INFORMATION VISUALIZATION CONCEPTS

What is Information Visualization?

Information visualization (also referred to as "business visualization", "data visualization of business data", or "visual analytics") is a means of transforming data, from spreadsheets, files, databases, web sources, etc., into visual representations that we can quickly comprehend trends, patterns and anomalies.

An effective information visualization presents the all the various information we need to make a decision in this intuitive visual form that permits to readily understand, interact with and make decisions.

Rows and columns of numbers and other data are tedious to analyze:

  • they require active attention,
  • they require the viewer to perform mental arithmetic to compare values,
  • they may require the user to create a mental model, that is, some model of understanding of how all these various pieces of information interrelate.

    Charts: Simple representations of one or two attributes

    A simple bar chart or pie chart transforms this data into a simple representation permitting the viewer to visually compare similar values thereby removing the need for the viewer to calculate the arithmetic. This includes the charts and graphs most people are familiar, such as those found in Microsoft Office and many "dashboard" products; as well as specialized charts such as financial charts (e.g. open-close-hi-low chart, candlestick chart), statistical charts (e.g. box-plots) and other chart types (e.g. bubble charts, network diagrams).

    However, simple charts are not adept at conveying multi-variate data, that is, data with many attributes. Some charts attempt to use visual attributes such as size, color and/or 3D to add additional variables, but overall charts are designed and limited to a few key variables. 

    Linked Charts: Adding variables with additional charts

    Linked charts overcome the limitations of a single chart, by linking various charts representing different attributes. The simplest form of a linked chart, is simply to align two charts with a common variable. For example, many stock charts show daily stock price as a line chart, and show daily volume as a bar chart immediately aligned underneath the price chart. A spike in the volume chart typically is aligned with a strong change in the price of the stock.

    Another means of linking charts together is through interaction. For example, one chart may indicate the number of owners of each automobile type, a second chart may indicate the salary levels of the automobile owners and a third chart may indicate their age. Clicking on the bar representing owners of sports cars, updates all the other charts to indicate the proportion of owners owning these sports cars and perhaps may reveal a large number of middle aged men in upper income brackets own these sports cars.

    ADVIZOR Solutions' ADVIZOR and COGNOS Visualizer represent two commercially available systems utilizing some form of linked charts

    Integrated Charts: Merging representations together

    Linked charts force the viewer to understand the data through a palette of pre-constructed visual representations - e.g. bar charts, line charts, pie charts, etc. The most appropriate representation is dependent on the task: what is the viewer trying to solve. If the viewer is interested in geographic or spatial attributes, maps may be useful. If the viewer is interested in time attributes, animation or cyclic representation may be useful. However, data that has geographic attributes does not necessarily imply that the viewer is interested in geographic analyses. Hence "automatic" visualizations may not generate what is in a viewer's interest and my central belief that good visualizations must be designed! (Just as good architecture, good web pages and good interfaces are all the result of thoughtful design.)

    These visualizations, referred to as "custom visualizations", "tailor visualizations", "designed visualizations", or "information fusion" merge various data representation techniques together and are typically created by a multi-disciplinary team of designers, engineers and the users (subject matter experts). Since effective merged representations are tailored to the uses of their viewer, examples of effective integrated visualizations can be seen in specialized applications such as: supply chain management (e.g. i2 technologies) stock market analysis (e.g. in use by NASDAQ, etc) data quality analysis (e.g. in use by Statistics Canada, etc) portfolio management (e.g. Northern Trust, etc). 

    See the data visualization gallery at www.oculusinfo.com and www.advizorsolutions.com for more examples of all kinds of visualizations.