Fifty years ago, the philosopher and social commentator Marshall McLuhan coined the phrase, "The medium is the...
message." While there are many perspectives on his intended meaning, this is a common interpretation: The method by which a message is conveyed may influence both the way the content is understood and the context through which the medium itself influences and affects that understanding.
Unfortunately, business intelligence practitioners and business analysts often neglect that concept when developing BI dashboards. While the availability and use of a broad palette of widgets for visualizing data suggests that there's ample demand for graphics in dashboards, the plethora of choices shouldn't dictate that flashy visualizations be employed whenever possible.
Instead, it raises two fundamental questions about dashboard development: How do different data visualization techniques convey messages to business users? And how should BI developers decide what kind of graphics are right for different dashboards?
In general, any type of visual representation has its pluses and minuses. Consider this basic issue: When is it better to use a pie chart versus a bar chart? Both methods can be used to convey comparative magnitude, such as the percentage of a population that drives different vehicles. But it might be difficult for users to differentiate between the sizes of the slices in a pie chart, especially when the numbers are relatively close -- in that case, a bar chart could be the preferred method. On the other hand, a pie chart is often better for showing how a specific group of drivers compares to the entire population, which can be more difficult to discern from a bar chart.
Proper visual balance a must to maintain
Viewed from that perspective, making the right choices on data visualizations becomes an art that balances the medium and the message. The decision must be driven by the business directive and user needs: What business decisions are informed as a result of the data presentation, what types of questions will data analysts and business users ask, and which visualization methods most effectively help to answer those questions?
First, consider which of these common data analysis activities are performed by the workers who will be using a particular dashboard in your organization:
- Finding specific values, such as the number of people driving SUVs in the Washington, D.C., metropolitan area.
- Finding aggregate values, such as the average number of minivans sold monthly by suburban car dealerships.
- Filtering down to subsets of data based on a set of desired conditions -- for example, focusing on households with both an SUV and a minivan.
- Identifying outliers and anomalies, such as households with more than 10 automobiles.
- Sorting a collection of data in ranked order using specific measurements, such as ordering communities based on their average household income.
- Clustering groups with similar characteristics -- as an example, organizing buyers of different car models based on age, sex and education levels.
- Correlating attributes to determine relationships, such as relating gender to automobile preferences.
Second, the decision process for determining the optimal course on data visualization will hinge on a number of factors that can influence dashboard design objectives. Some of them include:
Intended actions. What do you expect the user's next step to be? For example, do you anticipate that a specific business process will be triggered based on the presented information, or will users want to drill down into the data in search of additional insight? The answer can lead you down different visualization paths.
User sophistication. Different types of users absorb information in different ways, so you'll want to gauge who your audience is and how analytically sophisticated it is. Ensuring that the level of data visualization complexity is properly tailored for the users of a dashboard will keep them from being overwhelmed by the graphics.
Delivery requirements. The ubiquity of mobile devices makes it likely that users will view dashboards on different types of systems at different times. You likely will have to select a data visualization approach that takes into account the differences in usable "real estate" between a smartphone, a 10-inch tablet and a full-sized computer screen.
Geographic content. If the measurements to be displayed involve location data, it might be necessary to include visualizations that incorporate geographic information and make it easy to understand.
Time frame. It's one thing to provide a static view of data fixed at a particular time. But some analyses are designed to convey how sets of data have changed over time, which might call for the use of animations to help show the changes.
Another step is to examine the various operational scenarios in which data analysts and other users will be accessing data. Continuing with the automotive theme, one example might be a marketing application to help car dealerships determine the best locations for newspaper advertising; another might be a mobile app for prospective buyers looking for dealers with particular car models and configurations in stock.
With all of that information in hand, test different ways of visualizing data in dashboards to see which ones are best able to satisfy the business requirements. And in doing so, remember that the most visually appealing and flashiest graphics aren't always the optimal ones for communicating the desired information. When the medium engulfs the message, a visualization may look pretty -- but it won't accomplish your business goals. Two enduring words of wisdom from Henry David Thoreau's Walden fit here, 160 years after he wrote them: Simplify, simplify.
About the author:
David Loshin is president of Knowledge Integrity Inc., a consulting, training and development services company that works with clients on BI, big data and data management initiatives. He also is the author of numerous books, including Business Intelligence, 2nd Edition: The Savvy Manager's Guide. Email him at email@example.com.