The Eureka Project at Xerox
(From The Knowing Organization, Chapter 4)
The work of service technicians is vital to the photocopier business of Xerox. It is an important point of contact between the firm and its customers, and a primary means of maintaining customer satisfaction and brand loyalty. In the early 1980's, with a shortage of technicians, Xerox decided to use less skilled and experienced service staff by moving towards what was called "directive" repair and adjustment procedures. Service instructions would be documented in the form of a decision tree, and technicians need only be trained in using the documentation to be able to diagnose and repair machine failure.
A group in Xerox PARC (Palo Alto Research Center) had a background in artificial intelligence, as well as expertise in building programs to diagnose machine faults. In 1991, the group developed a model-based expert system covering one complex module of a particular photocopier that could support a technician in diagnosing repairing problems in this module. The system was shown to some technicians. Here's how the PARC group described the reaction:
"That's amazing," they said. "Would it really be useful if we had a complete model for the machine?" we asked proudly. "Not really -- though it is amazing - rather like a bear dancing. It is surprising to see it do it at all." (Bobrow and Whalen 2002, p. 49)
The research group was surprised by the unenthusiastic response and probed further. It found out that many products (e.g. those from Xerox's Japanese partners) did not have full descriptions of their operation. Diagnostic documentation was being produced by inducing faults in machines in a laboratory and then analyzing symptoms. However, the hardest problems in the field were new problems that were not covered by the documentation. This was understandable since photocopy machines operated in a broad range of complex environments. Extremes of temperature, humidity and dust; vibrations and network interactions; the age of the machine; all these could lead to new failure modes. Moreover, a fault may appear intermittently, making it hard to track down.
The PARC group decided to observe what service technicians actually did in the field by accompanying them on service calls. They saw that when technicians faced a new, undocumented problem, they might use their two-way radios to call on a buddy for ideas, or turn to the experts (former technicians now serving as field engineers) as part of the escalation process. When tough problems were solved, they would often tell stories about these successes when they met their peers in the cafe, the parts depot, or a workgroup meeting. Service technicians took pride in their work, especially in solving intractable problems that had stumped their peers. They enjoyed talking about solutions to hard problems, and gaining the respect and recognition of their peers.
Eureka in France
A member of the PARC group, a French national (Olivier Raiman), then spent time riding with French technicians to understand if their practices were similar to those in the US. One interesting observation was that many technicians carried cheat sheets of solutions their workgroup had invented to solve undocumented hard problems. New technicians would ask more experienced technicians for copies of these cheat sheets.
Through a series of workshops with technicians in France, the PARC research group concluded that a wider sharing of this knowledge across different workgroups would be valuable. The PARC group received backing from the French service organization - including management and expert field technicians ('tigers') - to try an experiment. An initial database of 100 to 200 tips was created by having the tigers edit and validate stories volunteered at workshops, as well as adding tips that the tigers themselves used. Tips were structured simply in terms of Symptom, Cause, Test, and Action. Access to the database was via a standard laptop running an easy-to-use program that included a basic search function. The experiment was a success, and it was decided to extend the use of technician-invented solutions to the entire French service force.
The community-based mode of knowledge sharing that was developed, called Eureka, would work as follows:
- Technicians who had found new ways of solving difficult repair problems would submit their solutions to a pending tips database.
- Pending Tips were vetted by validators -- senior field technicians who were recognized as experts for those particular products. "Validation isn't just a Yes or No, it's a conversation with somebody who understands the machine deeply, and this was a cool thing to do." (Tom Ruddy, KM director)
- After validation, a tip is entered formally into the 'knowledge base' where could be viewed by everyone who subscribed to the products for which the tips are applicable. When a technician viewed or used a tip, he was asked to vote on whether it helped. In this way, each tip developed a success rate for how many times that solution was applied and how many times it worked.
In 1995, the system was deployed with only three product databases. By the end of the first year, technicians had created over forty databases covering a wide range of products, and more than one new tip was being added each day. The service organization in France went from being an average or below average performer to being a benchmark performer: its service metrics were better than the European average by 5-20%, depending on the product (Bobrow and Whalen 2002).
Eureka in Canada
In June of 1996, the PARC group decided to introduce Eureka to another community. Canada was considered, partly because laptops had been deployed to all Canadian technicians, and partly because the Canadian service force was comparable in size to the French. The PARC group garnered support from one senior manager in Canada, and was able to team up with an experienced field engineer who would become a local champion for the development and deployment of Eureka. A server-client system was built, allowing a technician to use client software running on a laptop to access a local database that was synchronized with the community database on the server. Tip validation was done by product specialists in each customer business unit, similar to what was done in France. Unlike France, Canadian service management wanted to continue its financial incentive program for service suggestions, so technicians received a small financial reward for tips. Eureka was successfully launched for twenty products in early 1997. Extensive training of product specialists by the Canadian champion, with the specialists then training service technicians, took place over four months. After six months, Canadian Eureka had become accepted as the technicians' tool (Bobrow and Whalen 2002).
Eureka in United States
Eureka was rolled out in the United States in 1997 with a pilot program in several locations. Again, where these locations were supported were supported by local champions, the pilot took hold. In 1998, Eureka was distributed generally via CD-ROMs mailed to field managers who were then expected to pass them along to technicians in their workgroups. The PARC group did not favor this 'mass distribution' approach, but had difficulty persuading US management to adopt the 'participatory deployment' strategy that had worked well in France and Canada. Nevertheless, US technicians were accepting Eureka enthusiastically after they have learned the system. Bobrow and Whalen (2002) cite one technician's remarks as being typical: "In all my years in Xerox, the two best things ever given to us are the (two-way) radios and Eureka."
In early 2001, the tips database had grown considerably as the number of countries using Eureka increased. There were close to 50,000 technician-authored tips, and the number of problems solved using Eureka had increased to nearly 200,000 annually. Solving a problem with Eureka could mean saving several hours of down time, not having to escalate the problem to experts, or avoiding the replacement of a machine. Xerox reported that the use of Eureka provided many millions of dollars in savings annually, and had led to increases in both customer and employee satisfaction.
- What was the knowledge management problem to be solved?
- In Eureka, how was knowledge created, codified, and shared?
- What group and individual roles were instrumental?
- How did IT support the process?