|Montague Institute l Contents l Index l Digest l Courses l Calendar l Subscribe
Taxonomies in action at Unilever Research
by Adrian Dale, with commentary by Jean Graef
The following article, reprinted here with permission of the author, originally appeared in the Knowledge Management Review. In this "point-counterpoint" format, Montague Institute founder Jean Graef adds her comments on some of the issues raised by Mr. Dale. The text of his article is in black; Ms. Graef's comments are in blue.
Unilever operates in a diverse range of fast moving consumer goods markets. Its products span ice cream to tomato sauces and shampoo to washing powder. Its scientists work in diverse fields from anthropology to mathematics with every physical and life science in between. Building durable company taxonomies across these ranges has been impossible and will probably always be so. Instead, Unilever Research has taken a pragmatic "just-in-time" approach - creating taxonomies quickly and dynamically for a specific task and then throwing them away. To do this, mind mapping and creativity tools have been deployed in facilitated sessions to develop a shared language for the worlds in which research teams operate.
Purists versus Pragmatists
From the purist's point of view, this must surely be a problem. How can you structure the knowledge base of the organization without a taxonomy? How can you identify the gaps without a structure? How can you resolve ambiguity or redundancy of language without a company thesaurus to guide you?
To the pragmatist these questions simply don't matter and in our organization most Managers are pragmatists. They have long recognized that consumer goods innovation rarely emerges from deeper understanding of narrower fields. Innovation happens at the boundaries of disciplines or is inspired by research in completely unrelated areas. A deep taxonomy of a narrow field is of little practical value.
Dale touches on three interesting issues here:
1. "Purists vs. pragmatists." It's useful to recognize that, just as people learn in different ways, they also have different ways of dealing with information. In her 1993 study of team information sharing at H-P Labs, Vicki O'Day lists at least four kinds of behaviors - "purists vs. proliferators, semanticists vs. syntactics, scruffies vs. neatniks, and savers vs. deleters." (See "Where did you put it? Issues in the design of group memory.") As the team members worked to create a "group memory," they gradually changed their working style to reflect compromises among the different behaviors, and eventually decided to establish a "curator" role.
2. "To the pragmatist, these questions simply don't matter." Unfortunately, this is true for many managers and teams, pressured as they are to meet stringent deadlines and short term profit goals. But the whole point of managing knowledge is to leverage know-how across the enterprise. Should we abandon this goal, even though in practice there are many barriers to achieving it? If there is no reciprocity among teams and no norms for information sharing, the "hunters and foragers" will find slim pickings.
3. "Innovation happens at the boundaries of disciplines." Many observers of the innovation process would agree (see our review of Dorothy Leonard's Wellsprings of Knowledge) with the importance of "boundary spanning" roles, but does this necessarily mean that a deep taxonomy of a narrow field is of little practical value? A back-of-the-book index, with its hierarchical relationships and cross-references, covers a narrow topic and obviously adds value. At least the cost of producing it can be amortized over the number of copies purchased.
Maps for the hunter-gatherer
The modern Research Manager does not need detailed maps of the knowledge domains where their teams already live and work. These hunting grounds are well known and may indeed be exhausted. They do need high level maps of the likely external hunting grounds for new knowledge. Often these new grounds are well outside their normal experience and that of their teams. Getting there will be a daunting prospect and much effort is needed to give the teams the language and experience they need to operate there successfully in a short space of time.
For the hunting and foraging teams on the ground, there is also a need for tools that allow them visualize and locate niche knowledge areas with potentially rich pickings. These will probably not show up on the maps used by their managers and if they did they might initially appear to be barren or blind alleys.
In our experience, "hunting and foraging" can be very productive - not to mention fun. It can also be a time-consuming, frustrating, hit-or-miss proposition. If you need to reuse information you have previously gathered or if you want other teams to be able to use it, a little investment in "farming" can pay off handsomely. Furthermore, we think there is a great value in giving colleagues a tool that allows them to grasp your view of the world. A taxonomy is a table of contents of your knowledge, skills, and contact network. (See "Your taxonomy is your future.")
Although we agree that it's imprudent to ignore external information sources, we think many businesses overestimate their value. In the first place, everyone else has access to the same information. In the second place, customers and suppliers - people with first-hand knowledge of the products and their applications - are a mother lode of ideas and insights.
Mapping is not an individual activity
Individuals are always constrained in their thinking when mapping alone. The really valuable Knuggets  that drive innovation do not then tend to surface as they live outside or between the formal categories and definitions. More than one mind is needed to create the rich maps and structures that are needed.
Facilitated teams are the key enabler
Each mapping session lasts 2-3 hours and involves dialogue and arguments between the participants. In this process a shared context and language is developed, building on the combined knowledge and experience of the team. Between each session, time is required for the facilitators to tidy the documentation and for participants to reflect on their experience. The number of mapping sessions required depends upon the scale of the field but a minimum of 3 is recommended.
An alternative strategy for mapping out a taxonomy would involve a knowledge specialist in a series of structured interviews. In our experience this is much less effective. Snowden holds that "Knowledge can only ever be volunteered and not conscripted". A well-facilitated mapping session can create the right climate for knowledge volunteers to flourish.
We agree that teams play a pivotal role in information transfer and application, but we also think individuals must assume some responsibility for sharing their own knowledge. After all, there's a lot of truth in the old maxim that you don't really know something unless you teach it. Organizing the material - creating a taxonomy - is a key part of teaching. While it may be true that "knowledge is volunteered, not conscripted," more companies are making it clear that employees who do so will be rewarded. (See "Tuning up teams" and "On the front lines in knowledge base publishing.")
Building the initial Map
The facilitator will then invite the team to spend a few minutes thinking of sub-categorizations for each of the main terms on the wall. They should work quietly and alone recording their thoughts on Post-its™ appropriately numbered.
The whole team will then be asked to stand at the wall and each individual will be asked to talk through their Post-its™ as they are fixed to the wall in appropriate places. During this phase, the facilitator will start the process of dialogue inviting other participants to ask for clarification. Challenge or criticism is avoided at this stage so that the participants can begin to understand and feel comfortable with each other's expertise.
Once all participants have affixed their notes, the process of challenge starts. Are the original terms at the correct level? Do they overlap? Does everyone still agree they are appropriate? Can the original terms be clustered in any way? At the next level down - are the sub-terms complete and consistent? Are there any redundancies? Does anything need to be renamed? How can the sub-terms be further clustered? Are there links between main categories across the map? A level of disagreement and challenge is encouraged at this point so that the team can explore the limits and boundaries of their expertise.
At a lull point, the facilitator will draw attention to the empty sector. This space is reserved for the things the team has not considered - the missing angles. Usually a team denies anything is missing and the facilitator will need to bring in Synectics creativity techniques to push the team outside its normal frame of reference looking for new angles or dimensions to the problem.
Documenting the Map
Before the second session, the facilitator will provide participants with an A3 copy of the map and will invite them to make annotations as preparation for the next meeting. The second mapping session is run directly from the computer projecting the image onto the largest wall screen possible. The focus of this session is completion and accuracy of the map to the level appropriate for the problem being tackled.
Dialogue, reframing and challenge should be encouraged throughout this session. The team will have reached a level of shared context and language that enables them to be at their most insightful and creative. The team should be looking for hidden depth, unexpected links or different ways of viewing the knowledge or domain being considered. The facilitator will make changes to the map as they go, frequently base lining so that they can go back several stages if the thought train goes wrong.
Using the Map
1. Patent Prior Art Analysis and Searching
When preparing to submit a patent for examination it is essential that all previous prior art has been located and dealt with. Missing relevant material runs the risk of the patent being rejected or challenged later once it is granted. Many patents fall at this hurdle because the scientist or engineer is immersed too deeply in their problem. They are often well aware of prior art in their immediate area but have no conception of how publications from other fields might have dealt with the problem before.
To counter this, we have used the mapping technique described to build a mini-taxonomy of the knowledge bases relevant to a patent area. We emphasize the importance of "out-of-the-box" thinking, looking for any related field in which there might be prior art. Reframing of the inventive problem is particularly important and the facilitator often uses the TRIZ  methodology to describe the knowledge problem from a different perspective.
Once the knowledge map converges on a shape that everyone is happy with, an Information Scientist will begin to work through each leaf of the taxonomy tree to construct detailed prior art searches. Some of this work is done offline without the team. The Information Scientist will test and store each of the search strategies as notes behind the map. The searching is then done live with the team. Two PCs are used with 2 projectors. One keeps the map visible and the other projects the results of the search as they happen. By doing the search live with the team, changes can be made to both the search strategy and the map on the fly, taking into account the results found and any new insights from the team. Over another couple of sessions it is possible to explore huge areas of intellectual property space and to iterate towards any prior art that exists. The facilitated team session taps into team insights and creativity in a way not possible in traditional linear, off-line sessions.
2. Environmental Scanning
Our teams are frequently asked to operate in broad areas outside their immediate area of expertise. Collecting intelligence material or tracking potential competitors is very difficult and automated tools have been designed to assist with this. These tools are powered by taxonomies generated using the mapping technique described. A high level view of one such taxonomy is shown in Figure 2. This shows only 3 levels of detail for clarity. In reality, the taxonomy is 6 levels deep and fills an A0 poster when printed to a readable size. These posters are very popular with the team and their managers as a representation of their work and its complexity. It is well worth taking the trouble to produce this - it is an ideal marketing resource and an attractive, tangible output from many hours of work.
We have been experimenting with a visualization program called The Brain. You can either create "thoughts" (words or phrases) from scratch or add them from your Knowledge Base. You can attach Web sites or documents to a thought using drag-and-drop, and you can create cross references to related thoughts. Once you've created a "Brain" on your desktop, you can publish it to an internal or external Web site. The program is a nice way to visualize your taxonomy and the associated documents (see below).
3. Structuring an Intranet
This subject excites huge amounts of interest in many organizations and our observation is that political or hierarchical pressures dictate Intranet structure more often than sensible information designs. We have used the taxonomy mapping sessions to avoid this. We engage all interested parties in the discussion including senior managers. The objective is to explore the knowledge domains that need to be covered and to find a self-consistent representation for them.
The sessions usually start with a demand from senior managers to get straight into designing menus for Intranet access. For them this is something concrete. The skill of the facilitator is then to draw them back towards a more abstract representation of the knowledge domain so that the mini-taxonomy can be built. For such a group it is important to move quickly into the Mind Manager software, as the participants are technology "hungry". Fortunately, Mind Manager can generate good HTML pages on the fly allowing the facilitator to create web example menu pages at an early stage for the team to feel that concrete progress is being made. Figure 4 shows an Intranet structure in Mind map form.
For most companies, a "throw-away" taxonomy is a radical idea, although it may be appropriate for the ad hoc team projects that Mr. Dale describes. A related -- but not so radical -- idea is the "incremental" taxonomy. In the three years since we created our taxonomy, we have developed a work flow that allows the taxonomy to evolve as project life cycles ebb and flow through four phases:
1. In "steady state" mode - as we browse our alert services and casually explore the Web - we add citations and jot down "reminder" words and phrases in the free text Abstract/Description field in our Knowledge Base (see below).
2. During the research phase of a project, we use the Knowledge Base built-in search engine to identify source materials based on these "reminder" words. A special "multi-field" search on a combination of author, title, publisher, index terms, and a free text notes field is also helpful.
3. For those materials selected for use in a project, we flesh out the Knowledge Base record with enough information for a complete footnote. At this point we also assign index terms if we think the source will be used again for other projects.
4. At the completion of a project (e.g. this "dialog" article), we index the "deliverable" - a briefing, report, or article. For longer printed documents (e.g. Briefings, course guides), we produce a back-of-the-book style index and include a subset of the index terms in our electronic index. Also at this stage, we add new index terms, create related terms in our thesaurus, and add definitions (see below).
Since we are a research, publishing, and membership organization, an "incremental" taxonomy is more useful than a "throw-away" taxonomy. The bottom line is that taxonomies can take different forms depending on the applications they are designed to support.
1. Liberating Knowledge - Understanding the sense making communities in the complex ecologies of the modern organization; David Snowden, Introductory Chapter of the 1999 CBI Handbook on Knowledge Management.
2. "I only know what I know when I need to know it"; David Snowden; Knowledge Management [Ark Group] (March 1998)
3. Synectics Creative Problem Solving - http://www.synecticsworld.com
4. Mind Manager 4.0 Software - http://www.mindman.com
5. Mind Mapping : Your Personal Guide to Exploring Creativity and Problem-Solving; Joyce Wycoff; (May 1991); Berkley Publishing Group; ISBN: 042512780X
6. TRIZ Inventive problem Solving - http://www.airiz.org
7. Sopheon Agents - http://www.sopheon.com
8. Published in Knowledge Management Review Vol 3 Issue 6 pp30-34
9. Knugget - An unrefined gem of knowledge often buried deeply but sometimes lying visible yet unseen in organisations. Even when surfaced it is frequently ignored at first glance until polished by the expert or visionary who can appreciate its potential value. The real beauty of the Knugget is that it can be cut several ways and be reused in a variety of settings without diminishing its value.
About the author