Searching for a global solution

For global corporations, reconciling cultural and language variations is essential to ensure strategic advantage. Jean-Paul Lambermont-Ford shares his experiences of implementing an innovative multi-lingual taxonomy at Cabot Corporation, which allowed employees world-wide to share knowledge regardless of location.

We have all heard countless cultural jokes featuring Englishmen, Irishmen and Scotsmen, as well as tales of the foibles of lawyers or accountants. In their own way, these jokes illustrate an underlying awareness of cultural and language differences and the ways various groups view, and relate to, their environment. However, taking these opposing elements and transforming them into a sustainable operational reality can be challenging. Multi-lingual taxonomies have been developed as a tool to support global consistency and local flexibility.

Cabot Corporation, a chemical manufacturing company, has 4,400 employees operating in 21 countries world-wide. They specialise in making and handling fine particles: from manufacturing operations and research and development, to technical services and sales. Cabot has undergone a significant re-organisation in recent years to move from a regional-business structure to seven global businesses. The variety of job functions and the firm’s global focus mean that different user groups at Cabot have different operational requirements.

The context for KM at Cabot

A core value at Cabot is innovation, which applies to both its products and its use of a collaborative working and e-mail platform, IBM Lotus Notes, since 1994. Cabot implemented database-support teams and KM projects before the terms ‘communities of practice’ (CoP) and ‘communities of interest’ (CoI) became common currency. Knowledge repositories and workflow systems have been at the core of the company’s business ethos and form the basis of the company’s intranet. Since 1994, Cabot’s regional-business orientation has meant that a number of relatively consistent globally and locally developed systems, with differences in terminology and interfaces, had to be developed. On the positive side, this ensured that large numbers of employees were exposed to the concept of storing, sharing and retrieving information used mainly by engineering, managerial and administrative staff. And with the exception of the purely local databases, which were often in the local language, the majority of the databases were in English, the working language across Cabot world-wide.

In the late 1990s, Cabot adopted a shared-services model. The quality department in Europe recognised that the quality systems in each manufacturing site needed to be aligned. This would assist the model in the long term, be part of a continuous-improvement process, and help with ISO9000 registration and audits, which were resource intensive but essential to Cabot’s business.

The process started with an initial review of the quality systems of the Carbon Black manufacturing sites in Europe. The site quality managers identified issues with managing documentation and suggested solutions for overcoming them. This showed that managers were aware of the limitations of their existing systems and demonstrated a willingness to change. The managers concluded that:

Duplication should be avoided;
Documentation should always be current;
Quality of documentation should be consistent;
There should be improved use of documents;
Obsolete documentation should be avoided;
Management of documentation needed to be improved;
Document control and versioning problems should be solved;
The document review/approval/publishing time-cycle needed to be reduced.

The identification of the major issues, coming from those who would later be key users and implement the system on their own sites, was encouraging as it increased commitment to the project later down the line.

The project team was made up of members of the quality department and an internal consultant seconded from the IS department, with experience in implementing systems supporting CoPs and CoIs. An information audit was carried out for each of the local quality systems with the local quality manager, assessing where and how information was stored, structured and accessed, its source, audience and lifecycle. We then had a starting point: seven different systems in six languages, mainly stored on shared-network drives and delivered on paper. Each system was fit for purpose in its own right, but without structural consistency across each site, regional business needs would be difficult to anticipate.

Systems specialising in quality/ISO document management were then evaluated. One site was just starting to use a document-management system, Achiever Plus from Achiever Business Solutions. Early work gave an indication of the strengths and weaknesses of the product and we learnt from the way the quality departments had struggled with classification that this was a problematic area that would need careful attention.

The Achiever product was retained, as the supplier could provide it in the different language versions necessary in Europe, and it fit with the following criteria:

It was Notes based – users were already familiar with the underlying interface and the databases could be replicated to other sites without any additional investment in infrastructure;
The product could be supplied ready translated into the necessary languages – this was imperative as all employees at each site would be using the product and a hybrid English/local-language interface would not always be appropriate. In a production environment, there is no room for misunderstanding;
Minimal customisation – this was a deliberate choice, to minimise the IS resources required in implementing the project and to limit user requests for changes.

European implementation

From the outset it was clear that, to succeed, each site needed to ‘own’ its local system. The local quality manager would be the initial key user in each site and train further key users who would, in turn, train the rest of the site’s staff. This would ensure involvement at all levels and make the project local rather than regional from each site’s perspective, making it sustainable. From a regional stance, it was important that there was consistency across the sites in terms of structure and document format so that changes arising from business or external requirements could be co-ordinated and easily met.

The format of quality documents is fairly fixed, with standard sections imposed by ISO9000 requirements. Similarly, there are generally recognised types of document. In simple terms these are: policies – why something is done; procedures – what is done; work instructions – how to do it; and controlled documents – a grouping for any documents that need to be current and versioned. These categories are common to all existing systems, enabling the local quality managers to begin to relate to the proposed system. The major areas of change were evident in the classification of documents and the opportunities provided by a document-management system, with both hierarchical and poly-hierarchical tree structures, as well as full-text indexing features for searching and browsing. It quickly became apparent that this would be a contentious area, as the needs of all the stakeholders, including regional management, local management and users, would need to be satisfied equally, but an element of consistency across the different systems was imperative.

Taxonomy development

Each quality manager had developed their own system with its own classification structure, and was understandably reluctant to abandon it – after all, it had worked well at their site. While it would have been possible to transfer existing systems into Achiever Plus with the same classifications, this would not have satisfied the drive for consistency, nor would it have necessarily helped users to find what they needed. At a regional meeting, a focus group was convened to lay the foundations for the taxonomy, facilitated by the regional quality manager and the internal consultant seconded from the IS department. To demonstrate the possibilities of the new system, a pilot was set up and populated with a subset of documents from one of the quality systems that had been recategorised to show them in hierarchical, poly-hierarchical and full-text-search contexts. Use of analogies was made between these and tables of contents and indices in books, as well as practical examples, to assist in the concept change.

At this stage, two classification camps emerged: process-oriented and function-oriented, even though both could be accommodated within the system. This mainly reflected the protagonists’ existing systems. Using the information-audit results, we were able to compromise: within the hierarchical tree, policies and procedures would be classified by function, and work instructions and controlled documents were classified by process. This satisfied the stakeholder groups as management tended to deal more with policies and procedures by function and production staff needed information by process. Nevertheless, the poly-hierarchical tree could be used to provide a view across documents types by either process or function.

Quality managers were armed with a supply of Post-It notes and were asked to put down their first and second-level headings, which were then assembled into an outline subject tree. The first-level headings were straightforward, but second-level headings provided significant insight into the differences in granularity seen as necessary by the participants, as well as their placement in the subject tree. The granularity issue was resolved by using the guideline that, generally, no more than one screen of results should appear under a second-level heading, and that only in exceptional cases should a single document be assigned its own category. The placement issue was resolved by arbitration. The resulting taxonomy affirmed Peter Kibby’s suggestion that taxonomies in a corporate context may not be rational in a classical sense.[1]

This became the basis of the core common taxonomy, which was used as a controlled vocabulary for indexing. Local quality managers translated the terms into their own language, and when more than one site used the same language, they agreed on a common term. Ownership of the taxonomy was split between the region and the sites. The region was responsible for defining terms for policies and, to a large extent, for procedures. The sites determined their terms for local procedures and for work instructions and controlled documents. This balance allowed each site to mould its own taxonomy to suit its needs, while giving a degree of consistency across sites, reflecting TFPL’s suggestion that different taxonomies should support self-contained intranets [2], albeit with a twist: there was an overarching common core.

Going global

At this point of the implementation in Europe, the scope changed. A decision was taken at corporate level to extend the Achiever project globally to provide a single, consistent platform for managing quality documentation.

However, we were concerned about how this would affect the project in Europe. We re-examined the nascent federated model in Europe, where there was consistency emerging, but little opportunity for synergies because of the language differences. We compared other possible models – a single database for all documentation or regional databases, both of which were discarded because of the complexity of managing them, language issues and the need for each site to have a degree of autonomy in its taxonomy.

The federated model was retained but we were apprehensive about how easily it could be scaled up. We found that the principles established for taxonomy development and implementation could be transferred to other regions and businesses. Furthermore, the core taxonomy for the Carbon Black business in Europe could be used in other regions and could also be used as a model for other businesses.

We also needed a place for documents that applied globally, to a business or to a group of sites – for example, global policies or health and safety documentation relating to US sites. Rather than multiplying the number of systems, we opted for a single database covering these areas. Each site would thus have two databases: a local one that related to that site or business in the local language, and a global database, in English, to hold documentation available to all sites and businesses. This also took advantage of the underlying replication model of Notes, making updates to the global database available in all sites with a short time via Cabot’s wide-area network.

The core taxonomy structure was retained for the global database, with a top-level set of headings added for businesses and regions. This allowed a business or region to have the same basic structure as the local sites, ensuring consistency and ease of use because of users’ familiarity with it. Using the poly-hierarchical tree, it was also possible to view information by process or function across sites and businesses.

Initiating roll-out

The local/global database model was ready and the system started its roll-out phase in January 2000. A pilot site was used in France for the first implementation, taking place over a week, with a follow-up training session for the local quality manager. The implementation process and training were fine-tuned from the feedback that had been accumulated during the pilot. It was still apparent that classification was an issue at a local level, and so each site was given a ‘sandbox’ database (an extra copy of their local database) for training and experimentation so they could familiarise themselves with the database structure before transferring them to the production system. Similarly, greater emphasis was placed upon classification in the training programme, with half the time invested in classification and ensuring local/global consistency, and the opportunities this provided for sharing information with other sites. The remainder was spent on the document-management process itself and the mechanism for updating documents in the local databases from the global database. The sessions were hands-on, lasting three to four days depending on the participants’ languages, and were delivered at regional meetings with the local quality managers and key users from each site. During the training sessions, any existing synergies were exploited, for example the South American sites already shared much information, and the North American ones had very similar systems.

The roll-out continued with North America, South America and Asia-Pacific during 2000 and early 2001, with a final round for individual businesses in mid to late 2001. Finally, local databases existed in 35 sites and nine different languages. The global database was also available in the same sites and at office/research locations, bringing the total number of sites to 45.

Cascading information

Transferring information between the global and local databases has been achieved via a cascade mechanism, so that when a new document is published in the global database or an existing document is updated, the main users of the document at each site are alerted by e-mail. The recipient can then incorporate the document into their system and translate it where necessary. This ensures that each site has its ‘own’ complete set of documentation. Sites have latitude as to how and when they incorporate new or updated documents, allowing them to control and manage the change to fit with their own operational needs.

The cascade mechanism is also used to send changes in the core taxonomy to the key user at each site, which they can then incorporate into their own system. Again, the cascade model allows sites to own the process and to work at their own pace.

This allows the global database to be used as a repository for good practices across Cabot, for benchmarking between sites and businesses; and increasingly as a source for information that local sites can use and pull into their own systems. Its use has been extended from quality-related documents to encompass other areas and we are seeing synergies between plants at local levels in terms of taxonomy, especially where they share the same language. This has also led to a realisation that information is readily available elsewhere within Cabot and the emergence of a general knowledge-sharing culture as people are beginning to share information laterally as well as the vertically with the global/local model. For example, within one business with different languages, users at one site can see their colleagues’ information in English at another, and suggest changes that could be made based on their local experience, and then re-incorporate the updated document. Within a local database, the mix between local documents and those pulled from the global database is relatively consistent across sites.

Challenging times

During the roll-out, two further issues arose, which were due to the fact that the project was essentially locally implemented and grown. Paradoxically, the issues, which are outlined below, were also much of the reason why the project was such as success:

Taxonomy creep – some sites have created parallel taxonomies, reflecting their comfort in their old systems and enabling them to transfer their documentation en masse. In the short term, this is being remedied by incorporating the parallel taxonomy as an additional set in the poly-hierarchical tree, allowing for a transition period;
Insularity – a few sites were reluctant to make all their information openly available at first without it being filtered. This has now been resolved, as the site users saw that others were sharing their information openly. Cultural change thus became embedded.

Enabling and adapting

The Cabot case shows that while difficulties can arise in reconciling different elements of a business in terms of culture, language and conceptualisation of information, it is not an impossible task. The inclusive approach taken at Cabot, allowing local flexibility while maintaining global consistency through the use of federated systems bound together by a common core taxonomy, has helped to build a successful environment where employees are able to retrieve the information they need, as well as promoting a corps d’esprit for knowledge sharing. Users are starting to see the benefits of being able to access and contribute to the knowledge of others, as well as sharing their own knowledge. This has led to a common view of the company’s information resources via its taxonomy which, as Graef suggests, is an undeniable source of strategic advantage.[3]

References

1. Kibby, P., ‘The case for corporate taxonomy’ in Knowledge Management (volume 5, issue 5, Ark Group, November 2002)

2. Abell, A., Howells, A., Kibby, P. & White, M., Intranet Management – A TFPL Guide to Best Practice, (TFPL, 1998)

3. Graef, J., ‘Managing taxonomies strategically’ in Montague Institute Review (March 2001)

Jean-Paul Lambermont-Ford, former internal consultant, Cabot Corporation, jplf@easyconnect.fr

Box out: from the trenches...

Information classification is now consistent across languages, sites and businesses, and documentation has been reduced, with approximately 30 per cent of documentation eliminated at one site alone. The ease in information access across the organisation for benchmarking and identifying good practice has been invaluable at corporate and regional levels. Information sharing is evident within sites and there is a greater awareness of internal knowledge. Sites are also looking at one another’s systems to share information and build common processes.

We are already recognising the value of the system: employees retain knowledge when they change jobs and the system is a valuable information source for new employees, providing them with a consistent information structure wherever their Cabot career takes them.

We faced a lot of challenges when setting up the basic structure. In particular, changing people’s ideas about information classification and retrieval and attitudes towards information ownership was difficult. Striking the right balance between corporate, regional and local stakeholders to ensure smooth transition of ownership was an important consideration at the design and implementation stages. User participation is critical to future success, and it has become a reflex for some users to store, share and retrieve information via the system. However, some people still do not yet use its full capabilities.

In retrospect, we would probably have given sites less autonomy in the early stages until they were comfortable using a common classification system to maintain consistency within and across sites. However, looking forwards, the system should provide a solid base to enable Cabot to adapt to the changing needs of its customers and business environment.

Alain Benoist, quality director Europe, Cabot’s European HQ, Suresnes, France