Communications of the ACM
Data Completeness: a Key to Effective Net-Based Customer Service Systems
The Internet has created an environment of abundant consumer choice, and in this environment organizations seek to increase customer loyalty. Research suggests that as little as a 5% increase in retention can mean as much as a 95% boost in profit, and repeat customers generate over twice as much gross income as new customers [6]. Thus, it is imperative that organizations understand their customers' current behavior, preferences, and future needs. Many organizations believe one of the fundamental means for creating competitive advantage in this environment is to deploy IT that supports and fosters one-to-one relationships with customers [2]. This belief has helped make customer relationship management (CRM) one of the most significant current developments in the domain of enterprise software. CRM solutions are deemed to be so critical that investments in them continue to be funded, even in these days of shriveling IT budgets.
CRM is defined as a strategic posture that calls for iterative processes designed to turn customer data into customer relationships through active use of, and learning from, the information collected [4]. To implement CRM, organizations adopt a broad range of technology. Net-based customer service systems (NCSS) deliver service to a customer either directly (for example, via a browser, PDA, or cell phone) or indirectly (for example, via a service representative or agent accessing the system). The most visible instance of a NCSS is a Web site. For CRM, Web site personalization software provides navigational aids and keeps records of customers' past transactions. Internet-based recommender systems provide personal advice by using adaptive filtersboth content filters and collaborative filtersto deliver customized information and personal recommendations based on evaluations of users with similar interests [7]. NCSS are supported by database and data-mining tools used to analyze data to identify customer segments, match products to customer profiles, and better understand target demographics and psychographic characteristics. Customer service representatives may use "customer intelligence" to up-sell or cross-sell products and services.
CRM success is defined in three dimensions: increased profit, improved customer satisfaction, and enhanced customer loyalty. With such a broad range of outcomes, it is not surprising there are multiple orientations of CRM; different orientations have different approaches toward examining and leveraging customer data.
Profit-centric orientation to CRM focuses on capturing and analyzing historic transaction and preference data for existing customers. Data is primarily used to identify the most profitable customers who are expected to continue to contribute to the firm's bottom line in the future. No investment is made to retain unprofitable or marginally profitable customers.
Most organizations are still struggling to analyze the data they have and determine how to leverage it to increase profit, improve customer service, and build long-lasting relationships with customers. [F]irms need to start thinking more carefully about their CRM information processing strategy.
Customer-understanding orientation to CRM strives to understand the needs and preferences of current and potential customers and uses this information to better service them. This perspective allows the firm to benefit not only from those customers already providing revenue, but also includes customers who have future potential to do so.
Customer-relationship orientation focuses on managing individualized relationships with customers. The premise is that a relationship is something a customer values, and this relationship contributes to creating switching costs that aid in customer retention. The organization and the customer mutually benefit from the close-knit relationship that develops from the successful practice of CRM.
Thus, CRM is thought to be a strategy for delivering superior e-service, but, as we demonstrate in this article, success in this regard is founded on understanding the strategic fit between data and service. Fit is achieved when customer expectations, in terms of information-based services, are supported by the appropriate orientation to CRM; good fit is defined as data completenessa state where customers have access to all data they deem important to the information-based service in which they are involved. Lack of data completeness leads to a gap between what customers expect and the type of service they receivethe larger the gap the more dissatisfied the customer. As organizations move toward providing more and more information-based services over the Internet, it has never been more crucial that they narrow the data gap between themselves and their customers.
Managers that view CRM technology as a silver-bullet solution to improve customer relations are becoming increasingly frustrated by the results. Recent research suggests that 60% of managers will view their CRM implementations as failures. One primary reason is the lack of fit between CRM business processes and the capabilities of the CRM technology [1]. To examine technology and organizational challenges related to CRM, we conducted case studies in six organizations from four customer-intensive industries: two in finance, two in travel, one in transport, and one professional service. All six sites were large North American organizations that adopted CRM technology to enhance customer service over the Internet. Five semi-structured interviews were conducted in each participating organization. Participants represented three organizational levels: two marketing executives, two customer-service representatives, and a database manager. Based on insights from our investigation, we conclude that organizations generally fail to support their CRM orientation with complete data. Four principal CRM information processing strategies are expected to help guide organizations toward deriving more benefit from CRM investments.
CRM Information Processing Strategies
Every customer creates uncertainty on three levels for organizations: uncertainty about past behavior, current preferences, and future needs. Because organizations tend to increase information processing capabilities to manage increasing uncertainty [5], different firms implement different information processing strategies depending upon the type(s) of customer uncertainty they choose to address. The table on the preceding page displays four information processing strategies that differ by underlying CRM orientation, desired outcome, and type of uncertainty.
Transaction strategy. Data related to customers' current behavior is analyzed to identify profit concerns (for example, overdue accounts) and used to answer customer inquiries related to account transactions.
Data strategy. Historical data is captured and used to service customers by providing access to a detailed record of past transactions. Historical data is also processed to identify the most profitable customers expected to continue to contribute to the firm's performance. CRM analysis tools are used to segment customers based on their current and potential profitability, defined by many organizations as customer value.
Inference strategy. In line with understanding the customer, an inference strategy processes historical and revealed preference data to understand current and potential customers. By carefully analyzing this data, an organization assumes it can infer future customer needs based on past behavior and preferences. Inferring future behavior is risky and should be confined to select environments. We suggest it is more appropriate for organizations that deal with products where there is little variation between past and future behavioral patterns, and organizations that sell products often linked to previously purchased items or services. For example, a telecommunications firm might adopt an inference strategy, as long-distance calling patterns tend to be consistent (such as calling friends and family). Clothing retailers might infer a customer's future needs based on a historical analysis of past purchases and stated preferences for color and size.
Advice strategy. Using a customer relationship perspective, a firm adopts an advice strategy to provide accurate and reliable advice to customers based on knowledge of their future needs. The goal is to develop a long-term customer relationship. An advice strategy is appropriate for marketing products or services where expertise or trust is highly critical to the customer because this is what customers need and value. In the computer hardware business, for example, customers depend on vendors to provide advice to ensure that purchased products are compatible with existing equipment and readily upgradeable to meet foreseeable needs.
CRM is thought to be a strategy for delivering superior e-service, but success in this regard is founded on understanding the strategic fit between data and service.
Implementing a Complete Information Processing Strategy
Successful implementation of an information processing strategy requires that organizations have a data repository that provides an appropriate customer view for their strategy. All organizations begin with a functional viewa transactional database that stores current customer data related to a specific functional unit (for example, current accounts receivable). This is sufficient to support a transaction strategy.
As illustrated by the arrows in the figure, the next step to enhancing customer understanding and relationships is to develop a firmwide, cross-functional view of the customer to support a data strategy. To enhance customer profitability, organizations need to have a complete record of all customer behavior across all functions. For example, a financial institution needs to capture customer behavior pertaining to credit cards, loans, and savings accounts to accurately determine the overall profitability of each customer. Most organizations participating in our research had developed a repository of past transactions across the organization by implementing a customer-oriented enterprise data warehouse.
Customer service representatives consistently stated that more customers are expecting organizations to keep historical records of their personal transactions and provide Internet access to those records. Hence, organizations that have not integrated functional databases across the firm may experience a data gap. The lack of an accessible complete customer transaction history creates a gap between what data customers expect and what data firms can deliver. For example, customers might expect a retail Web site to have a record of their past purchases so that they can reorder shirts of the same neck size and sleeve length. If this data is not available, then the level of customer service is less than what today's customers expect.
We also found that organizations desired to move beyond a profit orientation toward enhancing customer understanding and customer relationships. Moving beyond a firm view of the customer is expensive and complex. Hence, we suggest organizations consider varying their information processing strategy by customer segment. For example, firms may invest in implementing an inference strategy for more valuable customers and a transaction strategy for less valued customers.
As illustrated in the figure, we also recommend that organizations think about inference and advice strategies as separate initiatives. The bottom arrow in the figure illustrates the approach to enhance understanding of customers. Firms that seek to adopt an inference strategy but only have a firm view of the customer risk experiencing an inference gap; they lack a multibrand view of the customer that captures historical data about customer behavior and preferences across multiple organizations in an industry. For example, a multibrand view might integrate data across several retailers (for example, Parisian, The Gap, and Ann Taylor). Parisian's view of a customer, for example, provides an incomplete picture of retail behavior and preferences because it does not record purchases in competing firms in the clothing industry. Incomplete data can lead to incorrect inferences about a customer. For example, a mutual fund company analyzing a customer's holdings might determine the person is overinvested in Latin America and suggest a reallocation of assets. Such an inference is likely to be erroneous if the customer's investments are distributed, and unevenly balanced, across several mutual funds companies.
Technology that supports the development of a multibrand view is less advanced. A large bank in our study used screen-scraping software to provide its customers with account aggregation capability through a Web-based view of financial products held across multiple firms. We expect third-party travel firms (for example, Travelocity) and online mass merchandisers (for example, Amazon) will be most effective at implementing an inference strategy, because they are already collecting data across multiple brands and products. Amazon.com's recommender system sets it apart in terms of advances in developing an inference strategy; it collects data related to multiple brands and products and uses adaptive filters to identify and capitalize on up-sell and cross-sell opportunities [7]. Amazon's newest effort, known internally as Ruby, is a new Internet apparel and accessories store with over 400 major clothing brands. By providing users with easy navigation and a single shopping cart, Ruby also allows Amazon to develop a more complete multibrand view of its retail clothing customers. We expect its success will be further enhanced with more initiatives to capture "why" data, such as the feature that asks customers to note if the good is a gift or a personal purchase. We concluded from our research that without data about why a product was purchased, it is difficult to make accurate inferences about future needs. Organizations that achieve an accurate multibrand view of the customer will be more effective at inferring future needs and better able to service customers due to an enhanced customer understanding.
The top arrow (in the figure) illustrates the approach to enhance a better relationship with customers. Firms that seek to adopt an advice strategy but do not have a future outlook view of the customer are at risk of experiencing an advice gap; they lack knowledge of a customer's future needs required to provide accurate and useful advice. It is very difficult, if even possible, for a firm to accurately predict a life-changing event from analysis of prior transactions. For example, the forthcoming birth of a child or a promotion that may lead to a move are certainly likely to change a person's future transactions. However, unless the firm captures details of this approaching event, the firm is unable to provide appropriate counsel. The firm's ability to provide relevant advice will be critical to enhancing the relationship between the firm and the customer.
Most advice tools available on current company Web sites are relatively simple. Tools such as hotel finders, retirement planners, and insurance calculators can provide an initial level of customer support. More complex tools require the ability to incorporate future outlook data. An early example is SmarterKids.com's system, which makes product recommendations for parents based on surveys they fill out or on test results that provide data about their children's learning styles. The company stores the rich profiles in its data warehouse and combines these with historical data to make product recommendations [3]. Organizations that achieve an accurate future outlook view of the customer will be more effective at providing valuable advice that will enhance the relationship with their customers.
Conclusion
Today, most organizations are still struggling to analyze the data they have and determine how to leverage it to increase profit, improve customer service, and build long-lasting relationships with customers. In order to move in the right direction, firms need to start thinking more carefully about their CRM information processing strategy. We propose the extent to which a firm can provide outstanding e-service is a function of its ability to achieve the appropriate level of data completeness. The greater the level of data completeness, the higher the likelihood that customers will feel their needs are understood and/or will want to build a relationship with the company. Organizations are encouraged to examine different CRM information processing strategies for customer segments across both general and corporate customers and identify where narrowing the gap will generate the highest return.
References
1. Cholewka, K. CRM: The failures are your fault. Sales & Marketing Management (2002), 2324.
2. Shoemaker, M.E. A framework for examining IT-enabled market relationships. J. Personal Selling & Sales Management 21, 2 (2001), 177186.
3. Sliwa, C. Web sites strive to personalize. Computerworld 54 (Apr. 17, 2000), 103.
4. Swift, R.S. Accelerating Customer Relationships Using CRM and Relationship Technologies. Prentice Hall, Upper Saddle River, NJ, 2001.
5. Tushman, M. and Nadler, D. Information processing as an integrating concept in organizational design. Academy of Management Review (July 1978), 613624.
6. Winer, R.S. A framework for customer relationship management. California Management Review 43, 4 (2001), 89106.
7. Zhang, Y. and Im, I. Recommender systems: A framework and research issues. In Proceedings of Americas Conference on Information Systems, 2002.
Footnotes
Teradata, a division of NCR, and the Marketing Science Institute (MSI) provided funding for the project.
©2003 ACM 0002-0782/03/0600 $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 © 2003 ACM, Inc.