Visualization and Business Intelligence

Data Visualization is an essential element of effective BI – perhaps the essential element of BI data display. The picture worth the million words.

The increasing scale and availability of digital data provides an extraordinary resource for informing public policy, scientific discovery, business strategy, and even our personal lives. Complex patterns that we would struggle to notice in tables of numbers become instantly clear when we see them graphed

Visualization and Business Intelligence

Each of its four sets of 11 data points looks very similar when listed in a table. But when graphed they show strikingly different patterns. If these were business trends, you would probably want to see the visual pattern before making any decisions.

GRT has successfully delivered projects following the visual analytics methodology outlined by Jeffrey Heer of Stanford University and Ben Shneiderman of the University of Maryland, College Park. These steps fall into three high-level groups: Source Data Identification - selecting data and views, Information Discovery - navigating and manipulating those views, and Collaborative Analysis - recording, sharing, and explaining the analysis.


Source Data Identification

  • Visual representation: First of all a method of visualization needs to be chosen. DV systems may offer users a chart of visual templates they can use, or graphs they can select. Formal grammars of visualization options add power, but require some level of programming skill from the user.
  • Data filtering: Once data has been selected and visualized it must typically be filtered – for example, if the relevant data points are between 65 and 75, the user needs to zero in from an initial display showing values from 0 to 100. Lassos, widgets, sliders, and similar basic tools serve this purpose.
  • Data sorting: The data being visualized may also need to be sorted. This may be as simple as sorting values of individual data points, or involve sorting by averages of groups, or node patterns in a social graph.
  • Deriving new data: Raw initial data is often used to derive new visual data, for example by normalization or log scaling. Providing more powerful data derivations is a major research frontier in data visualization technology.


Information Discovery

  • Selecting specific data: Once an overall set of data has been visually displayed, the user must be able to select particular data for closer analysis. This interactive selection process is closely related to filtering. Common tools are clicking, mouse hovering, lassos, and so forth. Selection using queries applied over the data provide additional selection power, at the price of greater complexity.
  • Navigating the visual scene: Users must be able to navigate visual data, using tools such as panning or scrolling across a visual display, or zooming in or out. Navigation tools provide a viewport on the data. Techniques such as visual distortion have been tested, but have not yet been shown to be aids to insight.
  • Coordinating multiple views: Visual views of data do not exist in isolation; users must be able to coordinate among alternative views. For example, selecting data in one view may bring up related data in other views. One common coordinating tool is a "trellis plot" in which a number of small graphs are displayed together in one view window.
  • Organizing multiple workspaces: In addition, users must be able to organize multiple workspaces and windows, for example providing a primary windows with others tiled to one side. This process is often automated to ease the process of navigating among workspaces.


Collaborative analysis

  • Preserving a record: Visualized data must be managed in time as well as space. At minimum, undo and redo buttons must be available. More sophisticated tools include time lines and "a;comic strip" views that display a sequence of actions.
  • Annotating work: Users must be able to make notes on their work. These may take the form of text annotations or visual highlighting, such as a circle drawn around a group of values.
  • Collaboration: Visual data analysts do not work in isolation. Users must be able to share their work. At minimum this calls for import-export capabilities. Publishing to a web page is often an important collaboration tool, while "application bookmarking" allows a new analyst to take up where another left off.
  • Providing a guided tour: Finally, the results of visual data analysis must be made available to a broader audience of data "customers," who are not themselves trained in using these tools. A picture is only useful when you know what you are looking at. Often this means telling a narrative visualization story about the visual patterns that the analyst has found or created.

These capabilities, with tools to perform them, are the basic functionalities that BI visualization solutions should be able to provide. They are the required internal features of a software package, whether it is implemented onsite or provided from the cloud.

But in real life, software packages do not work in isolation. If the software is to really provide a solution for your organization, it must work smoothly with its environment – the IT structure and organization that will be putting it to use. And successful implementation should include the following considerations:

What analytical engines does the platform support, and how does it access and interpret raw data? How does it manage in-memory data? What types of data can it analyze? Does the visualization platform require a specialized programming language? And what are the platform's integration capabilities?

This is where GRT Corporation stands ready to help. Our business is working with both technology and the businesses bringing you a partner you deserve..


To learn more about how GRT can help you meet your goals, contact us.