Matt Pearcey, Darren Pywell, David Tattersall²

1Intergral Ltd., ²Intergral GmbH

Extended abstract

This paper describes the design, development and deployment of a large-scale dynamic web-based information management system in the intranet and extranet of a large, multinational computer company. It attempts to identify and address some of the issues to be encountered in projects of such scope.

The core concept and focus of this paper is that the Enterprise Document Management System (EDMS) field has missed a fundamental point in deploying web-based information management. We supply the argument that information systems that provide a large body of information/knowledge should contextualise that information such that it is pertinent to the user’s requirements. Without this context, users encounter the well-known issues of disorientation and cognitive overhead [Ayers1994]. In addition, users tend to drift into unintentional usage models [Cockburn1996].

We suggest that a dynamic web-based navigational structure and content presentation provides more implicit context, and fundamental flexibility for business fit, than equivalent systems utilising only overviews and search engines for exposition and traversal of static structures.

However, the primary focus of this paper will be the issues encountered during the analysis, design and development of a number of fundamental aspects of the system, rather than the issues encountered on the organisational side (although these were major contributors to it’s ultimate success).

A client organisation retained our company to address an identified business case. Initially, we were retained to provide the following, for a European scope:

However, following further consultation, we were able to expand the scope of the project to cover all aspects of the information value chain, on a world-wide organisation-wide basis.

Identifying Requirements

Our approach utilised the MAD method [Pearcey1998], with the primary requirements gathering exercise roughly centred around a JAD/RAD-style technique. The approach involved user representatives from the outset, with requirements gathering being a process of negotiation, exploration and demonstration. The high-level requirements identified at this stage were as follows:

1. To provide an infrastructure for communication within MPG and between it, its channel partners and the public.
2. To provide secure access to information on the basis of clearly defined business rules.
3. To provide a single point of distribution of Marketing and Communications information both within the organisation, and to channel partners and the public.
4. To provide the information to the correct information consumers in a timely manner.
5. To provide a tangible ROI based on reduction of redundancy and duplication of information in the organisation.

To fulfil the requirements, an approach that utilises two discrete systems was proposed. This consists of a system that resides inside the corporate firewall, and provides the primary repository and communication infrastructure for marketing and communications material, and one that resides outside the firewall that provides a replica for external information consumers to access. The external replica is kept up to date by an explicit replication subsystem that releases information on a date and visibility basis. This structure is shown in Figure 2.

The metadata concept is fundamental to the approach. Our definition of metadata came from the realisation that information components were fundamentally related to business concepts (such as a product, or a product line) and that they had specific purposes (e.g. a technical data sheet, or an application note), and different user populations had differing usage of these. In addition, there were a number of other attributions that could be applied (a document has a title, it should be only visible to certain people etc). These metadata attributes define what is displayed, to whom, when, and where.

A virtual structure allows the user to determine the way in which documents appear. For example, a Product Manager user may require a view of the information that is structured around a product focus, whereas a world-wide MarCom (Marketing & Communications) representative may require the same information, but focused around a MarCom documentation perspective, rather than products.

In addition, the structure may be limited to specific attributes, based on user profiling (for example, if a dealer is only allowed to view product information relating to PL40, because that is in his contract with the company, then information shown to him would be restricted to that product line alone).

An example Dynamic View structure is shown in Figure 4, below.

Many of these dynamic structures are possible, as they are represented as an ordered set of attribute types that are expanded dynamically as the user navigates that particular view.

Every user has a profile. This identifies their position in the metadata structures. It defines what information they can see, and what they cannot. The profile also contains workflow-related information, such as their email address (for system-initiated communications), their backup, what language documents they are interested in seeing, what information they want pushed to them (selective on document types), and what language they wish the user interface to be displayed in.

As each document is loaded into the system, it is attributed with metadata by its author (Figure 5). The metadata that the author specifies enables the system to determine where in the system the document will be presented to another user. A document can be uploaded at any point to any other point. The document has no concept of location as its placement is determined purely by its metadata.

Every textual component of the system is localisable into non-English language. This covers every textual component of the application itself (such as form field headings) to the name definitions of every business-metadata component. This is enabled through the use of separate language-label tables that store the textual components in all translated language forms. There may be many translation sets in the system at any one time, with each user able to select the language that they wish the user interface to be displayed in. This also supports all non-European character sets, and extensions through Unicode.

Initially, the system was piloted in Europe prior to world-wide release. During this phase we identified a number of issues relating to its usability. In order to identify these further, we undertook a detailed set of usability tests across a set of test subjects, both in Europe and US. Generally, the main problems we found were not related to the dynamic nature of presentation. The main problems were related to layout, organisation and process. these identifications led to a set of redesigns of the user interface look and feel, rather than fundamental functionality (this was technically straightforward, as we separated the UI from the application functionality to a large extent). Finally, we applied the Questionnaire for User Interface Satisfaction [Chin1988] instrument from the U. of Maryland. Overall satisfaction was high (mean of 7), with the lowest area of satisfaction being ‘terminology and system information and feedback’, while the highest area was ‘system capabilities’.

We found that there is high correlation between specific questions (for example, 6.1 to 3.4, pertaining to the ease of use of the system), with high correlation between groups of related questions. On a per-group basis, overall satisfaction correlates comparatively highly with terminology and system information, and with questions related to learning.

As the ease of learning of the system increases, and the level of feedback, and accessibility of terminology increases, the overall satisfaction increases. This is in concordance with the focussed comments that users supplied in the post-test questionnaires. This is an expected result, considering that the system is new, and all users are in the process of learning to use the system, and adjust to the concepts presented within it.

As the overall experience level of users increases over time, it is expected that overall satisfaction will become less correlated to learning to use the system, and more towards task fit, and information presentation.

Two important conclusions were drawn from this work:

1. As Davenport [Davenport1998] highlights, one of the critical factors determining the success of any information system is the structure of the information and how it evolves over time. A dynamic structure is fundamentally capable of dealing with this, whereas a static one is not. While a static approach provides a single approach for communicating a single message to an entire population (e.g. a marketing message), sharing and adding value to information becomes impossible when a user is unable to locate the information, or is swamped by extraneous content.
2. EDMS approaches to web deployed information systems are centred entirely around either a search engine approach [e.g. Excalibur1998], or a static publication approach [Balasub1997, Balasub1998]. Our findings indicate that information systems users fall in to two distinct categories – those who search holistically, and those who navigate through data drill-down structures. Dynamic, data-driven web structures provide a drill-down metaphor that enables the latter group. Searching through a dynamic data-driven site using a search engine is simple, as the same queries that construct the site are used to provide the search results, which can in turn be used in a full-text search. This supports a wider range of user behaviour models.

[Ayers1994]

[Balasub1997]

[Balasub1998]

[Chin1988]

[Cockburn1996]

[Davenport1998]

[Excalibur1998]

[Pearcey1998]