Hsu, Sen-Tien Her, Yung-Han Chang and Jia-Jeng Hou 55
provided as references for future improvements of the case study company in terms of SIVA-Need attributes.
Following functions 4-15 and 4-16, data collected from the 326 participants was calculated. The results indicated that brand community value had the biggest performance gap (1.50) and the highest performance improvement score (1.284) whereas product benefit had the highest performance score (0.235). For more details, please see Table 7.
Table 7. Apple company SIVA-Need performance score and improvement score category attributes
relative weight ( )
expected performance ( )
actual performance ( )
Performance gap
performance score ( )
improvement score ( )
Amenity 0.0526 4.55 3.80 0.75 0.200 1.198 Commodity safety 0.0552 4.54 4.00 0.54 0.221 1.136 Sociability 0.0463 3.95 3.50 0.45 0.162 1.129 Exclusiveness 0.0485 4.45 3.70 0.75 0.179 1.203 Personal style 0.0463 4.40 3.60 0.80 0.167 1.223 Functionality 0.0465 4.55 3.86 0.69 0.179 1.180 Information safety 0.0491 4.50 4.25 0.25 0.209 1.060 Realtime interaction 0.0490 4.20 3.75 0.45 0.184 1.120
Respect for diversity 0.0479 4.30 3.75 0.55 0.180 1.147
completeness 0.0466 4.55 3.80 0.75 0.177 1.198 Product benefits 0.0610 4.55 3.85 0.70 0.235 1.182 Psychological benefits 0.0505 4.35 3.40 0.95 0.172 1.279
Brand community value 0.0631 4.75 3.70 1.05 0.233 1.284
Sense of honor 0.0537 4.25 3.65 0.60 0.196 1.166 Personal value 0.0488 4.40 3.65 0.75 0.178 1.204 Efficiency 0.0491 4.45 3.80 0.65 0.187 1.171 Consumption safety 0.0498 4.35 3.85 0.50 0.192 1.129
Belongingness 0.0420 4.05 3.45 0.60 0.145 1.174 Courteous reception 0.0476 4.35 3.71 0.64 0.176 1.174
Autonomous consumption 0.0465 4.55 4.35 0.20 0.202 1.047
5. CONCLUSIONS AND SUGGESTIONS
???? = ?????? × ???? (4-15)
???? = ???? ÷ ???? (4-16)
56 International Journal of Electronic Commerce Studies
5.1 Discussion Research on customer-focused marketing conducted within the past 30 years has
explained the basic work of marketing [34], the role of organizational structure [35, 36], marketing capabilities [38], the role of endorsers [39], performance [37, 38, 40, 42, 106, 107], market orientation-marketing mix capability-new product performance[41], price systems and product development [108], customer satisfaction [109, 110] and other research results concerning organizational and marketing strategies. There are also studies explaining how customer-focused marketing can satisfy customers’ needs and wants, and thereby enhance customer satisfaction [45, 46]. Although these research results have been verified on a theoretical level, there is as yet no practical marketing strategy evaluation model that can measure the key attributes of customers’ needs and wants. Analysis of key attributes can enable a company to determine those elements that should be improved as a first priority in order to ensure customer satisfaction [57, p458], which implies that the analysis of key attributes is a very important part of the development and selection of marketing strategies. A SIVA marketing mix can be facilitate the discovery of customer needs and wants [71], and the intent of hierarchy of needs theory is to explain human needs, from the most basic to the highest. The core aspects of these two theories both emphasize human needs, and the 20 attributes developed in this study by linking the individual aspects of the two theories are the “needs” attributes of customers at each level of the hierarchy of needs when engaging in consumer activities. The marketing strategy MADM model—SIVA-Need— constructed using these attributes possesses the three characteristics of depth, breadth, and concreteness, and can be used to measure the key attributes of customers’ needs and wants. This model can therefore fill the current gap in research on customer-focused marketing strategies.