Communications of the ACM
Communications of the ACM
Consumer Support Systems
Consumer support systems (CSS) are proposed as multilevel interorganizational systems to support business-consumer relationships. They utilize decision support technologies and adapt them to support consumer decision making. They also form interorganizational structures called "demand chains" to channel the flow of information between businesses and consumers. The proposed architecture suggests a new paradigm for business-consumer relationship, and may create a significantly more efficient marketplace due to highly structured information exchange between businesses and consumers.
Decision support systems are human-machine systems used widely to support organizational decision-making processes. They aid decision makers by extracting and filtering data from organizational databases, by processing it to make it directly relevant to specific decision makers, and by creating customized user interfaces to make the information easily usable by each decision maker [2, 6]. Decision support technology is commonly available to organizational decision makers, yet rarely deployed to support consumer decision making. There is considerable potential in providing this technology to consumers, both in creating new business opportunities, and also in strengthening the relationship between existing businesses and their customers. However, there are also both technical and organizational hurdles to the deployment of this technology for consumers, explaining the heretofore failure of the markets to develop such comprehensive consumer support systems.
Instead of comprehensive support systems, consumers are currently served by three distinct technologies, but each has serious shortcomings:
Consumer support is often provided by the customer relationship management (CRM) systems of businesses [3]. CRM is a vital and growing part of modern marketing, but the support it provides to consumer decisions is very limited. CRM systems belong to individual businesses, and consequently they have difficulty collecting consumer information, especially about consumers' needs and preferences, and certainly about their transactions with other businesses. Additionally, CRM systems also have difficulty delivering product information to consumers. Since they are owned by individual businesses, they have neither the resources nor the credibility to deliver comparative multi-vendor multi-product information, and provide comparisons and trade- offs. Dell Computers, for example, has an extensive CRM system that tracks its customers and its transactions and helps customers configure computers to fit their needs. However, Dell is unable to provide credible comparisons to the products of other computer manufacturers, and unable to collect detailed consumer information about their future needs and their past transactions with other vendors. Dell's interaction with its customers is limited to the context of conducting a transaction. It does not exist before the transaction, and it ceases to exist soon after the transaction when the service agreements expire.
Recommendation systems attempt to alleviate the problems associated with CRM systems by collecting multi-vendor product information, by surveying the consumers about their product preferences, and by recording their past transactions related to those products [7]. But they also have a number of shortcomings. They primarily rely on past transactions, and hence require a number of transactions before they can be useful. They survey the consumers for demographic information, but they do not collect information about consumers' intentions and tasks, and hence they often fail when consumers' intentions change. For example, Amazon's book recommendation system is unable to distinguish among one-time gift purchases, professional purchases made on behalf of others, sporadic special-occasion purchases, and personal purchases. Consequently, a gift purchase made for a child will lead to continuing recommendations for children's books irrespective of the consumer's current needs. Moreover, recommendation systems fail to explain their recommendations, and reveal the reasons for their choices except in terms of statistical correlations.
Intelligent agents are personal software products designed to alleviate some of these problems by collecting detailed personal information about individual consumers' needs, preferences, and past transactions, but they also have major shortcomings [11]. Intelligent agents belong to and serve individual consumers, and as such they resolve some of the privacy and disclosure issues plaguing the other solutions. Yet, because they belong to individual consumers, they do not have access to comprehensive product databases to draw on, and they need to crawl the Web through unstructured Web pages, or access multiple heterogeneous business databases in search of data. This search process is slow and inefficient, and locating and integrating some data is not even feasible.
Dell's interaction with its customers is limited to the context of conducting a transaction. It does not exist before the transaction, and it ceases to exist soon after the transaction when the service agreements expire.
Moreover, there are no general models of consumer decision making, and hence intelligent agents are often task-specific, and they need to be customized to capture the needs and preferences of specific consumers. Search agents, negotiation agents, and advice agents all perform different parts of the consumer support task, and they often require customized software to fit the needs of specific consumers. Finally, there is no mechanism to systematically capture and utilize past experiences, especially the experiences of other consumers, through a learning process. Merely recording and analyzing past transactions may not be sufficient, especially because intelligent agents owned by individual consumers do not have access to other consumers' transactions.
Some of these problems are the result of fundamental shortcomings of the currently available technologies of data management, artificial intelligence, and statistical analysis, and purely technological solutions may not be adequate to resolve them. Consumption may well be as complex a process as production, and may require the same degree of support from systems, technologies, and organizational structures as production does. Surveys and statistical analysis of past transactions may not be sufficient to understand and manage consumption, as they would not be sufficient to understand and manage production.
Advertising and promotion campaigns of businesses may not be sufficient to inform and educate the consumers, as they would not be sufficient to inform and educate the suppliers and trading partners. Modern supply chains of businesses have more cooperative and more complex sets of technological and structural arrangements, and similar arrangements involving both technological and structural components may be necessary to manage consumption. A demand chain may be necessary to manage the downstream relationship with consumers, as a supply chain is necessary to manage the upstream relationship with suppliers.
Requirements for a Comprehensive CSS
An effective mechanism is needed to deliver information to consumers. The information must be structured and easily searchable, not just a collection of Web pages or product catalogs; it must be comprehensive and comparative, not limited to a single vendor; it needs to be objective and reliable, not advertisement; and it needs to be customizable to each consumer, not mass distribution.
Dell Computers provides one of the most effective support systems to its customers, with detailed side-by-side comparisons of its computers, but it fails most of these criteria. It fails to provide comparisons to the products of other vendors, such as HP and Gateway; it fails to compare products from one class to another, such as desktops versus notebooks; and it fails to provide an effective search mechanism through the large number of possible product configurations. Humans have great difficulty doing multi-attribute comparisons among dozens of products at a time [6].
Compiling detailed product attributes from multiple vendors would certainly be useful, but consumer support requires more. Consumers do not start the market search process with a set of product attributes in mind, but usually with a collection of tasks to be performed. Identifying the product attributes necessary to accomplish a task is actually the first step of the search process, and it must be supported. A consumer does not start the search for a car with a certain horse- power engine and a certain fuel efficiency, but with a task of, say, commuting 20 miles a day to work, or transporting kids to their soccer games, or some combination of such tasks. Discovering the product attributes from task characteristics is complex, and needs effective support [10].
The starting point for a market search may not even be a task, since consumers may not be aware of the existence and relevance of tasks, and they may need to be alerted. Such alerts and triggers should be based on consumer attributes, and utilize information from past experiences of similar consumers. The starting point for a search then is often a collection of consumer attributes that lead to a set of relevant tasks, which in turn lead to desirable product attributes, and finally one can search for the products exhibiting those attributes. A university professor, for example, probably needs to attend an annual conference in his field, but he must be aware of the existence and importance of such a conference as a relevant task. Once a conference is identified, a search is necessary for all the products needed to attend the conference, ranging from flight and hotel reservations, conference registration, car rentals, and restaurant reservations.
Recommendations must be explained beyond mere statistical correlations. Such explanations should be in terms relevant to the consumers' world rather than the vendors' product attributes. A computer needs to be described not in terms of its memory size and CPU speed, but in terms of its usefulness in downloading music or executing a Web search, or in terms of its relevance to a geologist or a carpenter.
An Architecture of Technology and Organization
Any effort to incorporate consumers into business supply chains would require organizing consumers into more abstract entities similar to a business organization. These entities must be supported with information systems, which in turn could be linked to the systems of other supply chain members, creating the cooperative system-to-system linkages that characterize modern supply chains. We will refer to these extensions to the supply chains involving consumer organizations as demand chains, and refer to the systems built to support these chains as CSS.
CSS require three distinct sets of information: products, tasks, and consumers, as well as linkages among them. To support these three sets of information and their processing, a variety of technologies must be deployed; and a complex organizational structure may be necessary to support the flow of information.
Extensive product information should be collected from a variety of vendors and structured for comparative analysis. Data warehousing technology is the critical technology for drawing data from multiple vendor databases, and reorganizing it for efficient analysis [3]. However, CSS would require an interorganizational data warehousing project drawing data from multiple competing vendors. As such, it could not be credibly developed by a single vendor, but it would require an independent marketplace to collect and restructure the data. Moreover, product comparisons would require extensive technical data characterizing the products, often more than what vendors typically disclose on their Web sites.
An independent marketplace must accumulate considerable market power to pressure the vendors to disclose such detailed product data. Some product attributes are not objectively measurable, and require qualitative judgment. These attributes must be collected from independent evaluators, or aggregated from consumer feedback, and that would require the participation of third-party evaluators, government agencies, and consumers. There are some portal Web sites now that attempt to play this role, and present side-by-side product comparisons. Carpoint.com is one such Web site that contains considerable publicly available technical data about cars from all major manufacturers, in addition to third-party evaluations and ratings, and consumer feedback.
Product information about dozens of competing products involving hundreds of technical attributes is difficult for consumers to process in side-by-side comparisons. Decision support technology is critical to process such information, to aggregate and summarize it into a hierarchical drill-down structure, and present it in a customized user interface [6]. However, customization requires extensive consumer information, and consumers may not be willing to disclose such extensive personal information even to independent markets. Moreover, the process of customization to serve the decision processes of individual customers is not well understood, and may not be completely automatable. Business organizations certainly do not automate decision support completely, but use technology to aid an extensive organizational structure that processes information and feeds it to decision makers. A similar human-machine system is necessary to aid consumer decisions.
The tasks that consumers perform can be used to form new intermediaries to link consumers to products. Tasks can play a significant intermediary role both to summarize and aggregate extensive product information, and also to filter and reduce the information consumers have to disclose. Tasks can be used to isolate consumers from vendors so the consumer information can be kept private, and disclosed only to the extent they relate to a task. Tasks can also be used to reduce the information load on consumers since not every product has to be considered by every consumer, but tasks filter and aggregate product information. However, tasks are independent of consumers and products, and task information is specialized and possessed by professionals and craftsmen who practice the task.
If tasks are to be formalized and serve as intermediaries between products and consumers, these task professionals must be organized to collect task information, and to evaluate products in terms of their relevance and effectiveness for specific tasks.
Some task-oriented organizations exist now, including travel agents that aggregate and evaluate products for cruises and vacation packages; conference and convention organizers that aggregate hotel, transportation, speaker, and food products; and real estate agencies that aggregate mortgage, insurance, legal, and home assessment products to serve the single task of home purchase.
Large-scale implementation of such task-centered services would require a technology to automatically draw data from a multitude of vendors. The critical enabling technology is known as Web Services and it allows businesses to post their product and process data dynamically over the Web, which can be collected dynamically by other businesses such as task specialists, and aggregated and processed directly to serve various tasks [4, 9, 10].
Task specialists are a new breed of intermediaries, and Web Services is the technology that encourages the development of such intermediaries by automating the posting, collection, and aggregation of dynamic and heterogeneous data. Existing task specialists largely operate with manually collected data, and have difficulty maintaining their knowledge bases. The new technology of Web Services is likely to encourage the expansion of such services. Task information is closer to the consumers' world than product information, and it can simplify the consumers' decision processes. For example, a consumer can search for appropriate cars for specific tasks, such as urban commuting, or all-terrain driving for wilderness camping, as opposed to searching for a car with a certain amount of horsepower or gear ratio.
We have now extended the demand chain from products and product vendors to tasks and task services; this is, intermediaries that come between the product vendors and consumers in the demand chain. Consequently, we have replaced the product search problem for consumers with a task search problem.
Tasks are complex entities and searching for a task still requires a significant effort, especially because consumers may not know task attributes, or even be aware of the existence of relevant tasks. A university professor may not know what attributes make a conference relevant to him, or may not even know that attending annual conferences is a relevant and useful task for university professors. Moreover, tasks must often be customized for individual consumers, and task specialists would find it difficult to customize tasks without additional technology and organization.
The technology critical to organizing consumers and linking them to tasks is "object clustering and typing," and the necessary new intermediaries are "consumer communities" [1, 5]. Consumers are organized into a hierarchy of communities in terms of their personal characteristics, and each community collectively relates to a collection of relevant tasks and exchanges information with those tasks.
A community can be an independent business, or a cooperative that serves a distinct group of consumers by identifying the relevant tasks, classifying, customizing, and evaluating them, and presenting them to its members. It also collects extensive consumer information from its members, both to characterize and classify its membership, and also to customize tasks for each subclass. A community of technology professors, for example, would be a subclass of the community of professors; it would inherit information about tasks relevant to all professors from its super-class of all professors; but it would also identify tasks relevant specifically to technology professors, and even further classify them with respect to academic specialties of its members.
Communities are the last intermediary in the demand chain, linking consumers to task services, and eventually to products and product vendors. But, as promising as they are, communities can only be effective in the context of a complex demand chain deploying a variety of technologies and organizational structures. On the one hand, they require extensive product and task information that should be available dynamically, and on the other hand, they require access to extensive consumer information that must be isolated and protected for privacy and competitive reasons. That kind of bidirectional information flow with extensive processing while protecting the privacy of the parties requires extensive intermediation as described.
The three-level demand chain is now ready to be connected and integrated as shown on Table 2. The technology of integration is similar to supply chain integration utilizing Electronic Data Interchange and business-to-business Web connections [8]. Information flows from consumers to their communities, from communities to the relevant tasks, and from tasks to the products involved. At each step, the information is aggregated, and detailed individual information is protected.
Product information flows in the opposite direction: from vendors to task services to be aggregated into packages; from tasks to communities to be ranked and evaluated and recommended to the community members. At each step, the detailed product information is processed and aggregated, and the consumer is protected from the minute technical details of product attributes. In this environment, consumers think in terms of their membership in various communities based on their personal characteristics; communities think in terms of tasks that have to be performed by their members as a result of their membership in those communities; task services think in terms of the products necessary to perform those tasks and selecting the most appropriate products. At each step, the information is processed and analyzed with the help of appropriate technologies. Neither technologies nor the organization of intermediaries alone is likely to be sufficient for success.
When consumer characteristics are readily available, and volunteered by consumer communities, the dominant marketing paradigm is reversed, and focuses on identifying relevant transactions from consumer characteristics.
This process is likely to significantly change the current marketing paradigm that focuses most attention on characterizing consumers by using their past transactions. When consumer characteristics are readily available, and volunteered by consumer communities, the dominant marketing paradigm is reversed, and focuses on identifying relevant transactions from consumer characteristics. Intermediation by task specialists facilitates this new paradigm as it serves the dual purpose of channeling product information to the right communities and protecting the consumer communities from flooding with irrelevant product information. They not only protect consumers from irrelevant information, but also serve businesses by channeling their products to precisely the right consumers. Such an efficient information exchange has the potential to benefit all, and create a significantly more efficient marketplace. In this marketplace, the information exchange between businesses and consumers would be significantly different, since businesses and consumers would not interact directly, but through intermediaries such as consumer communities and task services. Customer relationship in this marketplace would resemble the business-to-business relationships of supply chain members, rather than transaction data collection, statistical data analysis, and direct marketing and mass media campaigns prevalent today.
Much of the literature on supply chains would be immediately relevant to demand chains, and the combination of supply and demand chains would stimulate a comprehensive research field combining and cross fertilizing the supply chain field with the customer relationship field.
References
1. Armstrong, A. and Hagel, J. The real value of online communities. Harvard Business Review (May 1996), 134141.
2. Berndt, D. Consumer decision support systems: A health care case study. In Proceedings of HICSS Conference on Systems Sciences 34 (2001), 110.
3. Brohman, M.K. et al. Data completeness: A key to effective Net-based customer service systems. Commun. ACM 46, 6 (June 2003), 4751.
4. Hagel, J. Out of the Box: Strategies for Web Services. Harvard Business School Press 2002.
5. Lechner, U. and Hummel, J. Business models and system architectures of communities. Intern. J. of Electronic Commerce 6, 3 (2002), 4182.
6. Marakas, G. Decision Support Systems. Prentice Hall 2000.
7. O'Mahoney, M. et al. Collaborative recommendation: A robustness analysis. ACM Trans. Internet Technology 44, 11 (2004), 344377.
8. Orman, L. Optimum design of electronic communities as economic entities. Information Systems Frontiers 8, 3 (2006), 179194.
9. Rust, R.T. and Kannan, P.K. E-service: A new paradigm for business in the electronic environment. Commun. ACM 46, 6 (June 2003). 3642.
10. Sawhney, M., Balasubramanian, S. and Krishnan, V.V. Creating growth with services. Sloan Management Review 45, 2 (2004), 3343.
11. Wagner, C. and Turban, E. Are intelligent e-commerce agents partners or predators? Commun. ACM 45, 5 (May 2002), 8490.
12. Womack, J.P. and Jones, D.T. Lean consumption. Harvard Business Review 83, 3 (2005).
Tables
Table 1. Summary of the capabilities and shortcomings of three common approaches to consumer support.
Table 2. The three-level demand chain including the two new intermediaries and the information flows along the chain for the specific example of getting a hotel reservation to attend a conference.
©2007 ACM 0001-0782/07/0400 $5.00
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.
The Digital Library is published by the Association for Computing Machinery. Copyright © 2007 ACM, Inc.
No entries found