In-Depth

Separation of Powers: Drawing the Data Viz Divide

What qualities help separate best-of-breed data visualization offerings from their general purpose competitors?

• By Stephen Swoyer
• 06/13/2007

Data visualization didn’t abrupt entire out of a product pipeline—much less the head of a slide-rule-toting Zeus.

Of course, data viz—i.e., the art and science of using common (or increasingly innovative) visual metaphors to help represent and elucidate data—has been around for decades. Seminal data viz research took place during the 1970s, when Herman Chernoff used facial representations to better elucidate trends in multidimensional data. The textbook use-case for Chernoff faces involved a representation of living conditions in a major metropolitan area, using happy faces (with other signifiers, such as arched eyebrows, eccentric head shapes, and so on) to represent additional aspects or trends.

With the rise of dashboards—or of the next-gen performance dashboard, in any event—data visualization seems to have hit the big time.

After all, performance dashboards—including offerings from Business Objects SA, Cognos Inc., MicroStrategy Inc., Oracle Corp., and SAS Institute Inc.—use a variety of visual metaphors, typically yoked to interactive technologies such as Adobe Flash or asynchronous Java over XML (AJaX), to quickly and intelligibly represent data, trends, or alert conditions.

Not surprisingly, however, data visualization experts tend to differentiate between conventional dashboard offerings and the best-of-breed data visualization tools (which also have dashboard elements) marketed by vendors such as Advizor Solutions Inc., Tableau Software Inc., and the former Spotfire Inc. (which was acquired last month by enterprise application integration specialist Tibco.) There’s good reason for such differentiation, they argue.

"There is a huge gulf between the few vendors that understand data visualization and everyone else. When you ask if the best-of-breed products are worth the price, the answer is … [that] their prices are not all out of sync with the prices of other, more mainstream BI vendors … [and that] the data visualization products of vendors other than the few who get it aren’t worthwhile at any price," comments visualization expert Stephen Few, a principal with consultancy Perceptual Edge and a lecturer at UCLA’s Haas School of Business.

What kinds of differences does Few have in mind? It all depends, he says.

"We could definitely speak of baseline capabilities that ought to exist in data visualization software, but the requirements vary depending on the nature of the software," he comments. "At a minimum, you must make a distinction between software that uses visualization to present information and software that uses it to analyze information. Some products can be used for both, but many are only designed to do one or the other." A complete run-down of what separates data viz best-of-breed from chaff is admittedly beyond the scope of this article. But—in a research article published on his Web site—Few does go into more detail.

"Data analysis—except for routine analytical procedures that are always done in the same exact manner—requires the ability to add, remove, and change views of the data on the fly, following the trail of analytical pursuit as quickly as questions arise," he writes. "Any software tools that require more than a few seconds for the analyst to add another chart to the screen or to modify an existing chart will not work for ad-hoc analysis."

This is one of the reasons he says it’s important to differentiate between best-of-breed offerings from Tableau, SAS JMP, and the former Spotfire and those of non-specialty vendors. Such tools typically boast at least two features—namely, global filters and "brushing"—that are missing from, or imperfectly implemented in, non-best-of-breed offerings.

"Global filters provide the means to remove unwanted data from the entire display," he writes. "Filters in some analysis tools can only be used to affect individual charts. Global filters, however, work across the entire collection of charts, keeping the data in sync across the entire display. When we examine data from multiple perspectives simultaneously, we almost always want to filter the data in a consistent manner across all of the charts. The easier the filter controls are to use and the faster they can be set, the more powerfully they support the free flow of analysis."

Ditto for brushing, which Few says "extends our ability to see connections in the data by highlighting selected items across all of the views."

He uses the example of a dashboard which breaks down sales down in terms of common values—e.g., yearly sales, per-customer sales, sales by customer size, change (on a percentage basis) in sales year-over-year (and so on).

In a conventional dashboard, Few points out, a user would simply be bombarded by charts, which—independent of deeper analysis—don’t tell much about anything. But by "brushing" the data to selectively highlight common items across several different charts, analysis becomes much easier.