A Better View Into RelationshipsA Better View Into Relationships

Some of the most exciting opportunities for business intelligence today can be glimpsed in technologies that are just beginning to tap into the incredible potential of information visualization. Not all information lends itself to the same means of analysis and presentation. Effective discovery sometimes involves reading through stacks of text documents or laboriously studying row after row of details in tabular reports, but often our greatest insights come when we look at pictures of data. Vision is our dominant sense. While scanning a properly designed visualization of data, sometimes we experience a flash of recognition that would otherwise take hours of painful study to discover. A U.K.-based company, anacubis, is diligently working to leverage this potential through its visualization software: anacubis Desktop.

Anacubis, established in 2001, is a division of i2 Group (no relation to the U.S. supply chain software vendor i2). I2 Group has been providing information visualization software for the intelligence and law enforcement communities for 13 years. Current efforts by the U.S. military to monitor the actions of al Qaeda and locate Osama bin Laden are supported in part by software from the i2 Group called Analyst's Notebook. Identifying connections between things is an application ideally suited to visualization. The visual representation of entities — whether terrorists, companies, or products — and the relationships between them is the sweet spot of anacubis Desktop.

Before proceeding, I want to clarify the meaning of "information visualization." Like many terms that we use in the BI industry, this one is often applied more broadly than intended. Here's how those who coined the term define it: "The use of computer-supported, interactive, visual representations of abstract data to amplify cognition" (Stuart K. Card, Jock D. Mackinlay, and Ben Shneiderman, Readings in Information Visualization: Using Vision to Think, Academic Press, 1999). Successful information visualization uses visual perception to amplify our ability to think about and understand abstract data (distinct from physical data, such as a map of the world or a visualization of the human brain). A computer presents a visual representation, enabling us to manipulate the display in various ways to unveil its meaning. Static graphs, however useful, aren't an example of information visualization. But if you graphically represent abstract information on a computer screen in a way that enables exploration through dynamic manipulation of the display (filtering, highlighting, rearranging, and so on) resulting in the discovery of meaning, you're engaged in information visualization. Do it well, and the resulting insights can be extraordinary.

A new version of this software, anacubis Desktop 3.0, is scheduled for release in October. Consistent with earlier versions, anacubis Desktop 3.0 provides a node-link type of display that represents entities and the interconnections between them, revealing complex networks of information that would be difficult to display in any other way. A good example is the network of entities and connections that exist in any Web site, with pages as the entities and links between pages as the connections. Figure 1 shows a network, visualized using anacubis Desktop 3.0, consisting of multiple types of entities and relationships relevant to a recent legal case between Microsoft and Sun.

FIGURE 1 - Connections to and between Microsoft and Sun regarding a recent legal case.

Anacubis Desktop uses a combination of visual components and text to represent the data. Icons represent the entities (companies, people, and news releases, for example). Lines represent the connections between entities, with different colors for relationships of different types (red for competitor, black for cooperation, green for officer, and so on) and arrows when the connections are directional. This display was constructed from anacubis-enabled information about companies, industries, and markets — freely available from Hoover's — along with articles about the legal case retrieved from the Web, which I accessed and integrated using anacubis. Let your imagination loose for a moment, and you will begin to see the potential of this approach to information visualization.

Anacubis Desktop 3.0 performs the following primary functions:

FIGURE 2 - The anacubis Desktop 3.0 interface.

Anacubis organizes a single collection of related information into a document, and you can create separate displays of that data as sheets, similar to Excel documents and worksheets. The conveniently designed interface is divided into five sections, each of which you can selectively display or hide, as shown in Figure 2 (the sections labeled in red):

Anacubis Desktop 3.0 isn't a general-purpose BI tool. It's an information visualization tool that specializes in the analysis and presentation of entities and relationships that make up a network of information. This is relatively unexplored territory for commercial software.

I'm encouraged that anacubis has stepped into this promising space of network visualization and taken on its inherent challenges, but so far anacubis hasn't ventured very far into this territory. In some unfortunate ways, the design of anacubis Desktop actually undermines the strength of visual perception, borrowing GUI conventions that weren't designed for information visualization. The most apparent example is the use of icons to represent data entities such as companies, people, and news articles. Icons like these are too visually complex for the role they're being asked to perform in information visualization, which is to enable rapid visual perception of meaningful patterns in a rich set of data. To support this perceptual process, the entities should be represented by simple visual objects that rely on no more than one or two visual attributes (position, color, size, and so on) to encode their meaning.

The most interesting and useful information about a network of entities and their interrelationships isn't its static structure, but the variable nature of those entities and their interactions. Think back for a moment to my earlier example of a Web site, with its many Web pages and the network of links among them. While it's useful to visualize the Web site's structure, the insight that this provides is limited. What might be more interesting is a visualization of the relative popularity of the site's pages as well as the traffic volume through each of the links for a given period of time. Even more revealing might be a visualization of how these measures vary through time. For example, wouldn't it be helpful to see the popularity of each Web page visually encoded as the size of the entity or through a range of color intensity from light to dark, and the traffic along the links encoded as the proximity of the entities to one another or the thickness of the lines that connect them? At present, these interesting quantitative measures are buried in the text of the anacubis display, untapped by the rich potential of information visualization.

Rich network visualizations like the one I just described do exist, but aren't readily available in commercial software. Not yet. Perhaps anacubis Desktop will evolve into the software that delivers these riches to the BI market. As anacubis continues to mature, here are a few of the features and functions that I'd like to see:

Some of these features are, no doubt, technically challenging, but this direction is necessary to exploit the tremendous power of visual perception for data analysis. Anacubis is already on an ambitious release schedule, which I suspect will rapidly produce deeper and more exciting forays into the realm of information visualization, so keep your eyes peeled.

Stephen Few is the founder of Perceptual Edge, a consulting firm that specializes in information design for analysis and communication. His new book, Show Me the Numbers: Designing Tables and Graphs to Enlighten, is now available from Analytics Press.