Business Intelligence Best Practices -

Collaboration. Communication. Community.

 E-mail to friend
Seeing It the User’s Way: Interfaces for Real-World Decision Making

by Dr. Steve Roth
There is no more important data warehousing challenge than how people can use the system to make faster, better decisions.

The Perils of Decision Making

To understand how visual and collaborative BI application/data warehouse interfaces promote more effective decision making, it is first necessary to examine the challenges users face when making important, data-driven decisions. Few big decisions are unambiguous yes or no questions. Typically, significant organizational decisions are fraught with four main difficulties:

  • Time pressure
  • Lack of a well defined problem/opportunity
  • Data problems, including unanticipated information needs and hard-to-find, incomplete, or conflicting data
  • The tendency for problems and opportunities to transcend single job functions, entailing the involvement of multiple decision makers with different areas of expertise

The complexity of the real-world decision-making process has implications for every phase of business intelligence tool and data warehouse development. For the purposes of this interface-centered discussion, it will be assumed that the underlying information and systems architecture are structured for maximum flexibility.

Understanding the User

When considering potential interfaces for data warehouse users, it is important for companies to first understand each user’s or user group’s organizational function and the corresponding decision-making process. Enterprise system realities generally require a standardized data warehouse interface for all users; at the same time, many interface options are flexible enough to allow customization for users’ particular decision-making needs.

  • A partial list of questions to consider:
  • What are the goals and responsibilities of each user, group, or job function?
  • What activities do these users commonly engage in to achieve these goals?
  • What types of decisions does each user face?
  • What information do these users need to make these decisions?
  • With what other people inside and outside the organization does each user interact when exploring and finalizing these decisions? What information is exchanged?
  • Do these users often engage in open-ended problem-solving and/or opportunity recognition?

It is also vital, when designing or evaluating a user interface, to consider how the four difficulties mentioned earlier hamper the user’s decision-making process. For example:


  • Is the user often forced to make important decisions under extreme time pressure? This is commonly seen in fields such as logistics and distribution, where even brief delays can cause a supply chain domino effect costing millions of dollars and incalculable customer goodwill.
  • Does the user face murky or undefined issues? People responsible for the management of organizations are plagued with this particular challenge: even if organizational goals are clear, there is no predefined menu of problems and opportunities from which they can choose to achieve them. This means they must spend much of their time just defining the situation and the options available to them.
  • Does the user commonly deal with incomplete or conflicting information? The classic example of this is found in the military, where commanders must figure out what is happening on the battlefield based on a steady stream of usually vague and contradictory field dispatches.
  • Is it necessary or helpful for the user to collaborate with multiple people in order to make a decision? Medical professionals in particular face this situation—for example, diagnosing and treating cancerous skin lesions often involves collaboration among primary care physicians, oncologists, dermatologists, lab technicians, and others.

Most users face a combination of these problems. In addition to exploring these issues in user interviews, when considering interface functions it is valuable to shadow users and document their daily activities.

Depending on the scope of the project, it may also be useful for companies to consult an expert in their industry to determine the state of the current business processes. While an in-depth review of this issue is beyond the reach of this discussion, it is noteworthy that a BI/data warehouse interface built to exploit flexible communication channels and information flows may diminish in value if grafted onto a rigid organizational structure. The optimal environment for group decision making involves teams whose members and roles are configured flexibly and dynamically across departments and organizations. Although such sweeping organizational reviews may be outside the normal purview of an interface development effort, it is an important piece of design for the most effective use of information.

Interface Implications

Once companies have an understanding of their BI application/data warehouse users’ needs, it becomes possible for them to choose an

interface that provides the appropriate level of decision-making support.

Interfaces that offer the most advanced visualization and collaboration capabilities are not necessarily the best choice for all situations. These advanced interfaces can significantly boost the productivity of users who must frequently analyze complex problems and opportunities. On the other hand, many users face relatively straightforward decisions based on well defined data. For these users, interfaces that provide for basic data retrieval are simpler and more cost effective. In short, the key consideration when designing or evaluating an interface is not how sophisticated it is, but rather how well it fits a given user’s or user group’s requirements.

Data warehouse interfaces have come a long way since the days of complex database queries and results reported in lengthy text-heavy form. It has long been accepted that most users will not take

advantage of even the most advanced data warehouse if the act of accessing and sharing the information is viewed as a job in itself.

Enter visualization and collaboration utilities. Most modern data warehouse interfaces feature GUI tools; increasingly, they also feature various levels of collaboration capability. It’s important for interface developers/evaluators crafting the appropriate mix for their users to understand their options and analyze the cost-benefit ratio associated with different levels of functionality. The following explanation of the latest advancements in visualization and collaboration functions provides a basis for this analysis. 


To understand why visualization techniques are such a significant advancement, it is useful to briefly look at how the human brain works. Reading tables, comparing rows of numbers and sorting through long texts are acquired skills; because these tasks don’t come naturally, they are inherently time consuming, attention demanding, and error prone. On the other hand, our brains are hardwired to very quickly process visual data, such as geographical, temporal (chronological), and relational (bigger/smaller) information.

Accordingly, with data visualization users can quickly see and understand large amounts of abstract information. Users also pick up data patterns that help them define the issue at hand and develop options for dealing with it. This helps relieve two major decision-making difficulties by helping users to act more quickly and to more easily define vague problems and opportunities. In this way, the interface itself enhances user productivity.

Visualization serves two distinct functions in an interface: information visualization for understanding, and information manipulation.

Information Visualization for Understanding

This function refers to visual displays that show relationships among many attributes of many types of data objects, including viewing the same data in different ways that may enhance comprehension.

To illustrate this, consider Figure 1, an example of a visual interface from a military exercise. First, it presents numerous data attributes. The map shows the locations and types of military units, as well as a quantitative/geographic abstraction associated with the units called a force projection. The cloud-like patterns represent an abstract measure of strength and ability to control a region relative to opponent forces. It summarizes numerous factors that might be considered in such an abstraction. These include unit mobility with respect to the terrain, range of fire, and readiness. Of course, it could be used to express other quantitative unit attributes, such as the areas in which different forms of support could be provided to them (e.g., supply, Internet connectivity, communication) or the visibility of units to enemy positions.

Figure 1

Situation Map

To further illustrate the integrative effects of the visualization, the map has been created in a coordinated manner with a temporal display that shows events and activities at the top using coded interval bars, as well as several what-if analyses. Each of these is also represented in terms of a sequence of views of combat states at different points in the timeline. The image at the top represents a particular point in time on the middle time-space, as well as the zoomed-out time-space at the bottom.

An interesting feature of this visualization is its component-based approach. Visualizations integrate numerous attributes through a process of composition. Composition occurs by embedding (e.g., map displays and interval bars within a single timeline), by alignment (e.g., the two time-lines), by using multiple visual properties of objects (e.g., the shape, color, position of graphical objects), by clustering of objects (e.g., combining elements as in military icons or labels of points), and through coordination of multiple displays (e.g., in the above display, moving the dotted line connecting the large map and the time-line changes the contents of the map to reflect the point in time).

This type of visualization has many applications. For example, Figure 2 depicts a logistics visualization. The circles represent warehouse locations; the size of each circle immediately cues the user on the relative size of each warehouse. The small squares represent the locations to which each warehouse ships. This is an example of how much information users can take in at a glance via visualization techniques.

Figure 2

Logistics Visualization

Figure 3 shows a medical visualization. Unlike the military and logistics visualizations, which were based on a geographical perspective, this example is based on a temporal view. In this case, a timeline approach makes sense because timing of symptoms and treatment is the primary data consideration for a medical context.

Figure 3

Patient Timeline

In each of these examples of advanced visualization for understanding, the display integrates a large amount of diverse data. This is very helpful to users who frequently must synthesize information in order to make a decision or where the problem is loosely defined with multiple potential outcomes. However, it’s overkill for users whose decisions rely on simple data retrieval to answer a straightforward, repetitive question.

Information Manipulation

For some basic, non-interactive tasks (such as importing data into printed proposals), it is sufficient for an interface to provide static data visualizations. However, for interactive and collaborative work, and significant, data-driven decision making, it is extremely

advantageous for an interface to provide information manipulation capabilities as well.

With this type of visualization, users can navigate to new data, separate relevant data, reorganize and transform data, and perform basic operations of drill-down and aggregation—all with direct manipulation operations rather than with complex commands. An important element of this is the ability to transfer information using basic drag-and-drop operations from a visualization embedded in one application to a visualization embedded in another application. The latter helps to achieve a visualization-based user interface environment that transcends otherwise “stovepipe” data architectures.

Figures 1, 2, and 3 also illustrate information manipulation. In digital interface form, any icon on the map, interval bar on the timeline, and embedded map in the timeline would be subject to all of the following operations:

  • Drilled into to get at more detailed information in the underlying data model
  • Dragged from one kind of visualization to another (so a unit icon can be dragged from a map to a table to a bar chart)
  • Selected or painted a color so it can be identified and viewed across many coordinated visualizations at once

So, for example, users can drill into warehouse icons in Figure 2 to see increasingly detailed lists of real-time data on warehouse stocks, delivery schedules and other key information.


Collaboration is the second major advance in data warehouse interface technology. Interface designers are increasingly building collaborative functions into BI and data warehouse front-ends as a response to the prevalence of matrix organizations and multi-party decision making (rather than hierarchical directives).

Again, a careful assessment of user needs is the first step in determining what types of collaboration tools are necessary and beneficial. Often, users simply need a way to share documents. In these cases, a standard groupware application will meet their needs cost effectively and without any unnecessary consumption of information system resources. For users dealing with heavy analysis demands and complex issues spanning several organizational functions, the goal of collaborative data warehouse interfaces is to go beyond the approach of making data sources available to all decision-making participants. Collaboration tools make it possible to capture, maintain, and give team members visibility into each other’s multiple, often divergent and inconsistent, perspectives of commonly accessible information. This is especially useful when data is absent or uncertain because team members can rely on expression of their insights and knowledge.

In this way, collaborative interface functions can mitigate the remaining two difficulties associated with analysis-heavy decisions by helping users overcome data inconsistencies and bring multiple perspectives to bear on a problem or opportunity. There are seven primary functions collaborative interfaces serve:

1. Information-Based Collaboration

This is the coordination of expertise of multiple users with shared workspaces containing the same data coordinated with different, local, and/or private views of related data.

2. Collaboration for Recognition of Complex Patterns

Visualization alone cannot make some complex situations understandable to any one person. Because of interacting processes, each perhaps only understood by a different expert, it is necessary for a group to view and convey their mutually guiding interpretations of the picture. A collaborative visualization is more than a picture—it is an environment for enabling the coordination of expertise. People share their interpretations about the fragments each understands so patterns can emerge.

3. Collaboration for Optimization of Multiple Interests

This allows multiple collaborators to maintain continuous visibility of each other's changing constraints and goals, enabling them to make choices that are more globally beneficial.

4. Collaboration to Support Finding the Appropriate Information for Solving a Problem

People become experts with different information types and sources. Making all sources available to all people (e.g., an interface to all data in the data warehouse) isn't effective unless it is possible for data to be dynamically customized to be understandable and navigable for each user and for each situation. The value collaboration brings to problem solving is that each collaborator can help others get to what they need rapidly.

5. Collaboration to Manage Uncertainty by Providing Parallel but Coordinated Perspectives

As with recognition of complex patterns, cooperative interpretation is useful in helping team members form conclusions in the face of inconsistencies, inaccuracies, and insufficient information. A collaborative interface doesn’t rest on the assumption that the information it links to will be free of ambiguity and uncertainty. Rather, the assumption is that there will always be conflicting data; the way to mitigate it is through multiple-perspective collaboration. Often this is not just a data-driven, bottom-up process of cleaning data. More important, this is a sharing of top-down expectations, intuitions, and assumptions in the face of missing, uncertain, and inconsistent data.

6. Collaboration that Spans Synchronous, Asynchronous, and Unpredictably Discontinuous Communication

It is often useful for interfaces to allow for discontinuous connectivity. It is important that the interface design include a way to gracefully reconcile differences in collaborators’ workspaces produced because of connectivity differences.

7. Collaboration for Coordinated Action through Topsight

Topsight is the visibility each member of a team has of the insights, thinking, and perspectives of others. Collaborative interfaces allow people to view each other’s developing inferences about a situation and their developing ideas for action. This occurs not because they provide interim plans and reports to each other, but because they have visibility into each other’s ongoing analysis and planning and the plan fragments that are considered.

Regardless of the specific collaboration functions companies build or buy for their users, three keys are common to all advanced collaborative BI and data warehouse interfaces:

  • They present pictures of people’s assertions about data in addition to visualizations of the data. This often includes visual, gesture-based, drawing vocabularies that enable decision makers to sketch their intuitions of situational awareness.
  • They encourage and make visible differing assumptions and conclusions that arise because people have different perspectives on situations. This is in contrast to enforcing uniform interpretation of uncertain data or attempting to resolve uncertainty prior to making data visible.
  • They make ongoing decision processes visible and public, in contrast to visibility only on end products.

While these collaborative interface features can promote great productivity gains among user groups tasked with large and intricate decisions, they do not come without a price. Advanced collaboration functionality is more costly than groupware and requires appreciably more bandwidth on the client side. In addition, for advanced collaboration to fulfill its promise it must be tailored to a specific user-group—by nature it is not a shrink-wrapped solution.


The most notable interface revolution in modern times was the switch from DOS systems to GUI and Windows. It was marked by massive user adoption and improved productivity, speed and decision making.

Similar to the way GUIs replaced the incredibly cumbersome text directions people used to control their DOS systems, visual, collaborative BI tool/data warehouse interfaces are rapidly replacing the laborious verbal and text-based processes people used to rely on to understand data, and communicate their interpretations of data and the situations that data represent. In many cases burdensome for both the speaker/writer and the listener/reader, words are often inefficient and ineffective as a currency for complex thought exchange. Building collaborative visualization into a data warehouse interface makes perceptual communication possible and is leading to marked productivity, speed, decision making, and team functioning. As in the case of Windows-based interfaces, efficiency improvement and rapid user adoption happen because people can devote most of their attention to the problem or opportunity at hand.

With data warehouse interfaces, one size does not fit all—it all begins with user needs. Decision makers who mainly rely on “what, when, where” data are generally better off with basic GUI and collaboration tools. Those who face with complex decisions with major organizational implications—based on analyzing the whys and hows—can enrich their decision making and boost their productivity through advanced visualization and collaborative techniques. Either way, understanding the promise of the interface could be the secret weapon in a successful BI/data warehouse mission.

Dr. Steve Roth -

For more than 15 years Dr. Steve Roth has served on the faculty at Carnegie Mellon University's School of Computer Science. At CMU, Steve directed the SAGE Group, which performed groundbreaking research on visualization and intelligent user interfaces. The group designed an expert system that can automatically design visualizations and user interfaces based on the characteristics of the data, user sketches, graphical examples, and user preferences. Steve is also CEO of MAYA Viz, a company that creates software that helps organizations eliminate boundaries. He has focused much of his career developing innovative approaches to developing computer systems with which people can use and create information. Steve holds a Ph.D. in cognitive psychology and has published more than 50 papers related to visualization and user interface technology.