Electronic Commerce Research and Applications 11 (2012) 89–100

Contents lists available at ScienceDirect

Electronic Commerce Research and Applications journal homepage: www.elsevier.com/locate/ecra

Exploring consumer adoption of new services by analyzing the behavior of 3G subscribers: An empirical case study Li-Chen Cheng ⇑, Li-Min Sun Department of Computer Science and Information Management, Soochow University, Taipei 100, Taiwan, ROC

a r t i c l e

i n f o

Article history: Available online 17 June 2011 Keywords: 3G application services Association rules Data mining Mobile commerce RFM model

a b s t r a c t As the profit margins of 3G mobile network operators gradually decline, and market competition becomes increasingly intensive, they must develop rich and diverse varieties of brand new application services to attract new subscribers and retain old ones. Understanding the customer’s purchasing behavior is a key issue in this process. The operator must accurately grasp movements in the market based on analysis of the behavior of 3G subscribers. This study proposes a comprehensive customer relationship management strategy framework to furnish a beneficial plan to overcome such challenges. First, we propose a new model to identify who are the high-value customers related to the characteristics of new telecommunication services. After segmenting the customers, we propose a procedure to provide different kinds of usage analysis, including inter-cluster analysis and intra-cluster analysis. The experimental results are determined based on rules extracted from a large number of call detail records generated by the mobile subscribers of leading 3G mobile system operators in Taiwan. The dependency network demonstrates the relationship between voice services, data communications, message services, micropayments and entertainment. Finally, we propose some marketing recommendations for 3G system operators based on these interesting rules. Ó 2011 Elsevier B.V. All rights reserved.

1. Introduction Diverse mobile applications and services undoubtedly influence the daily life of mobile users around the world significantly. Furthermore, the maturing of mobile technologies effectuates e-commerce over mobile platforms. Mobile commerce (m-commerce) means that any activity can be transacted business via a mobile device (Clarke 2001). M-commerce can be considered an extension of e-commerce (Coursaris and Hassanein 2002, Ngai and Gunasekaran 2007). According to a study by Juniper Research, the global m-commerce market will total an industry worth of US$115 billion by 2015 (Juniper Research 2011). 3G telecommunication operators have an important role in the m-commerce value chain, according to Kuo and Yu (2006). Several studies have suggested that the next phase of e-business growth will occur in wireless services and m-commerce (Fouskas et al. 2005, Smith 2006, Baldwin et al. 2007). This trend is creating promising new business opportunities for m-commerce development. The International Telecommunication Union (ITU) reported that in 2009 mobile global subscriptions numbered 4.6 billion users (ITU 2010). The ITU expected this number to reach 5 billion in 2010, while the world population in April 2010 was 6.8 billion ⇑ Corresponding author. Tel.: +886 2 23111531x3812; fax: +886 2 23756878. E-mail address: [email protected] (L.-C. Cheng). 1567-4223/$ - see front matter Ó 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.elerap.2011.06.005

people. Though the number of 3G subscribers is raising rapidly, average mobile revenue per user (ARPU) is dropping in the majority of Asian Pacific markets. Mobile operators must manage nonvoice ARPU more effectively to overcome this business challenge (Kuo and Yen 2009, Deng et al. 2010). To fulfill this requirement, the telecommunications industry is constantly developing new and innovative 3G value-added services to meet the various needs of consumers. Although a variety of mobile value-added services have been released, whether consumers will purchase these services remains unknown (Teng et al. 2009). Understanding customer-purchasing behavior has become a key issue. 3G operators employing objective data analysis to identify heavy users of new products and investigate the behavioral differences between heavy and non-heavy users has become a critical exercise (Koivumaki et al. 2006, Schierza et al. 2010). These results will support the development of different consumption behavior models, providing differentiated products and services regarding marketing and maintenance of customer relationships, which will significantly enhance future marketing strategies (Wu and Wang 2005). Data mining generally entails technologies discovering previously unknown information and summarizing relevant information from a vast number of databases to assist business decisions (Cabana et al. 2008). Appropriate customer relationship management (CRM) strategies can be adopted with the assistance of data mining technologies, which can manage the data required to

90

L.-C. Cheng, L.-M. Sun / Electronic Commerce Research and Applications 11 (2012) 89–100

enhance understanding of customers (Ngai 2005). Successful applications of data mining techniques in CRM include churn prediction, fraud detection, cross selling, customer segmentation, and customer retention (Berson et al. 2000, Wei and Chiu 2002, Liu and Shih 2005, Hung et al. 2006, Cheng and Chen 2009, Chen et al. 2009). One of the most well-known examples is that of Wal-Mart, which used association rules to discover that beer and diapers are often purchased together (Hughes 1994, Groth 2000, Han and Kamber 2007). This study uses the data mining technique of association rules due to its efficiency (Agrawal et al. 1993, Liu and Shih 2005). Previous research in the area of m-commerce focused on exploring mobile consumer intention or acceptance (Hsu et al. 2007, Wang et al. 2006, Koivumaki et al. 2006, Schierza et al. 2010). There is a lack of knowledge about how to investigate the consumption behavioral differences between early adopters and other users. Studies of the behaviors that analyze the behaviors of valuable 3G mobile consumers and real consumer usage are rare and a fitting methodology for analysis of consumer behaviors in the case of 3G telecommunications needs to be developed (Bose and Chen 2010). This paper provides two main contributions: from a conceptual viewpoint, this study proposes a comprehensive CRM strategy framework that contains a customer segmentation process and a behavior analysis process. The proposed framework can address the lack of knowledge regarding consumption behavioral differences between early adopters and other users. Firstly, we built a customer segmentation model, the TFM, which depends on the value determinations of 3G mobile services. We define heavy users as early adopters who use 3G services frequently (F), accumulate a greater volume of service time (T), and have large bills for a time period of one month (M). Other studies also define heavy users similarly (Lim et al. 2005, Gensler et al. 2007). This study then proposes a procedure for consumer behavioral analysis and a comparison of different behavioral clusters. 3G operators can use these tools to understand the behaviors of valuable users and propose novel marketing strategies to increase customer adoption. From an empirical viewpoint, the proposed framework draws on 3G value-added usage data obtained from a leading telecommunications firm in Taiwan. To the best of our knowledge, no behavioral analysis of 3G mobile subscribers using association rules extracted from real usage patterns exists for understanding behavioral differences between valuable clusters. Taiwan is one of the most crucial m-commerce markets in the world. At the end of December 2009, the number of mobile phone subscriptions in Taiwan had risen to 26.96 million, a penetration rate of 116.6% (in other words, 116 mobile phone numbers for every 100 people). Furthermore, 58.7% of subscribers used 3G services. This rate implies a potential representative market for enhancing understanding of 3G valued-added service adoption. This paper hence provides sound strategies to increase new service adoption for 3G companies. The rest of the paper is organized as follows: Section 2 briefly reviews existing literature; Section 3 describes the proposed research modules; Section 4 presents an analysis of the proposed procedures in a series of experiments; Section 5 provides several managerial implications for marketing reference; and finally, the last section offers conclusions such as the limitations of this study and considerations for future research. 2. Related work After the application of customer value analysis for customer segmentation, this study conducted behavioral analysis of the different segments to propose marketing recommendations for mobile operators. This section reviews literature related to customer relationship management, customer segmentation, association rule studies, and research studies of 3G mobile service adoption.

2.1. Data mining in customer relationship management After surveying several studies, Ngai (2005) argued that CRM is a comprehensive set of strategies managing relationships with customers related to the overall process of marketing, sales, service, and support within the organization. Payne and Frow (2005) proposed a strategic framework for successful CRM implementation that contains five key processes: strategy development, value creation, channel and media integration, information management, and performance assessment. In Payne’s framework, the most critical customer strategy should comprise examining the existing and potential customer base and identifying which forms of segmentation is most appropriate. Upon conducting customer segmentation, enterprises could focus on the character of different user groups to propose more efficient operations when acquiring and retaining the relationship to increase overall business profits (Payne and Frow 2005, Thomas and Sullivan 2005). The objective of these strategic operations is to build long-term relationships for increasing customer satisfaction, strengthening customer loyalty, and raising profitability (Swift 2000, Mithas et al. 2005). Data mining is regarded as a powerful tool to find relevant rules and behavior patterns from the analysis of a large amount of data (Ngai et al. 2009). Appropriate data mining tools, which are suitable for extracting and identifying useful information and knowledge from massive customer databases, are one of the best support tools for forming various CRM decisions (Berson et al. 2000, Bose and Chen 2010). Association rule is the best-known technique for customer purchase analysis. Liu applied association rules to extract knowledge from customer purchase history for the development of one-to-one marketing (Liu and Shih 2005). The method has been widely used in various areas, such as for mining user access patterns on websites, using POS information to extract interorganizational retailing knowledge, recommending products to users, and cross sailing (Berson et al. 2000, Lin et al. 2003, Sohn and Kim 2003, Liu and Shih 2005, Chen et al. 2009). The American Management Association estimates that attracting new customers costs five times the amount of retaining existing ones. Therefore, business organizations can also use association rules to study customer behaviors and provide healthier management of customer relationships. 2.2. Customer segmentation and RFM models Customer segmentation divides all customers into an appropriate number of clusters to effect customized strategies for meeting different customer needs. Early approaches for segmentation include the use of demographic, geographic, situation (Gehrt and Shim 2003), lifestyle and behavioral characteristics of consumers (Plummer 1974, Assael and Roscoe 1976, Punj and Stewart 1983, Hoek et al. 1996, Schijns and Schroder 1996, Gehrt and Shim 2003, Bose and Chen 2010). However, Gupta et al. (2006) suggests that the past purchases of consumers are superior predictors for determining future purchase behavior compared to demographics. Enterprises can apply proper techniques to discover consumerpurchasing patterns by examining the customer sales records to propose useful marketing strategies (Chandon et al. 2005, Bose and Chen 2010). Recency, frequency, and monetary (RFM) variables are the most frequently used behavioral data in marketing research (Bult and Wansbeek 1995). Hughes defines recency as the period since a customer’s last purchase, frequency as the number of purchases made within a certain time period; and monetary as the amount of money that a customer spent during a certain period (1994). The RFM model as a market segmentation tool not only quantifies customer behavior, but also identifies the most profitable customers (Goodman 1992, Miglautsch 2000). Incorporating the concept of

91

L.-C. Cheng, L.-M. Sun / Electronic Commerce Research and Applications 11 (2012) 89–100

RFM modeling into the data mining technique is common practice, such as in customer loyalty prediction, churn prediction, customer retention, and cross selling (Cui et al. 2006, Cheng and Chen 2009, Etzion et al. 2005). Several applications based on the RFM model have been proposed for use in the mobile communication industry, including churn prediction, customer retention, and cross selling (Mozer et al. 2000, Wei and Chiu 2002, Weiss 2005). For telecommunication and Internet service providers, highvalue customers are those with high monthly telecommunication expenses. However, these customers do not necessarily have high incomes; in fact, they could even be students. Segmenting customers by actual consumption behavior rather than demographics is thus a reasonable and appropriate approach. This study develops a new model targeting products with special attributes, utilizing the model to segment consumers and analyze consumer behavior.

This study defines heavy users as early adopters who use 3G services frequently, accumulate a greater volume of service time, and have large bills during a time period of one month. Early adopters spend more time surfing the 3G services compared to other users. Studies of the behaviors of valuable 3G mobile consumers and real consumer usage are rare (Bose and Chen 2010). A fitting methodology to conduct analyses of consumer behavior of 3G telecommunications must be developed. 3. Model and methodology In this section, we briefly introduce the proposed framework and describe procedures for analysis of real call data produced by 3G mobile users. 3.1. Research model

2.3. Association rules Association rule mining is another major data mining technique mostly used to discover multiple independent elements that frequently co-occur, and seek to extract rules in association with the co-occurring elements in a given dataset (Agrawal et al. 1993). A well-known application of association rules is market basket analysis, which contains a customer’s purchasing transactions, that is, the itemsets that are purchased by a customer in a single transaction. The number of customer transactions is typically quite high, and the number of items in frequent itemsets can increase exponentially. Association rules can be used to examine as many frequent itemsets as possible, answering queries such as what products tend to be purchased together. Association rules can reduce a large amount of information to a small and more comprehensive set of statistically supported statements (Han and Kamber 2007). For example, they can reveal that customers who purchased Product A have an 81.25% probability of buying Product B. Conversely, customers who purchased Product B have a 65% probability of buying Product A. A typical association rule assumes the form A ? B, where A and B are in an item set that contains only a single atomic condition. A rule also can be represented as an antecedent (left-hand side) and a consequent (right-hand side). The intersection between the antecedent and the consequent is empty. The support of an association rule is the percentage of records containing items A and B together. The confidence of a rule is the percentage of records containing item A that also contains item B. The confidence represents the strength of the rule, and the support indicates the frequency of the patterns occurring in the rule. Rules with high confidence and robust support are referred to as strong rules. 2.4. Research regarding 3G mobile service adoption Prior studies in m-commerce and related fields have focused on examining mobile consumer intention or acceptance (Wang et al. 2006, Koivumaki et al. 2006, Hsu 2007, Schierza et al. 2010), churn prediction or loyalty (Wei and Chiu 2002, Hung et al. 2006), value of consumers or services (Kim et al. 2007), and so on. Numerous studies have applied innovation diffusion theory to analyze the manner in which new innovations spread among groups of people (Moore and Benbasat 1991). Consumers who adopt new innovations can be categorized as innovators/early adopters, early majority, late majority, and laggards (Rogers 2003). Karahanna et al. (1999) indicated that different beliefs regarding IT use exist between potential adopters and users. The gap between early adopters and early majority is the most challenging when concerned with reaching the mass market (Moore and Benbasat 1991). There is a lack of knowledge about how to investigate the consumption behavioral differences between early adopters and early majority.

The conceptual framework for the CRM strategy proposed Payne and Frow (2005) mentions the information management process (Payne and Frow 2005). The proposed framework completely maps the customer strategy process, customer segmentation versus segment granularity, rules extraction versus customer characteristics. Thus, our results can be beneficial in the value creation and multichannel integration process. The aim of this study is to produce a strategy for developing marketing and product recommendations based on real data generated by customers, in this case mobile operators. These providers have to continually meet new customer demands and be able to dynamically shift their business strategies in order to gain a competitive advantage among intense rivalries. The analysis of the consumer behaviors of valuable customer segments is a starting point to accomplish this goal. We construct a model for clustering customers based on some special attributes and then analyze customer behaviors by finding association rules from each segment. The proposed framework is illustrated in Fig. 1. We further explain the procedures for data analysis produced by mobile users. The proposed model is described in detail below. 3.2. Data preprocessing We concentrate on data from mobile users who subscribe or use 3G application services. The real call data are provided by a wellknown telecommunications company in Taiwan. Each of these records represented a user’s usage of services. They reflected customers’ behaviors such as usage of application type and the

Data preprocessing

Customer segmentation

Proposed Model TFM model

Rules extraction from each segment Mining the difference of customer behavior Process for behavior comparisons Fig. 1. The proposed CRM framework.

92

L.-C. Cheng, L.-M. Sun / Electronic Commerce Research and Applications 11 (2012) 89–100

interval of time for application using. First, we extracted the experimental dataset from the databases and calculate the average billing amount per month. Second, mobile users whose billing amounts per month are larger than NT$850, the monthly fee for unlimited 3G mobile Internet services, are noted. Third, the training dataset is compiled. It is found that 70% of mobile users have billing amounts larger than NT$850 per month and 30% have billing amounts smaller than NT$850 per month. This critical amount is used to define valuable customers for the operator. The cut in the dataset is determined as suggested by marketing practitioners in the 3G mobile services industry. Lastly, the experimental dataset is organized and cleaned.

Table 1 Description of product categories. Product category

Product sub-category

Voice services

Intra-network call Inter-network call International roaming call International direct dial CHT 2G call CHT local call Non-CHT 2G call Non-CHT local call 070 VoIP call

Message services

Intra-network message Inter-network message Message package Value-added message Multimedia value-added message MMS International message

Micropayment services

Micropayment-Internet multimedia Micropayment-destiny Micropayment-online game Micropayment-entertainment Micropayment-parking chargess Micropayment-donation

Data communication services

WLAN service

3.3. Customer segmentation with the proposed new model According to the definition of the RFM model proposed by Hughes (1994), the three attributes are equal in importance. That is the three attributes for evaluating customer value should be equally weighted. However, the RFM model was originally developed to analyze products within each transaction for each consumer in a specific time in the retail industry (Goodman 1992, Miglautsch 2000). 3G mobile applications users may subscribe to applications every few minutes. In other words although most retail buyers only carry out one transaction in a short interval of time, mobile users may conduct many transactions during the same interval of time. Their purchasing behaviors are quite varied. In addition, since the prices of 3G services are determined by quantity, there will be many new product attributes. Three key attributes are used to find heavy users who accumulate a greater volume of service time (T), purchase 3G services frequently (F) and amass large billing amounts per month (M). Thus, we propose a fitting model, the TFM, which will cluster consumers based on some special attributes. The aim is to realizing not only how to cluster but also the scales of each attribute in each segment. The three key attributes are defined as follows:  Average call (usage) time per month (T). The T value means the average time spent calling or using the application services in the dataset in one month. When the call time of the user is larger than the pre-calculated average, the value of T is 2; otherwise, it is 1.  Average call (usage) frequency per month (F). The F value refers to the frequency of calling or using application services in the dataset in one month. If the calling (using) frequency is larger than the pre-calculated average, the value of F value is 2; if less than the pre-calculated average it is 1.  Average billing amount per month (M): The M value refers to the billing amount in the dataset for calling or using application services in one month. The scheme for calculation of charges for 3G application services in Taiwan is complicated because of diverse monthly fees and service packages, including unlimited services like 3G mobile Internet and other high monthly-fee services. Users subscribing to this kind of service and paying monthly fees to the operator can certainly be classified as belonging to valuable segments. The rules can be used to categorize high monthly billing amounts. (See Table 1.) The M value is set from 1 to 4 corresponding to the billing amount, from low to high. The threshold for the billing amount categorization is chosen based on specific statistics and recommendations offered by the operator and domain experts. Thus, target users are segmented into clusters 1–5 using the TFM model, where the high-value cluster is comprised of cluster 1, the medium-value cluster contains clusters 2–4, and the low-value cluster is distributed to cluster 5. According to innovation dif-

Internet surfing through emome-monthly charge MDVPN communications Unlimited monthly Internet surfing Internet surfing-monthly charge Entertainment services

Video call JAVA application Multimedia application MP3 application Mobile TV application Gaming application

fusion theory, the adoption of innovative technology is closely related to willingness to try and accept new things (Morre 1991). The greater the volume of these 3G services used, the greater the billing amounts they will accumulate. Heavy users of 3G services are early adopters that belong to cluster 1. 3.4. Rule extraction for each segment After the preceding steps are completed, consumer behavior in relation to new products in the market can be analyzed using the clustering mechanism. Comparison of behavioral differences between early adopters and early majority users, for example, may help marketing managers to align their strategies for mass adopters to ‘‘cross the chasm’’ (Moore and Benbasat 1991). Next, clusters will be analyzed based on their use of 3G services and association rules. Rules generated from each segment will be inspected for diverse levels of support and confidence. Implicit rules reveal what kinds of services are usually used together. The association rules that emerged from the experimental dataset will be summarized in the next section. There may be thousands of rules related to hundreds of products output by the data mining tool; such rules become complicated and are not easy to interpret. The only thing we can do is to reduce the number of rules without altering their essence. We will adopt the concept of hierarchical technology to simplify the rules (Han and Kamber 2007). We interviewed domain experts to determine abstract product categories. A summary of the product categories and sub-categories based on the opinions of experts and product statements of the firm is given in Table 2.

L.-C. Cheng, L.-M. Sun / Electronic Commerce Research and Applications 11 (2012) 89–100 Table 2 Segmentation rules of billing amounts. Cluster 1 2 3 4 5

High-value (early adopters) Medium-value (mass adopters)

Low-value

T

F

M

2 2 2 1 –

2 2 1 2 –

4 3 3 3 2

Traditional voice calls including local, international, intra-network and inter-network calls are categorized as voice services. Message applications including text messaging, MMS and valueadded services are classified as message services. Prepaid services for a variety of products belong to the category of micropayment services. Data communication applications for mobile users to connect to the Internet or VPN are categorized as data communications services. Multimedia applications involving MP3, games and mobile TV are included under mobile entertainment services. Finally, we will discuss both within-cluster and between-cluster aspects of rules for these five clusters. The association rules for each cluster have been grouped by selecting the right-hand side (consequent) item to which product category it belongs. 3.5. Process for behavior comparison The whole process of behavior comparison can be divided into two steps, including intra-cluster analysis that can help to understand specific cluster characteristics, and inter-cluster analysis for discussing the differences for one-on-one behavioral comparison. 3.5.1. Intra-cluster analysis At the beginning, we discover the general rules within each cluster that have the same support value. Then, the dependency network can be determined to illustrate the relationships among five product categories. Among these general rules, we filter out some interesting rules for marketing purposes. Changed rules show the implicit patterns that target products are consequents under different support values. We want to discover significant rules that change as the support values decrease, such as the change from A?B to C?B. These patterns can describe various services associated with new services. Otherwise, the trend for adopting services appears while adjusting the support values. 3.5.2. Inter-cluster analysis In this study, customers are divided into three groups, based on high, medium and low values. The goal of the analysis is to discover meaningful market knowledge. Inter-cluster analysis may be used to generate special patterns for unexpected rules and inter-cluster changing rules. The unexpected rules are filtering the changing rules among clusters with the same support value for designing suitable marketing packages. Finally, we observe changes in consumption behaviors among clusters under different support values. The changes in adopting specific services among the three clusters may be found. 4. Empirical case study In this section, the experimental results are discussed. The results are obtained based on the rules extracted from a large amount of call detail records (CDRs) generated by the mobile subscribers of a leading 3G mobile operator in Taiwan. This corporation is the most experienced and integrated leader in the telecom industry in Taiwan, and mainly provide telecommunication and information-related services. The experimental dataset in this

93

study is sampled from its CDR database, excluding demographic data (to preserve personal privacy). This dealer was the first authorized Apple iPhone reseller and mobile network provider at the end of August 2009. Its 3G mobile application CDRs exceed billions of records every month. The sampling time period of the experimental dataset is from July 2009 to December 2009 with a total number of subscribers of approximately 650,000. All selected data were cleaned and processed for further analysis in advance. Before conducting data mining analysis, we first divided the subscribers into three clusters based on the TFM model. The association rules for each cluster are grouped by selecting the right-hand side consequent item associated with the product categories to which it belongs. The level of support will vary for the rules in each category. From the variation in support values, we can make relevant inferences. Rules can be discussed from both the intra-cluster and inter-cluster points of view for these two critical variables, support and category. In the following discussion, we will focus on emerging services. The goal is for the operator to increase the average revenue per user (ARPU). Voice, message and micropayment services are not of major interest in this study. 4.1. Intra-cluster analysis The three clusters are mapped into high-value, medium-value and low-value groups. Valuable clusters have greater influence on creating value and their characteristics help to decide on future promotion and marketing activities. Detailed rules have to be conferred within individual clusters, especially for high-value and medium-value clusters, in order to offer some relevant recommendations. The dependency network indicates the relationships among services. Associations that exist are represented with solid lines. The objective is to observe changed rules for emerging services of data communication and entertainment services. Given this condition, support values are set to decrease to be 0.7, 0.5 and 0.3. In this case, we will discuss rules that are changed as the support values decrease leading to adjustment for the intracluster view. 4.1.1. Results for the high-value cluster with early adopters Subscribers are confirmed to be of high-value to the operator when their volume of usage or monthly charge has exceeded a threshold. Fig. 2 indicates tight associations among voice, message and data communications services. The three services are critical components of marketing portfolios. The relation between data communications and entertainment services is weak, which suggests why they are not frequently used. Connecting micropayment services with the other four types of services would be an interesting trend. 4.1.1.1. Summarizing general rules. Most new services, including Internet surfing, MDVPN, MP3 apps and gaming apps are shown in Table 3. The results tell us that the high-value cluster (early adopters) tends to use such innovative applications. Traditional services also play an important role if usage for this cluster. Nevertheless, there are a few services such as video calling, multimedia apps and Java apps that are not generally accepted. The marketing team could investigate to find out reasons why these are so rarely used within the high-value cluster. Perhaps a bundle package offering both rare and typical services at a discount will be a more meaningful stimulus for subscribers who have not tried new services yet. All in all, the number of subscribers in this cluster is quite substantial, both because of their high loyalty and ARPU. The marketing strategy is designed to completely meet implicit and explicit subscriber demands. Implicit demand can be discovered by peri-

94

L.-C. Cheng, L.-M. Sun / Electronic Commerce Research and Applications 11 (2012) 89–100

Voice services

Voice services

Entertainment services

Data comm. services

Message services

Data comm. services

Entertainment services

Micropayment services

Message services

Micropayment services

Fig. 3. Dependency network for the medium-value cluster (mass adopters). Fig. 2. Dependency network of the high-value cluster (early adopters).

odic rule mining. The operator should continue to observe behavioral trends within this cluster as long as possible. 4.1.1.2. Summarizing changed rules. The change of rules within the innovation cluster shows a special phenomenon that entertainment services will be fully present until support decreases to 0.3. (See Table 4.) There are two basic reasons this service is needed, one is for the data communication group; the other is for the entertainment group. Both demand the extent of importance to the service. Moreover, this link can be considered during process of product design and marketing. Similarly, we try to choose rules for this service that show implicit patterns. The results show that entertainment users will choose to use homogeneous services when the threshold is re-

duced to 0.3. See Table 5. This means that entertainment users have the characteristic of using analogous services, and it is not quite obvious. The potential for volume growth can be enhanced and revenues increased by providing combinations of new and old products or even offering interactive activities between users and products. 4.1.2. Results for the medium-value cluster It can be seen that the patterns in Figs. 2 and 3 are similar to those for the high-value cluster except for the partial relation between data communications and entertainment services. The marketing strategy could include various rewards for subscribers in this cluster to encourage consideration of voice, data communications or message services. The greater the volume of these new services used, the higher the discounts they will acquire. The

Table 3 Significant rules for the high-value cluster. Category

Consequent

Antecedent

Data communications

Unlimited monthly Internet surfing MDVPN communications Internet surfing through emome-monthly charge MP3 application Gaming application MP3 application

Micropayment-online game Message package Intra-network message Intra-network call Intra-network call Inter-network message

Entertainment

WLAN service Intra-network call Non-CHT 2G call Intra-network message Intra-network message Mobile TV application

Table 4 Significant rules for data communication services within the high-value cluster. Support

Consequent

Antecedent

0.7 ? 0.5

Internet surfing-monthly charge Internet surfing-monthly charge

MDVPN communications WLAN service

MP3 application MP3 application

0.5 ? 0.3

Internet surfing-monthly charge Internet surfing-monthly charge

Mobile TV application Mobile TV application

Video call Multimedia application

Table 5 Significant rules for entertainment services within the high-value cluster. Support

Consequent

Antecedent

0.7 ? 0.5

JAVA app. Mobile TV application

Intra-network call WLAN service

WLAN service Inter-network call

0.5 ? 0.3

JAVA application Mobile TV application

MP3 application Multimedia application

Mobile TV application JAVA application

95

L.-C. Cheng, L.-M. Sun / Electronic Commerce Research and Applications 11 (2012) 89–100 Table 6 Significant rules for medium-value clusters. Category

Consequent

Antecedent

Data communications

MDVPN communications Internet surfing-monthly charge

Intra-network call WLAN service

Intra-network message MDVPN communications

Entertainment

MP3 application Mobile TV application JAVA application Video call

Intra-network call Message package Micropayment-online game MMS

Inter-network call Intra-network call Intra-network call CHT local call

Table 7 Significant rules for data communication services within the medium-value cluster. Support

Consequent

Antecedent

0.7 ? 0.5

MDVPN communications MDVPN communications

International messaging Intra-network call

International roaming call International message

0.5 ? 0.3

Internet surfing-monthly charge Internet surfing-monthly charge

Intra-network call Intra-network call

MP3 application Mobile TV application

Table 8 Significant rules for entertainment services within the medium-value cluster. Support

Consequent

Antecedent

0.7 ? 0.5

MP3 application

WLAN service

Intra-network call

MP3 application

Intra-network call

Internet surfing-monthly charge

Multimedia application Gaming application

Intra-network call Intra-network call

JAVA application

0.5 ? 0.3

Voice services

Message services

Data comm. services

Mobile TV application

advantage of ARPU from emerging services will be inspired, and the loyalty of this cluster will improve. Alternatively, some or all fundamental services can be included in the reward plan for fulfilling diverse demands of subscribers based on cost concerns. 4.1.2.1. Summarizing general rules. Although almost all of the data communication and entertainment services have rules presented from Table 6, we still have the chance to continue to promote and increase the volume of these services. 4.1.2.2. Summarizing changed rules. As can be seen in Table 7 voice services will appear in MDVPN communications when the support is 0.5 within this cluster. This behavior tells us that business subscribers in this cluster are likely to use voice services and data communication services. These could be offered as a business package for users within this cluster. Entertainment services are involved as support is adjusted downwards to 0.3. This characteristic indicates that some users who have subscribed to the Internet surfing service may also be interested in entertainment services. The marketing plan for this cluster should also consider these types of entertainment services. When the support is 0.5, entertainment and data communication services will exist simultaneously again. See Table 8. When

Entertainment services

Micropayment services

Fig. 4. Dependency network for the low-value cluster.

support is set to 0.3, more entertainment services are present. These rules confirm the linkage between entertainment and data communication services. This really reveals the importance of bundling these two services. Nevertheless, we suggest that effective marketing strategies need to be drawn up considering the details of the products offered. 4.1.3. Results for low-value clusters This cluster uses few new services and their volume and monthly charges are low. Fig. 4 and Table 9 show that most such subscribers use voice as well as message services. This means that only a small number of subscribers are interested in a handful of new services. Due to limited contribution to revenues, the rate of return from such customers may be low. All the operator can do is to maintain the status quo and offer a basic discount.

Table 9 Significant rules for low-value clusters. Category

Consequent

Antecedent

Data communications Entertainment

Internet surfing-monthly charge MP3 application

Intra-network call Intra-network call

Inter-network call Intra-network message

96

L.-C. Cheng, L.-M. Sun / Electronic Commerce Research and Applications 11 (2012) 89–100 Table 10 Significant rules for data communication services. Cluster

Consequent

Antecedent

High-value Medium-value

MDVPN communications MDVPN communications

Message package Intra-network call

Intra-network call Intra-network message

High-value Medium-value

Unlimited monthly Internet surfing N/A

Micropayment-online game N/A

WLAN service N/A

High-value Medium-value

Internet surfing-monthly charge Internet surfing-monthly charge

Micropayment-parking charges WLAN service

CHT 2G call MDVPN communications

Table 11 Significant rules for entertainment services. Cluster

Consequent

Antecedent

High-value Medium-value

MP3 application MP3 application

Intra-network call Intra-network call

Intra-network message Inter-network call

High-value Medium-value

Gaming application Gaming application

Intra-network call Micropayment-online game

Intra-network message Intra-network call

High-value Medium-value

Mobile TV application Mobile TV application

Intra-network call Message package

Micropayment-online game Intra-network call

High-value Medium-value

JAVA application JAVA application

Intra-network call Micropayment-online game

WLAN service Intra-network call

40

12 10

30

Number of rules

Number of rules

35

25 20 15 10

6 4 2

5 0

8

0.3

0.5

0.7

0

0.3

0.5

Support value High value

Medium value

High value

Low value

Fig. 5. Voice service rules among the three clusters over distinct support values.

Medium value

Low value

Fig. 7. Micropayment service rules over distinct support values.

comparing pair-wise rules for as many valuable clusters as possible. The marketing team can conduct efficient marketing plans for the enlargement of ARPU to fulfill a variety of demands as determined in this initial study. We will compare rules in the inter-cluster view.

35 30

Number of rules

0.7

Support value

25 20 15 10 5 0

0.3

0.5

0.7

Support value High value

Medium value

Low value

Fig. 6. Message service rules among the three clusters over distinct support values.

4.2. Inter-cluster analysis The difference between clusters during segmentation helps to identify behaviors that are specific to each cluster. We want to know the difference between high-value clusters and mediumvalue clusters. In this section, we will define characteristics by

4.2.1. Rules for the same product category at the same support value This study found that it is worth discussing discrepant rules for data communications and entertainment services, but not conventional services. The support is 0.5. The following paragraphs contain an explanation of significant rules related to data communications and entertainment services. Pair-wise comparison is used to distinguish between high-value clusters and medium-value clusters. The behavior of both clusters using MDVPN services is similar. Subscribers also use intra-network call and message services. (See Table 10.) The evidence suggests that these services can be combined to form an effective marketing package for future promotion. Meanwhile, MDVPN is definitely used for business applications, so the promotion could be developed for customized needs. Extensive marketing plans can be based on related analysis. Unlimited Internet surfing could be part of a special ‘‘all you can surf’’ service. The subscriber can surf the Internet anytime anywhere. All you have to pay is the surfing charges month by month

97

L.-C. Cheng, L.-M. Sun / Electronic Commerce Research and Applications 11 (2012) 89–100

25

Number of rules

20 15 10 5 0

0.3

0.5

0.7

Support value High value

Medium value

Low value

Number of rules

Fig. 8. Data communications service rules for the three clusters over distinct support values.

20 18 16 14 12 10 8 6 4 2 0

0.3

0.5

0.7

Support value High value

Medium value

Low value

Fig. 9. Entertainment service rules for the three clusters over distinct support values.

Table 12 Significant rules for data communications services over distinct support values. Support

Cluster

Consequent

Antecedent

0.7

Highvalue

Internet surfingmonthly charge

Mediumvalue

Internet surfingmonthly charge

Micropaymentparking charges WLAN service

0.5

Highvalue Mediumvalue

Internet surfingmonthly charge Internet surfing with emomemonthly charge

Mobile TV application Intra-network call

Intra-network call MDVPN communications

0.3

Highvalue Mediumvalue

Internet surfingmonthly charge Internet surfingmonthly charge

Mobile TV application Mobile TV application

Video call

CHT 2G call

MDVPN communication

MP3 application

until you cancel it. The rule indicates that this type of service only exists in the high-value cluster due to the charges. If the marketing team could think of a way to promote or offer discounts for WLAN service or other homogeneous services, they could perhaps attract users within the medium-value cluster to this service. Data-limited and time-limited Internet surfing services face the same type of challenge in terms of offering tempting and targeted promotion packages for conventional services. Entertainment services are related to regular calls, messages and on-line game playing. See Table 11. The results suggest that subscribers interested in entertainments services may desire to have a flexible discount package that would combine general calls and messages with other related services. Accordingly, the ARPU could be progressively expanded by enlarging the usage capacity as much as possible. 4.2.2. Rules for the same product category over distinct support values Figs. 5–7 represent the number of voice, message and micropayment service rules, respectively, for which the support values exceed a certain value. This study concentrates on emerging services. Although message services are still profitable to the operator, there has been no progress with innovative product development. Capital investment and marketing expenditure by the operator have been concentrated on developing competitive services such as data communication and entertainment. This is main reason why we are keen to discussing these well-known applications. Figs. 8 and 9 show that both data communication and entertainment services have similar trends. The high-value cluster always has more rules than other clusters with diverse support values, and the medium-value cluster is next. The difference for entertainment services is slight. Tables 12 and 13 show a comparison of rules for these two services over distinct support values. The table above delineates pair-wise data communication service rules with high and medium-values over perspicuous support values. When the support is 0.7 or 0.5, the innovator cluster has data communication services along with entertainment services and the medium-value cluster reflects the definite characteristics of business applications. The results show that the early adopters cluster can likely be defined as a ‘‘data-or-entertainment’’ group, and the medium-value cluster can possibly be defined as a ‘‘business’’ group. When the support is 0.3, data communication and entertainment services are connected to some extent according to our rules. In fact, further work needs to be done to establish more accurate definitions of our clusters. There was evidence that both clusters will adopt entertainment services when support is lower. These two clusters are obvious for potential users of entertainment services. Our results show that entertainment services are subscribed to within both clusters at every support level. The discount package may attract potential users. Innovation and development of entertainment services are also critical to gain competitive advantage.

Table 13 Significant entertainment service rules over distinct support values. Support

Cluster

Consequent

Antecedent

0.7

High-value Medium-value

MP3 application MP3 application

Inter-network call Inter-network call

Mobile TV application Inter-network call

0.5

High-value Medium-value

Mobile TV application Mobile TV application

Intra-network call Internet surfing through emome-monthly charge

Micropayment-online game Intra-network call

0.3

High-value Medium-value

Internet surfing-monthly charge Internet surfing-monthly charge

Mobile TV application MP3 application

Intra-network call Inter-network call

98

L.-C. Cheng, L.-M. Sun / Electronic Commerce Research and Applications 11 (2012) 89–100

5. Discussion and implications The goal of this experiment and case study has been to understand the behaviors of users in order to give some marketing knowledge to mobile network operators. On the basis of our categorized mining results, we propose some recommendations for mobile network operators from the marketing, product and operational points of view. This could be a useful reference for developing business strategies. The following sections discuss the results and implications based on our findings. 5.1. Discussion on intra-cluster rules In intra-cluster analysis, the trend for service adoption is shown while adjusting the support values. 5.1.1. High-value clusters for early adopters The early adopters tend to use innovative applications, such as Internet surfing, MDVPN, MP3 apps, Mobile TV and so on. This observation is consistent with Rogers’ viewpoint. The dependency network for the high-value cluster indicates tight associations among voice, message and data communications services. It is interesting that the dependency network of the high-value cluster indicates that micropayment services are connected with the other four services. The early adopters always adopt unlimited Internet surfing services and micropayment services. This makes for reliable payments for most users. Additionally, the early adopters may not be interested in entertainment services. The higher the usage of data communications services, the less usage there is of other services. It’s because that early adopters can download games, Mp3 from the web site via a mobile network. Since this is such a special phenomenon, the operator should become a content provider just like the App Store which contains various categories contents including the e-books, games, apps and so on. Therefore, the operator’s revenue can come from not only the application stores but also the micropayment service fees. In addition, operator should notice that mobile TV may become a killer application for 3G entertainment service. 5.1.2. Medium-value clusters It is possible to define the medium-value cluster as the ‘‘business’’ group. Most patterns in the dependency network are similar to those in the high-value cluster except for the partial relation between data communications and entertainment. Most users subscribe to Internet surfing services with monthly charges are also interested in entertainment services. As shown in Table 8, voice service always appears with MDVPN communications and WLAN service. Meanwhile, MDVPN and WLAN are definitely used for business applications, so their promotion could be developed for customized needs. This means that the operator can offer discount packages including these services simultaneously. 5.1.3. Low-value clusters Most users still subscribe to traditional services such as voice and messages. According to the Pareto principle, the contribution of revenues from this cluster may be low. It is reasonable to assume that most such users still use 2G phones. The operator should offer promotions to attract these users to 3G services. At first, the subscribers should be offered a free 3G phone with some marketing plans. 5.2. Discussion rules for the inter-cluster view From intra-cluster analysis, we can observe changes in consumption behaviors among clusters under different support values.

There are changes for adopting specific services between the three clusters. A comparison of behavioral differences between early adopters and mass adopters may help marketing managers to align their strategies for mass adopters to accept new services. Mobile voice calls still play an important role in contributing revenue to the operator. However, emerging services such as mobile Internet may have a direct or indirect influence on voice revenues. If the operator offers attractive packages, they could maintain stable growth of profits derived from voice calls. Although mobile message services are critical applications, we find in the experiments that the use of text or multimedia message services is still not very effective. The growth of messaging services resulting from aggressive marketing activities should occur though. It is surprising that the usage and volume of mobile entertainment is not significant in the valuable users group. Therefore, the operator could consider offering a discount combination of mobile entertainment, video calls and data communications services as an incentive to potential users. Furthermore, the geographic, demographic, lifestyle, and socio-graphic characteristics of valuable segments of 3G users can be analyzed jointly for discovery of detailed usage. An interesting finding is that the unlimited Internet surfing service only exists in the innovative cluster. In addition, this service always comes with micropayment services. Incidentally, more than 2 billion applications have been downloaded from Apple Inc.’s App Store. The more such devices there are out there, the greater the number of people that want to download software. They see that it is an easy and fun experience but how to pay for the service is another problem. Micropayment may be an important service in m-commerce. The use of MDVPN services is similar for both high-value and medium-value clusters. The evidence suggests that these services can be combined to form an effective marketing package for future promotion. 6. Conclusion At present, 3G operators are facing fierce competition. They need to constantly introduce new products and services to attract the customers. Whether consumers will purchase their new services is a problem. 6.1. Contributions Most existing studies of m-commerce and related issues have focused on exploring mobile consumer intention or acceptance. This study is intended to be a valuable source for further empirical and conceptual research on 3G value-added services. We proposed a strategy framework to make up for the lack of knowledge about how to investigate the consumption behavior differences between early adopters and other users. We first proposed a new customer segmentation model that will help to identify high-value 3G consumers and to understand customer purchase behavior. By knowing the differences between early adopters and potential users, companies can develop different marketing strategies to attract new mobile services users and retain existing ones. To the best of our knowledge, there has been no research conducted on behavioral analysis of 3G mobile subscribers using association rules extracted from their real usage patterns with the goal of ascertaining behavioral differences between value clusters. The experimental results are based on rules extracted from nearly 50 million customer data records generated by the mobile subscribers of one of the chief competitors among 3G mobile system operators in Taiwan. This rich data set supported our development of results that provide valuable knowledge about their

L.-C. Cheng, L.-M. Sun / Electronic Commerce Research and Applications 11 (2012) 89–100

customers’ behavior patterns. Furthermore, this study provides several important practical guidelines for 3G operator. The purpose of fulfilling the explicit and implicit demands of valuable customers and increasing revenues for operators can be accomplished, creating a win-win situation. Despite plentiful services and applications now provided by mobile network operators, particular ones like the entertainment and payment have not been extensively adopted within the valuable customer groups. Operators should not only concentrate on capital expenditure and development of brand-new services but also be aware of expectations or feedback from target users. This study offers opportunities for mobile network operators and researchers to confer about issues related to services and future users. 6.2. Limitations and future research Our study has some limitations. The first limitation is that all the respondents came from one mobile operator. Although our sampling data come from the biggest mobile operator in Taiwan, there may be a sampling bias. Second, the result cannot be generalized to interpret the behaviors of all mobile phone users. The threshold amount per month for the value segments should be adjusted dynamically for the analyzing the behaviors of adopters of homogeneous or heterogeneous 3G applications and services for the sake of observing long and short-term users. Third, users’ demographic data, such as gender and age, may affect the adoption of new services too. However, the sample data do not include the customers’ demographic information to preserve personal privacy. Despite these limitations, however, this study provides insights into the adoption behavior of mobile services with a new proposed framework. In the future, we will extend the empirical study to other countries to explore different behavior patterns. The relationship between purchase intention and actual purchase behavior is an interesting and critical issue (Chandon et al. 2005). Since the purpose of this study was to observe purchasing behavior, the tracing of consumer intentions was not conducted. Future studies could be carried out to fill in the gap between purchase intentions and actual purchase behavior. Also, sequential patterns mining methods can be applied to investigate changes in the behavior of the early adopters. Acknowledgments It is our pleasure to acknowledge the anonymous reviewers for their valuable suggestions and the careful reading of our manuscript. The authors would like to express our gratitude to these reviewers for their suggestions that helped to substantially improve our paper. This research was supported in part by the National Science Council of the Taiwan (Republic of China) under the Grant NSC 98-2410-H-031-001. References Agrawal, R., Imielinski, T., and Swami, A. Mining association rules between sets of items in large databases. In P. Buneman and S. Jajodia (eds.), Proceedings of the ACM SIGMOD Conference on Management of Data, May26–28, 1993, ACM Press, New York, NY, 1993, 207–216. Assael, B., and Roscoe, A. Approach to market segmentation analysis. Journal of Marketing, 40, 1976, 67–76. Baldwin, L. P., Low, P. H., Picton, C., and Young, T. The use of mobile devices for information sharing in a technology-supported model of care in A&E. International Journal of Electronic Healthcare, 3, 1, 2007, 90–106. Berson, A., Smith, S., and Thearling, K. Building Data Mining Applications for CRM. McGraw-Hill, New York, NY, 2000. Bose, I., and Chen, X. Exploring business opportunities from mobile services data of customers: an inter-cluster analysis approach. Electronic Commerce Research and Applications, 9, 3, 2010, 197–208.

99

Bult, J. R., and Wansbeek, T. Optimal selection for direct mail. Marketing Science, 14, 4, 1995, 378–394. Cabana, P., Hadjinian, P., Stadler, R., Verhees, J., and Zanasi, A. Discovering Data Mining: From Concept to Implementation. Prentice Hall, Upper Saddle River, NJ, 2008. Chandon Morwitz, V. G., and Reinartz, W. J. Do intentions really predict behavior? Self-generated validity effects in survey research. Journal of Marketing, 69, 2, 2005, 1–14. Chen, Y. L., Kuo, M. H., Wu, S. Y., and Tang, K. Discovering recency, frequency, and monetary (RFM) sequential patterns from customers’ purchasing data. Electronic Commerce Research and Applications, 8, 5, 2009, 241–245. Cheng, C. H., and Chen, Y. S. Classifying the segmentation of customer value via RFM model and RS theory. Expert Systems with Applications, 36, 3, 2009, 4176–4184. Clarke, I. III., Emerging value propositions for m-commerce. Journal of Business Strategies, 18, 2, 2001, 133–148. Coursaris, C., and Hassanein, K. Understanding m-commerce: a consumer centric model. Quarterly Journal of Electronic Commerce, 3, 3, 2002, 247–271. Cui, G., Wong, M. L., and Lui, H. K. Machine learning for direct marketing response models: Bayesian networks with evolutionary programming. Management Science, 52, 4, 2006, 597–612. Deng, Z., Lu, Y., Wang Zhang, B. J., and Wei, K. K. An empirical analysis of factors influencing users’ adoption and use of mobile services in China. International Journal of Mobile Communications, 8, 5, 2010, 561–585. Etzion, O., Fisher, A., and Wasserkrug, S. e-CLV: a modeling approach for customer lifetime evaluation in e-Commerce domains, with an application and case study for online auction. Information Systems Frontiers, 7, 4–5, 2005, 421–434. Fouskas, K., Giaglis, G., Kourouthanassis, P., Karnouskos, S., Pitsillides, A., and Stylianou, M. A roadmap for research in mobile business. International Journal of Mobile Communications, 3, 2005, 350–373. Gehrt, K., and Shim, S. Situational segmentation in the international marketplace: the Japanese snack market. International Marketing Review, 20, 2, 2003, 180– 194. Gensler, S., Dekimpe, M. G., and Skiera, B. Evaluating channel performance in multichannel environments. Journal of Retailing and Consumer Services, 14, 1, 2007, 17–23. Gupta, S., Hanssens, D., Hardie, B., Kahn, W., Kumar, V., Lin, N., Ravishanker, N., and Sriram, S. Modeling customer lifetime value. Journal of Service Research, 9, 2, 2006, 139–155. Goodman, J. Leveraging the customer database to your competitive advantage. Direct Marketing, 55, 8, 1992, 26–27. Groth, R. Data Mining: Building Competitive Advantage. Prentice Hall, Upper Saddle River, NJ, 2000. Han, J., and Kamber, M. Data Mining: Concepts and Techniques, 2nd edition. Morgan Kaufman, San Francisco, CA, 2007. Hoek, J., Gendall, P., and Esslemont, D. Market segmentation: a search for the Holy Grail? Journal of Marketing Practice, 2, 1, 1996, 1355–2538. Hsu, C. L., Lu, H. P., and Hsu, H. H. Adoption of the mobile Internet: an empirical study of multimedia message service. Omega, 35, 6, 2007, 715–726. Hughes, A. M. Strategic Database Marketing. Probus, Chicago, IL, 1994. Hung, S. Y., Yen, D. C., and Wang, H. Y. Applying data mining to telecom churn management. Expert Systems with Applications, 31, 3, 2006, 515–524. ITU, International Telecommunication Union. ITU sees 5 billion mobile subscriptions globally in 2010. Available at www.itu.int/newsroom/ press_releases/2010/06.html. February 15, 2010. Juniper Research, Global mobile statistics forecast. March 2011. Available at mobithinking.com/stats-corner/global-mobile-statistics-2011-all-qualitymobile-marketing-research-mobile-web-stats-su. Karahanna, E., Straub, D. W., and Chervany, N. L. Information technology adoption across time: a cross-sectional comparison of pre-adoption and post-adoption beliefs. MIS Quarterly, 23, 2, 1999, 183–213. Kim, H. W., Chan, H. C., and Gupta, S. Value–based adoption of mobile Internet: an empirical investigation. Decision Support Systems, 43, 1, 2007, 111–126. Koivumaki, T., Ristola, A., and Kesti, M. Predicting consumer acceptance in mobile services: empirical evidence from an experimental end user environment. International Journal of Mobile Communications, 4, 4, 2006, 418–435. Kuo, Y. F., and Yen, S. N. Towards an understanding of the behavioral intention to use 3G mobile value-added services. Computers in Human Behavior, 25, 1, 2009, 103–110. Kuo, Y. F., and Yu, C. W. 3G telecommunication operators’ challenges and roles: a perspective of mobile commerce value chain. Technovation, 26, 12, 2006, 1347– 1356. Lim, J., Currim, I., and Andrews, R. Consumer heterogeneity in the longer-term effects of price promotions. International Journal of Research in Marketing, 22, 4, 2005, 441–457. Lin, Q. Y., Chen, Y. L., Chen, J. S., and Chen, Y. C. Mining interorganizational retailing knowledge for an alliance formed by competitive firms. Information and Management, 40, 5, 2003, 431–442. Liu, D. R., and Shih, Y. Y. Integrating AHP and data mining for product recommendation based on customer lifetime value. Information and Management, 42, 2005, 340–387. Miglautsch, J. Thoughts on RFM scoring. Journal of Database Marketing, 8, 1, 2000, 67–72. Mithas, S., Krishnan, M. S., and Fornell, C. Why do customer relationship management applications affect customer satisfaction? Journal of Marketing, 69, 2005, 167–176.

100

L.-C. Cheng, L.-M. Sun / Electronic Commerce Research and Applications 11 (2012) 89–100

Moore, G. C., and Benbasat, I. Development of an instrument to measure the perceptions of adopting an information technology innovation. Information Systems Research, 2, 3, 1991, 192–222. Mozer, M. C., Wolniewicz, R., Grimes, D. B., Johnson, E., and Kaushansky, H. Predicting subscriber dissatisfaction and improving retention in the wireless telecommunications industry. IEEE Transactions on Neural Networks, 11, 3, 2000, 690–696. Ngai, E. W. T. Customer relationship management research (1992–2002): an academic literature review and classification. Marketing Intelligence & Planning, 23, 6, 2005, 582–605. Ngai, E. W. T., and Gunasekaran, A. A review for mobile commerce research and applications. Decision Support Systems, 43, 1, 2007, 3–15. Ngai, E. W. T., Xie, L., and Chau, D. Application of data mining techniques in customer relationship management: a literature review and classification. Expert Systems and Applications, 36, 2, 2009, 2592–2602. Payne, A. F. T., and Frow, P. Strategic framework for customer relationship management. Journal of Marketing, 69, 2005, 167–176. Plummer, J. The concept and application of lifestyle segmentation. Journal of Marketing, 38, 1974, 33–37. Punj, G., and Stewart, D. W. Cluster analysis in marketing research: review and suggestions for application. Journal of Marketing Research, 20, 2, 1983, 134– 148. Rogers, E. M. Diffusion of innovations, 5th edition. Free Press, New York, NY, 2003. Schierza, P. G., Schilkeb, O., and Wirtzc, B. W. Understanding consumer acceptance of mobile payment services: an empirical analysis. Electronic Commerce Research and Applications, 9, 3, 2010, 209–216.

Schijns, J. M. C., and Schroder, G. J. Segment selection by relationship strength. Journal of Direct Marketing, 10, 3, 1996, 69–79. Smith, A. Exploring m-commerce in terms of viability, growth and challenges. International Journal of Mobile Communications, 4, 6, 2006, 682–703. Sohn, S. Y., and Kim, Y. Searching customer patterns of mobile service using clustering and quantitative association rule. Expert Systems with Applications, 34, 2, 2003, 1070–1077. Swift, Ronald S. Accelerating Customer Relationships-Using CRM and Relationship Technologies. Prentice Hall, Upper Saddle River, NJ, 2000. Teng, W. C., Lu, H. P., and Yu, H. J. Exploring the mass adoption of third-generation (3G) mobile phones in Taiwan. Telecommunications Policy, 33, 10–11, 2009, 628–641. Thomas, J. S., and Sullivan, U. Y. Managing marketing communications with multichannel customers. Journal of Marketing, 69, 2005, 239–251. Wang, Y. S., Lin, H. H., and Luarn, P. Predicting consumer intention to use mobile service. Information Systems Journal, 16, 2, 2006, 157–159. Wei, C. P., and Chiu, I. T. Turning telecommunications call details to churn prediction: a data mining approach. Expert Systems with Applications, 23, 2, 2002, 103–112. Weiss, G. M. Data mining in the telecommunications. In O. Maimon and L. Rokach (eds.), Data Mining and Knowledge Discovery Handbook: A Complete Guide for Practitioners and Researchers, Kluwer Academic Publishers, New York, NY, 2005, 1189–1201. Wu, J. H., and Wang, S. C. What drives mobile commerce? An empirical evaluation of the revised technology acceptance model. Information and Management, 42, 5, 2005, 719–729.