Cliches about customer knowledge abound. Know your customer. Fire your worst customers. Have a relationship with your customer. Market one-to-one. Virtually every manager would agree that more and better customer knowledge can bring economic benefits to a company. Yet after participating in two recent studies on customer knowledge management (sponsored by the Concours Group and the American Productivity and Quality Center) I've concluded that the state of the art in the field of customer knowledge is fairly artless. Certainly, some companies are doing impressive things with customer data. A few even manage to turn some of it into knowledge. But I can't point to any single company and say that it's got it right.\n\nOne difficulty is that customer knowledge is widely dispersed around a company. Many different business functions face off to customers, including marketing, sales, service, logistics and even financial functions. (IS, of course, doesn't typically interact with customers, which may also be part of the problem.) Each function usually has its own interests regarding customer information, its own way of recording what it learns and perhaps even its own customer information system. The disparate interests of departments make it difficult to pull together customer knowledge in one common format and place.\n\nCustomer information and knowledge also inspire a high level of politics and passion. If knowledge is power, customer knowledge is high-octane power; it is unlikely to be shared without reluctance. The salesperson with valuable customer information on index cards in his trunk or laptop, the service department with valuable information on what customers think about new products, the marketing organization with highly detailed customer attitudes and behavior from focus groups and surveys--all have some reason to keep control of what they know about customers. Senior general managers, however, generally prefer to make customer knowledge an organizational resource. And therein lies the conflict.\n\nAnother factor making the management of customer knowledge difficult is the fact that there are several different types, each of which must be managed with a different approach. The first type is data-derived customer knowledge that originates in transaction systems. We typically think of this type of knowledge as involving consumers, but it can also be about business customers. Managing this stuff involves several S's:\n- Strategy. Defining what information is really important and what customer behavior really counts.- Standards. Ensuring that "customer" and other related terms mean the same thing throughout the organization.-Systems. Allotting sufficient processing power to chew through all the data.-Statistics. Turning data into knowledge through statistical processing (although a neural network sounds sexier than statistics, that is really all it is).-Smart people. Finding smart people to structure and interpret the analysis of customer data, regardless of what the data mining theory says.\nSome companies are starting to get quite good at turning customer data into knowledge. MasterCard International Inc., for example, does it on a regular basis for member banks that want to understand the behavior of their credit card customers. FedEx uses it to increase market share and profitability among small shippers. Several regional Bell firms, including US West and the late Nynex, have used it to try to become customer-oriented before deregulation hits. The impetus to do so usually comes from marketing, but IT people are inevitably involved as partners. Companies that are successful at managing this type of customer knowledge put considerable resources into the problem. MasterCard, for example, has a group of 28 people who help spin data into knowledge.\n\nBut those who are good at data-derived knowledge are not always particularly outstanding at managing the second type. I call it human customer knowledge because it typically comes from interaction among people. It includes experiential observations, comments, lessons learned, conclusions and qualitative facts--e.g., organization charts--about the customer. This type of knowledge generally pertains to business customers or high-value consumers, though sometimes it is also useful to capture human knowledge from large numbers of consumers.\n\nOne good example of managing consumer human knowledge comes from General Electric Co.'s appliance division, where the GE Answer Center has provided service to GE appliance customers since the early 1980s (and frequently, though not intentionally, to customers of other GE divisions). In the late '80s, the GE Answer Center also became the "GE Appliance Customer Knowledge Center," although it was never given that name. Each service representative is supposed to--and usually does--record customer comments about appliances, dealers, service or other relevant issues in a central system. That's a good idea by itself, but what GE does with the data is an even better idea. It's reviewed by marketers, dealer relationship managers, customer service managers and, perhaps most important, manufacturing and new product development managers. The service-based customer knowledge is also used to rapidly identify appliance problems in the field (e.g., an improper installation approach) that can be addressed with product or documentation changes, saving vast sums in field dispatch costs.\n\nHow, in general, should companies go about managing human customer knowledge? Systems are important for capture and distribution, but their creation doesn't have to be difficult; GE's, for example, is an old, traditional database. What is difficult is motivating employees to record, share and use the knowledge. This won't happen by accident unless you have an extremely pleasant corporate culture and highly motivated people. Instead of counting on these things (I've been in your company and I know it isn't that wonderful!) it's better to mandate or strongly encourage knowledge behavior by evaluating people on the basis of it and rewarding those who display it. At GE, for example, participation in the customer knowledge system is an integral part of performance evaluations and bonuses. Also, as with data-derived knowledge, good management of human customer knowledge doesn't come without considerable human resources. The human knowledge has to be pruned, summarized, edited, put in context and otherwise massaged by smart human analysts. Perhaps someday we'll have machines that can do that, but I wouldn't suspend respiratory processes while you wait.\n\nThe third type of customer knowledge is tacit--unstructured, difficult-to-express knowledge that we observe or sense about our customers. We have a bit of difficulty with this as Americans; tacit knowledge feels squishy, and it often is. But the voice of the market never speaks clearly, and we often have to intuit messages from customers at the subrational level. Every good salesperson tries to elicit some tacit knowledge from a customer in the form of body language, facial expressions or other "vibes." Some market research experts now argue that customer opinions about products and marketing messages can best be understood in tacit forms of expression--say, how far the customer twists a dial--rather than explicitly stated verbal arguments. Tacit customer knowledge can be just as important to sales and marketing functions as the other types of customer knowledge.\n\nThe primary home of tacit customer knowledge management, however, is product development. And more specifically in Japanese product development. Ikujiro Nonaka, the UC Berkeley "knowledge professor," has been writing and speaking for a decade about how large, export-oriented Japanese companies turn tacit, metaphorical ideas about what customers want into successful products. Honda and Lexus, a division of Toyota Motor Sales U.S.A. Inc., for example, turned the shorthand phrases "Let's gamble" and "Western luxury," respectively, into highly successful new cars. Meanwhile, Sharp translated "Don't imitate" into a series of successful optoelectronic products. This concept is not exclusively Japanese; several U.S. automakers have also employed obscure-sounding metaphors for customer needs in designing new cars.\n\nThe good news about tacit knowledge is that much of it doesn't have to stay tacit forever; it can be converted through various means to explicit human knowledge and thus made more permanent and transferable. The bad news is that doing so is difficult. It requires continual observation, careful analysis of customer behavior and a patient disposition.\n\nCustomer knowledge isn't a highfalutin abstraction. When managed well and applied to various customer-facing business processes, it can increase customer purchase and retention levels, save money by directing marketing efforts at customers who will respond to them, and yield products and services that customers really want in the first place. If you're just getting started in knowledge management and wonder where to focus your efforts, don't overlook knowledge about customers--the business resource that keeps you in business.\n\nThomas H. Davenport is a professor and director of the Information Management Program at the University of Texas at Austin. He welcomes reader comments at firstname.lastname@example.org.