This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Journal of Industrial Engineering and Decision Making
Journal of Industrial Engineering and Decision Making 1 (1) (2020) 49-61
54
3.2. Conjoint Analysis (CA)
CA is a multivariate technique which is useful for understanding customer preferences for product or services. The product or service is composed of attributes and their levels. In the ordinal approach, linear models are applied based on the assumption that the value or utility of the product or service is the linear summation of values or utilities of attribute levels. The total part-worth of a product or service with m attributes with the maximum number of n levels for each attribute can be obtained via the following formula:
1 1T=
m n
ijj i
ijx w
(5)
Where, T is the total part-worth of the product, ijx is equal to one, if jth performance attribute is set
to level i and is equal to zero otherwise. ijw is the part-worth of the j attribute at the i performance
attribute level.
4. The framework
In this section, the proposed framework is explained. The framework which is depicted in Figure 1 has five phases. In the first phase, the utmost important criteria for credit card design based on an in-depth literature review were gathered. Then, a group of experts evaluated the criteria and the final list was constructed. In the second phase, the AHP method was used as a group decision-making tool by experts to rank criteria. Next, the validity of evaluations was checked with the C.R. index. In the third phase, CA applied the selected criteria to product profile design. Then, the sampling was done. Finally, in the interpretation phase, obtained results were applied to provide useful information for deciding the best credit card.
Figure 1. The proposed procedure
5. Case study
To demonstrate the applicability of the model, we conducted a case study in one of the very famous and reputable banks in Iran. XYZ Bank is the fourth major Iranian private bank, headquartered in Tehran, Iran. Based on the CEO’s comments, in 2018, 23.7 billion shares have been exchanged on the Tehran Stock Exchange and counts over 70,000 shareholders, and 6.5 million customers of the bank used transactions for the majority of imports of foodstuffs, medicine, and other humanitarian trade items.
Eliciting consumers’ preferences in service sector via Conjoint analysis: A case study on credit card
55
This section concentrates on obtained numerical results. In this study, we use expert’s evaluations, and sampling data to illustrate the proposed framework empirically. The next section explains about choosing the most important criteria by using pairwise comparisons. The conjoint analysis procedure is provided in the last section.
6.1. Criteria Selection and Pairwise Comparison
In this phase, a group of experts participated to run the project. So, the literature review was conducted to gather the most suitable criteria for credit card design. Table 3 shows the selected criteria list and their definitions.
Table 3. The definitions of credit cards factors
Variables Definitions
Balance Transfer Fee (C1) The money that one is charged when they transfer credit card debt from one card to another.
Annual Fees (C2) The fee that is charged for using the credit card service. Over-the-Limit Fee (C3) It is the money that you pay when your balance goes over your credit limit Interest Rate (C4) The price that a customer pays for borrowing money. Late Payment Fee (C5) The money that is charged to a borrower who misses paying at least their
minimum payment by the payment deadline. Credit Limit (C6) The maximum amount of credit that is extended to a customer. Brand (C8) The brand of the credit card. Foreign Transaction Fee (C7) It is money that a customer is charged for each transaction made abroad. Brand (C8) The brand of the credit card.
After creating the list of criteria, the group of experts was asked to do the pairwise comparisons in order to select the most relevant features. They used the scale provided in Table 1 to do the comparisons. All the pairwise comparisons and the calculated weights of the criteria are showed in Tables 4. We use Equations 1 to 4 for AHP calculations. The last column of Table 4 shows the weight of each criterion. Using the last column, one can see the rank/priority of each criterion. Based on Table 4, interest rate has the highest rank. The calculated C.R. factor for Table 4 is 0.07 and it is in the acceptable range (see Table 2). The top six ranked criteria are selected for CA in the next step.
Table 4. Pairwise comparisons between any two attributes
CA is one of the most suitable tools for understanding customers’ preferences. It is a widely accepted market research tool that is capable of designing and pricing a new product or service. The following steps are essential to perform for CA (Wedel and Kamakura, 2012; Gustafsson et al. 2003):
1. Identifying the attributes and levels: The first step of CA is to choose the attributes and their levels. The features/attributes are the key components that a customer uses them to evaluate a product or service. We used the AHP method for the selection of the attributes in order to reduce the number of features in CA. Table 5 shows six attributes and their levels. The main attributes are: Interest Rate, Annual Fees, Late Payment Fee, Credit Limit, Over-the-Limit Fee, and Brand.
Table 5. Conjoint attributes and attribute levels
Journal of Industrial Engineering and Decision Making 1 (1) (2020) 49-61
2. Stimulus set construction: In this study, we used a full-profile for CA. The full-profile has been a mainstay approach for decades in CA (Orme, 2005) and it can be applied to measure up to six attributes (Green and Srinivasan, 1978). To create a set of profiles for CA, we need to make a list of potential products based on different possible combinations of selected attributes. More precisely, the total number of hypothetical profiles of credit cards is calculated by multiplying the number of levels associated with each attribute. In our case study, it is 4 × 4 × 4 × 4 ×4 × 3 = 3,072 hypothetical profiles, which is a big number.
From the customers’ point of view, it is impossible to evaluate all the hypothetical profiles; and it is necessary to select a subset of them in a way that the performance of the evaluations stays the same. To solve this problem, fractional factorial design frequently is applied by many researchers and fractionates (Naes et al. 2001). To generate the profiles, we use SPSS-18.0. 29 profiles were generated by the software. Out of 29 profiles, 4 of them are called holdout profiles. Holdout profiles are designed to assess the reliability and validity of the response. An example of a profile card is shown in Table 6. Table 7 depicts a few numbers of profiles. A ten-point scale is used for evaluations. 3. Stimulus presentation: For collecting data we used a questionnaire.
4. Part-worth utility calculation
5. Calculating the relative importance of each attribute
6. Evaluating and interpreting the results
Table 6. Example of a profile card
Profile number: 9 How likely are you to purchase this Credit Card? .........
Interest Rate 40% Annual Fees 2,500,000 Late Payment Fee 0.03% Credit Limit 200,000,000 Over-the-Limit Fee 1,500,000 Brand B1 Least preferred …………………………………….......... Most preferred
Eliciting consumers’ preferences in service sector via Conjoint analysis: A case study on credit card
57
1 2 3 4 5 6 7 8 9 10 □ □ □ □ □ □ □ □ □ □
Table 7. Design of profiles
Profiles A1 A2 A3 A4 A5 A6
L11 L12 L13 L14
L21 L22 L23 L24
L31 L32 L33 L34
L41 L42 L43
L51 L52 L53 L54
L61 L62 L63 L64
1
… 10
… 20
… 29
The data was gathered using a questionnaire survey. The total sample size consists of 600 respondents, 260 females, and 340 males. All respondents were Iranian, aged 18 years old or older. The questionnaires were distributed in the Bank’s branches. The branches and the customers are selected randomly. Table 8 shows a wide range of socio-demographic information such as age, gender, marital status, education, occupation, and income.
Table 8. Socio-demographic characteristics of the sample (% of respondents, n = 600)
Gender Income class ($) Male 56.7 <= 400 71.3% Female 43.3 400-800 15.1% 800-1200 7.1% 1200-1600 3.6% >=1600 2.9% Age Education 18 to 24 40.3 Diploma or under 6.1% 24 to 30 33.4 Bachelor degree 58.4% 30 to 40 14.5 Master degree 26.7% > 40 11.8 PhD 8.8%
Table 9 shows the relative importance of each attribute and its levels. The last column of the table shows the average utility scores. For example, respondents preferred a high credit limit with a low interest rate. The negative sign shows that the attribute’s preference declines when the value increases. Based on the results, interest rate has the highest utility.
Table 9. The part-worth utility and relative importance for all the customers
To check the reliability, the goodness of fit for the estimated conjoint model is calculated using two measures. we found out that the value of Kendall’s tau is 0.903 and the value of Pearson’s R is 0.976 for all the samples. We have also used four stimuli as validation or holdout stimuli to determine internal validity. Parameters from the estimated conjoint model (using 25 stimuli) were used to predict preferences for the holdout set of stimuli and then they were compared with actual responses by calculating correlation. Considering the table (Table 10), we have found out that the value of Kendall’s tau is 1.000 for the four holdout cases in the overall sample and two segments. So, we can say that our conjoint model has high predictive accuracy and, internal validity.
Table 10. Correlations
Measure Overall
1 Pearson's R .976
2 Kendall's tau .903
3 Kendall's tau for Holdouts 1.000
7. CONCLUSIONS AND DIRECTIONS FOR FUTURE RESEARCH
In the preference modelling literature, the question of how human beings evaluates and chooses an alternative, for example buying a product is of paramount importance. Since one of the goals of marketing is providing a customer-oriented product or service, market researchers always seek to find how consumers evaluate and choose a product or service out of myriad goods in the marketplace. Answering this significant question can pave the way for other company’s strategies and plans. Moreover, the market is bursting with demanding customers, a multitude of competitors, and the overly dynamic environment. Similar to many companies, banks are invariably seeking for finding novel ways to elicit their consumer’s minds to absorb or keep them. Considering the above-mentioned conditions, banks need to provide the best possible services for their customers via eliciting their wants and needs.
The aim of this paper is to use AHP as a MADM tool and CA as a preference modelling tool to elicit preferences of credit card users. More precisely, AHP was applied to rank the selected criteria from the most to the least important ones. Then, the CA process used them to find the optimum combination of the factors for designing the most suitable credit card based on customer evaluations.
The paper has some limitations and drawbacks. Firstly, the sample size and sampling process are based on the available time and budget of the researchers. Secondly, the case study is conducted in one of the banks in Iran and the results and discussion are based on that. Thus the scale of this study is somehow small.
Future works will be based on addressing these themes. One possible theme is using fuzzy logic in conjoint and AHP methods and compares the results. Furthermore, this study showed that the selected factors can significantly influence the part-worth utilities so it is highly recommended to add or delete some of the evaluation criteria to see whether the ranking may change or not. In this paper, AHP was applied to select the factors, however, using other MADM methods can be considered as a new study. Moreover, applying other CA methods, such as Adaptive Conjoint Analysis (ACA) and choice-based conjoint analysis (CBA) or MaxDiff analysis is also recommended. Additionally, it is not possible to design one product for the whole market. Thus it would be a good idea to combine conjoint analysis with market segmentation (See Aghdaie et al. 2014). Although the model was applied in the service industry, in our study banking industries, it is possible to use the model in other industries, or in other countries.
Eliciting consumers’ preferences in service sector via Conjoint analysis: A case study on credit card
59
ACKNOWLEDGMENTS: The authors express their gratitude to the respectful editors for their constructive, valuable and encouraging comments.
REFERENCES
Aghdaie, M. H., Tafreshi, P. F., & Behzadian, M. (2014). Customer-oriented benefit segmentation: an
integrated approach, International Journal of Business Innovation and Research, 8(2), 168-189.
Altun, A., & Gök, B. (2010). Determining in-service training programs’ characteristics given to teachers by