Question and Answer: Making Data Meaningful with Visualization
How data visualization techniques can encapsulate vast amounts of data into intensely visual and data-rich displays.
- By Linda L. Briggs
- September 9, 2009
When it comes to viewing data, “how do we give users not just a dashboard but the [entire] windshield?” asks Andrew Cardno, CTO of BIS2. He’s discussing the possibilities offered by data visualization, a technique beginning to be used in business intelligence in which thousands or even millions of pieces of information can be summarized in a single data-rich display.
Cardno says that rather than presenting a few dozen key numbers or facts, as dashboards often do, data visualization techniques leverage the immense processing power of the eye and mind to combine and present a far greater amount of information. A geospatial system, which combines spatial software and analytical methods with geographic data sets and offers up data in visual formats, is an excellent example of what is possible with data visualization in BI, Cardno says.
Cardno is known for his pioneering work in the data visualization industry and is the recipient of two Smithsonian Laureates, in 2000 and 2001, for his contributions to visualization technology.
BI This Week: What is your definition of the term “data visualization” as it’s beginning to be used in BI?
Andrew Cardno: I believe that data visualization is no longer a term; it’s now a field. The field ranges from the charts that tools such as Microsoft Excel produce with ease, to high-end displays made with mapping tools such as those from geographic information system leader ESRI or Pitney Bowes with its MapInfo. Within the field of data visualization, those sorts of “super graphics” are just beginning to be used in business intelligence.
The study of great cartographic works such as the famous chart by Charles Joseph Minard portraying Napoleon’s losses to the Russians in 1812, enables us to realize that we are constrained only by our imaginations in how we represent huge densities of data. Geospatial systems are great examples of this: they are rich in display methods and intensely visual. The geospatial world is a good example of bringing together the art and science of visualization.
Where is data visualization headed, and what do you think its impact is going to be on BI?
The world is headed toward massive data sets, there’s no question about that. This massive growth in data, and the exploitation of its value, produces a need for faster understanding of more complex problems.
The BI industry is rapidly expanding its reach and broadening its mandate. There are two driving forces in the world of business intelligence: we can call one computationally centric analysis, and the other visually centric analysis. I believe the future of BI will combine the power of both.
Let me explain: Computationally centric analysis is analysis in which the applied expertise focuses on computation and numeric analysis; it’s not aligned toward data presentation. Many data mining techniques fall strongly into this category, such as k-means clustering (a method of cluster analysis) or regression analysis.
Visually centric analysis, on the other hand, is analysis in which expertise is applied that focuses on presentation of the data. Geographic mapping and traditional charting fall strongly into this category.
Do you see the two types of analytics eventually merging, and to what result for the user?
Yes, I see the combination of visually centric and computationally centric coming together into a natural partnership which provides both analytical rigor and human understanding.
After all, visually centric analytics and computationally centric analytics are naturally linked. In many ways, the world of science has used the power of observation (visually centric analysis) for years to establish a hypothesis, then has used experimentation and modeling of results (computationally centric analysis) to prove or disprove the hypothesis, then the presentation of results (again, computationally centric analysis) to show the answers. This scientific method, which in my view should form the basis of all analysis, is accepted as an effective means of discovery.
I also believe this: That it’s irresponsible to model data in a form we don’t understand -- yet that is something that often happens today. The poster children for that are the models that drove the financial services industry, an industry that prided itself on layers of complexity so deep that critical elements of risk were hidden from management.
It’s the critical role of all modeling to produce visually centric displays that enable a clear understanding of both the data and the data outliers. Nassim Taleb’s book, The Black Swan, about randomness and uncertainty in the world and our reaction to it as humans, describes the occurrence of rare and improbable events, or outliers, and how we cope with them. In a world of black swans, we must ensure that all models are understandable and all predictions are visually displayed. That helps ensure that outliers can be seen as opportunities.
Basic data visualization isn’t really new -- Excel’s chart functions offer data in a visual form, for example. However, it seems that most users are still trying to do even basic charting well. Why is that?
Charting is surprisingly hard to do well. Computers have provided us with enormously powerful tools to generate all manner of graphs. The challenge with formalizing the methods is that it’s a combination of art and science. As with many areas that straddle the line between art and science, such as building a data warehouse, human decisions in the presentation method are key.
One example is the use of interpolation. There are many instances where interpolating between points is in a strict sense incorrect, as with a map showing revenue by customer location. However, this interpolation often provides a clearer picture of patterns in the data.
Artwork is often exciting or stimulating and thus attracts attention. The art of visualization often takes us to a point where we say, by making these numbers or this display more interesting, we can improve our communication with users.
A very senior analyst with a major international company recently told me that while it was not strictly correct, he wanted to display a series of results as a line graph rather than a bar graph because he felt the results would be easier to understand. Thus does art meet science.
Where does BIS2 fit into all of this?
BIS2 is focused on extending the value chain by providing high-intensity visualizations within the enterprise environment. At BIS2, we start with spatial visualization, leveraging the now-mature ability of the relational database to store, query, and manage spatial data. We then move to exploiting techniques such as those shown by the “cyclogram” created by Russian cosmonaut Georgi Grechko, where non-spatial but correlated data can be shown using cartographic techniques.
At BIS2, we’ve found that the application of our Quartal super graphic provides the perfect analytical front-end for the customer segmentation schemas. We’re working with some of the largest retailers and telecommunications companies in the world, helping them adjust their customer marketing programs by giving deep insight into the changes they face.