A model is a representation of an original, normally omitting some features but retaining others, usually for a specific purpose. Thus, a model airplane may capture the appearance or the flight characteristics of the original, while a business model usually represents the financial structure and dynamic relationships between demand, supply, and financial performance. For new products, a marketing model should capture not only demand patterns (the behavior of buyers with varying attributes) but also changes in these patterns as the market gains experience with the new product.
The evidence suggests that operators have yet to form a clear picture of their MDS markets. Typically, mobile operators segment their markets into two or three tiers, based on the average monthly bill. This retrospective view provides little insight into preferences for future services. Also, the adoption pattern for MDS is strikingly inconsistent. In Europe, where text messaging is hugely popular, take-up of WAP-based data services is sluggish . The combination of geographic isolation, a social system that supports early adoption of new technologies, and operator focus on new market segments is the basis for a "new geographic focus for mobile technologies in the Nordic region of Europe" (McKnight, 2001, p. 6). In North America, where legacy infrastructure and multi-bearer interoperability issues represent additional barriers to adoption, the MDS outlook is even more uncertain . Industry watchers see Asia as a bright spot for MDS, ranging from the startling growth of text messaging in China, the Philippines, and other markets to the accelerating adoption of multimedia services in Japan and South Korea (McFetrich, Chaudhry, & Bauer, 2001).
In Singapore, WAP services have yet to attract the necessary critical mass despite heavily promoted launches by three mobile operators (Gilbert, 2000). In Japan, DoCoMo's i-Mode service was a huge success, signing up 10 million users in less than one year and soon tripling that (Kuchinskas, 2000). However, its 3G service has not met expectations (BBC News Business, 2002). Clearly, the dynamics of the fit between service offerings and demand are a key success factor.
Many of the causes of new product failure ‚ small market scale, a poor fit between product features and demand, a lack of innovation or buyer value, poor product positioning, channel problems, and forecasting errors ‚ are market related (Urban & Hauser, 1980). Management must balance development cost, time to market, and risk with the potential value of new products, market them to the potential buyers most likely to derive immediate value, then leverage the successful experiences of early adopters as a base for market expansion. Risks along this path are high, especially when the new product or service rests on digital network platforms that demand substantial initial investments, yet enable a wide array of value propositions .
These risks are compounded by the huge license fees and massive infrastructure costs required to deploy "third-generation" (3G) broadband mobile telephone infrastructure. The high fixed costs to develop service offerings should motivate providers to minimize uncertainties regarding the dynamics of MDS adoption. Fortunately, the recent emergence of "2.5G" upgrades to existing second-generation infrastructure provides relatively low-cost opportunities to learn in the current context. Thus, the IMARC research sought to develop a marketing model that identifies products that will appeal to early adopters in each segment and that targets the most effective information channels for influencing these lead customers.
For voice calls and peer-to-peer SMS, the operator provides only the connection, while users provide the content. The MDS value proposition is quite different: as with the Internet, value is created through the delivery of data. As individual users are unlikely to assign similar value to each MDS product, it is necessary to develop a map of the value assigned to specific products for various user types. Market segmentation categorizes subjects according to known attributes that are believed to predict future behavior. Such attributes may be demographic, geographic, socioeconomic, or psychographic (Weinstein, 1994). However, in a dynamic environment such as MDS, simple demographic or economic data are decreasingly useful as a basis for segmentation (Urban, Weinberg, & Hauser, 1996). This is because while certain aspects of segment behavior are homogeneous and membership in a segment can be global, the socioeconomic profiles of members may be heterogeneous. New psychographic clusters based on perceived usage and benefits, known as lifestyle and need-based segmentation, have gained credence for explaining how adopters of new technologies behave (Michman, 1991). Needs-based segmentation is appropriate for information- intensive environments, where marketing and communication strategies are shifting from the traditional mass media orientation to a personalized approach (Hoffman & Novak, 1996).
Subscribers are likely to adopt cellular telephones both to aid in their work and also due to social pressure and their perceived ideas of the cell phone as a security device and source of enjoyment (Davis, 1993; Kwon, 1994). Davis, in his study of the social impact of cellular telephones, reports that they are especially useful for maintaining interpersonal relationships and are also seen as a powerful medium for decision making. The study identified both internal and external motivations for early adoption of cellular telephones. Deci (1971, 1972, 1975) contends that intrinsic and extrinsic motivation determines behavior. Extrinsic motivation refers to activity in pursuit of external rewards such as income or some other perceived utility, while intrinsic motivation refers to activity in search of satisfaction and enjoyment (Davis, Bagozzi, & Warshaw, 1992).
The Rogers (1995) innovation model posits five segments based on the timing of their adoption behavior: (1) innovators, (2) early adopters, (3) early majority, (4) late majority, and (5) laggards. As members of each segment perceive innovations differently and may acquire information about their characteristics from different sources, knowledge of their needs and behaviors can focus the use of resources to help prevent a newly introduced innovation from failing (Miller, 1993). Service providers can devise a dynamic marketing plan by cross-referencing these time-based segments to the relationships among their demographics and psychographics (Weinstein, 1994, pp. 205-223). An operator can also use this knowledge to develop a positioning strategy and guide other tactical activity (Kotler & Armstrong, 1999, pp. 46). In each segment, new products can be positioned dynamically based on perceived value, price, and expected adoption rates over time in the target segment and in related segments.
The researchers selected innovation theory as the primary lens through which to view the MDS phenomenon across a series of four studies. Each study applied different research methods to answer specific questions. The first, conducted in late 2000, examined WAP services as the mostly widely available interactive platform in use at the time and derived a segmentation model (Gilbert & Sangwan, 2002). Two follow-on studies, conducted a year later, applied both survey and focus group techniques to examine emergent demand patterns for a specific MDS category: mobile entertainment. A fourth field study, conducted in early 2002, captured cross-border preferences for MDS through multilingual surveys of several hundred mobile phone users in Singapore and Malaysia.
All field research followed the general sequence and protocols described below:
Identify current marketplace trends by examining recent industry publications and mass media.
Organize focus groups to explore emerging issues, refine research questions, and gather impressionistic data.
Design survey instruments to capture data about the most interesting research questions from the perspective of the relevant theory.
Field-test survey questions with members of the target groups and refine research instruments.
Collect and interpret data, avoiding systematic bias to the extent possible.
Survey data was entered into spreadsheets, edited, imported into SPSS, then analyzed using a variety of statistical tools to reveal segments based on attitudes or needs (Urban & Hauser, 1980). Relationships reported below were significant at P<.05, except where otherwise noted.
The first round of field research, completed in early 2001, explored the early stages of mobile data services deployment in Singapore, on SMS and WAP platforms (Chia, Hazam Aris Bin, Ho, 2001). This began with a focus group composed of 20 GSM subscribers. None were WAP users. The focus groups helped clarify research questions and refine the survey, distributed to 300 undergraduate and postgraduate students. The unit of analysis was the individual decision to adopt WAP services rather than the device. Of 300 forms distributed, 198 were returned, for a 65% return rate. Respondents were evenly distributed along age, gender, and occupational lines. More than 85% owned and used cell phones, a figure slightly higher than the national average for adults.
In contrast to other cell phone users, WAP subscribers in the survey panel were more likely to be male and to perceive themselves as technologically savvy persons to whom others turn for guidance. They were more likely to use SMS and personal organizer functions on their cell phones and to use cable and wireless modems to connect their personal computers to the Internet. For information about technology and technology-based services, compared to nonsubscribers, WAP subscribers were significantly more dependent on mass media channels and less dependent on advice from family, friends , and colleagues. These findings confirm the need to match the communications channel to the intended target (Rogers, 1995) over time.
Exploratory R-type factor analysis examined relationships among key interval-scaled survey questions regarding the intention to use MDS services. This analysis, combined with the focus group findings, helped identify segments within each service market. The analysis revealed relationships among the intention to use MDS, specific service requirements, and demographic variables , allowing identification of early adopter segments.  Ordered based on their time of adoption, five needs-based segments emerge:
TechnoToy: adopt services that fill needs for hands-on knowledge about technological developments. Their adoption decisions are influenced by technical reports and journals.
Mobile Professionals: adopt services including calendaring and access to mobile e-mail and intranet/extranet services. Influenced mainly by opportunities to create new value related to work life or by decisions to adopt mobile technology by employers or client organizations.
Sophisticates: adopt services and products that fill needs for status in terms of material style. Influenced by images that are projected by celebrity users and in mass media.
Socialites: services that meet needs to keep in touch while on the go. Influenced mainly by family and friends.
Lifestyle: these services, partly overlapping the categories listed above, fill convenience needs related to mobile lifestyles, such as delivering information or directions to people who are in an unfamiliar location and helping people fill "dead time" with time-critical tasks. Examples of such tasks include messaging and bill paying while in waiting lines or on public transport or facilitating meetings among friends who are on the move.
Two added segments, containing users whose needs were unlikely to motivate them to adopt, emerged:
Misers: members of this segment were unwilling to pay for wireless data services.
Laggards: were the last to know about and adopt new technologies.
The model identifies seven distinct segments (see Figure 1), differentiated by behavior linked to time of adoption, the primary value derived from MDS, and the primary information channel for information that influenced their adoption decisions. The model has potential utility in that it focuses marketing decisions on shifts in application types and information channels over time. However, the validity of the model is bounded by the small sample size , while its generalizability is limited by the narrow selection of available MDS applications in many categories and the single-country sample.
The next phase sharpened our focus. Wireless entertainment includes a wide range of products such as games , ring tones, logos, simulated voice mail from celebrities such as Arnold Schwarzenegger, and access to entertainmentrelated information. Ovum Consulting Group (2002) predicts that global wireless gaming, now booming in Japan, will generate US$1 billion in revenue by 2004, rising to US$4.3 billion by 2006, while other analysts place an even higher value on this new market. By 2006, 30% of the world's 53 million wireless gamers will be in the Asia-Pacific region. By 2004, ring tones alone are expected to be a billion-dollar business (The Economist, 2002).
To improve the validity of our model, focus group and survey research were combined to explore mobile gaming. Forty-five respondents (including young working adults, undergraduate university students, and school leavers) participated in nine focus group sessions. The process followed the information acceleration methodology by Urban et al. (1996), a technique that has been tested in various contexts, ranging from new medical technology to car sales. As MDS products are new, consumers have little knowledge of their value. All panel members interacted with a laptop-based cell phone simulator running mobile gaming.
The panel members concurred that most currently available mobile games were essentially too simple to be compelling. When bored, they "turn to their cell phone games for companionship only out of sheer desperation". They felt this could change with multiplayer capability, larger color displays, better input methods, and compatibility across different playing platforms. Three distinct segments emerged from our focus groups: hardcore gamers, casual gamers, and non-gamers (Han et al., 2002).
Analysis of survey returns from 300 cell phone users in Singapore, with demographic profiles similar to the first-phase MDS subjects, revealed three adopter segments at the application level: hardcore, social, and casual gamers (Ovum Consulting Group, 2001). As there are (yet) no professional mobile gamers, the Mobile Professional segment did not surface. Also, current mobile gaming technology may not be sufficiently advanced to interest the TechnoToys segment.
There were few significant differences between gender, age, or income groups within the three segments, outlined below.
Dedicated gamers prefer mobile games like board, card/tile, puzzle, word, and quizzes. They see themselves playing mobile games when they are waiting for an appointment or traveling, provided they like the game. Members of this group are likely to be digital gamers, male, currently studying , younger , and with a below-average income. While the demographics of members of this group resemble the TechnoToys segment identified in Phase 1, knowledge acquisition was not their source of motivation.
Social gamers are likely to adopt mobile gaming for interactivity. They prefer dating games and enjoy meeting people of the opposite gender. They are also likely to play mobile games when their friends are also playing. The attributes of social game players resemble those of the Socialites identified in our earlier study.
Casual gamers , who make up the majority of the sample, neither prefer a specific type of mobile game nor reveal unique mobile gaming behavior or lifestyle needs. This result agrees with Ovum Consulting Group's (2001) white paper: casual gamers make up the mass market. This group parallels the Lifestylers segment in the earlier study.
A great majority of the respondents stay abreast of the latest technology (75%), although only 35% consider themselves early adopters. Three major reasons to why respondents play games on their mobile phones are because they (1) are traveling, (2) are waiting for an appointment, and (3) like the game. Those who did not play these games reported reasons including (1) they have no time, (2) they have no interest, and (3) existing games are not appealing.
The survey findings converged with opinions recorded during independently conducted focus group sessions: "Games now are too limited and too simple and are interesting only at the beginning, when it is a new phone, or when I'm stranded with nothing but a cell phone". The focus group sessions revealed three distinct segments: hardcore gamers, casual gamers, and non-gamers. The hardcore gamers are already active local area network (LAN) players of PC games such as Counterstrike, War Craft 3, and Diablo. They are keen to see popular games on the mobile platform, especially multiplayer games. The attributes of this segment parallel those of dedicated gamers found in the survey research. They were also similar in terms of digital games preferences and behavior.
Casual gamers do not actively play but do occasionally join in. The segment appears to represent the majority of the Singapore population, an impression from the focus groups that our survey confirms. The focus group and survey findings varied in several interesting respects. The focus group sessions detected a segment that was not keen on any form of gaming capabilities. To them, the only purpose of a mobile phone was its communication function, and games were perceived as a gimmick. However, some members of this segment had a utilitarian interest in information-based MDS. The survey also found a social gamer segment, those who play games for social interactivity.
To extend the research scope beyond local markets, the next step was a comparative survey of attitudes and adoption behavior in Singapore and Malaysia, sponsored by a regional mobile operator (Olsen et al., 2002). A survey instrument was created and pretested in English, Chinese, and Malay languages to reach out to all respondents, then administered through face-to-face contact. To ensure that the survey captured the same information in each language, a multilingual panel crosschecked the translations. The questions focused on current and planned behaviors rather than attitudes and values.
The target population was cell phone users between the ages of 15 and 60 living in major metropolitan areas in Singapore and Malaysia ‚ where wireless networks are fully developed. The sampling frame consisted of hand phone users walking in busy city street areas in Singapore, Kuala Lumpur, and Johor Bahru. To reduce sampling frame error, participants were screened for whether or not they used a hand phone. A central location-intercept quota sampling technique was used. Quotas were based on ITU estimates of user demographics (Weinstein, 2001). The technique, even though non-probabilistic, enjoys conditions roughly similar to that obtained under probability sampling (Kalton, 1983) in being representative of the population. Survey data was entered into spreadsheets, checked and edited, imported into SPSS 11, and then analyzed to reveal segments based on attitudes or needs (Urban & Hauser, 1980). The resulting sample was evenly balanced in gender terms and approximates the demographic profiles of cell phone users in the two markets.
To link between what an operator knows about its customers to a hypothetical services bundle, we explored the data using noncausal statistical techniques including cluster and factor analysis. Cluster analysis classified cases into relatively homogeneous groups, using cluster variables based on behavioral characteristics and a combination of demographic information (average monthly mobile bill, gender, age, education, and income), personal characteristics, and derived indices for specific service needs. Derived indices were categorized prior to analysis. Hierarchical clustering (using Ward's method with squared Euclidean as the interval distance measure) identified four segments as broadly consistent with the data and compatible with theoretical considerations (Malhotra, Hall, Shaw, & Oppenheim, 2002).
Addressing the remaining piece of the puzzle involved factor analysis. Factor analysis explored how well the responses to the new mobile data services survey questions distil into bundles of services. These factors, embodying the essence of the survey questions, grouped by attribute, then provide us the means by which to map back to user types identified earlier, using one-way ANOVA and post hoc tests from SPSS. Factor analysis utilized the principal axis factoring method. The mean factor scores for entertainment, information, and functionality preferences were compared across segments to reveal significant differences. Three separate one-way analyses of variance isolated relationships among behavioral and demographic data and combinations of these and resulted in the relationships portrayed below.
Table 2 reveals that behavioral indices were no better across the board in their ability to identify meaningful segments than a traditional demographic approach. However, these results also suggest that the validity of segmentation can be improved by combining selected demographic and behavioral variables.
Across borders, the revealed needs of members of these segments were parallel, except that Malaysians in the Mobile Professional segment were significantly more interested in mobile payments than their Singapore counterparts, but less interested in messaging. Such differences may result from crossborder differences in economic geography and barriers to interoperability, respectively. Singaporeans in all segments were more interested in the functional service bundle, while Malaysians were more interested in the fun bundle, suggesting that Singapore might serve as a lead market for some, but not all, service bundles (Olsen et al., 2002).
 The subset of participants reporting MDS use totaled 149. Principal axis factoring was carried out, followed by varimax rotation with Kaiser normalization. Rotations converged in six iterations. The KMO measure of sampling adequacy and Bartlette's tests of sphericity provided support for the validity of factor analysis on the dataset. Varimax rotation facilitated interpretability. Initial runs showed, on the basis of a scree plot and eigenvalues, support for five factors, which explained 48% of the total variation.