INFORMATION TECHNOLOGY

Beautiful reports, better decisions

November 29, 2006

It may be unwelcome news for report authors struggling with data and keeping up with demand, but there's an emerging issue in business intelligence that's going to make their jobs even busier: the importance of proper data presentation.

Report authors can now control nearly every aspect of report design. Some, like fonts, are usually a matter of preference. Others, however, can lead end users to make the wrong conclusions about the data they contain.

Consider the difference between the two graphs below. Suppose your CEO needs to know the market share of yours and other companies in your industry, and she's presented with the pie chart in Figure 1.

Figure 1

Figure 2

In Figure 1, it's clear that "Company D" has the greatest market share. That's about all that's clear. Your CEO would be hard pressed to determine your company's market share vis a vis your competitors, with or without the 3-D effect. In this example, the design choices work against the data.

Figure 2 presents the same data in a bar graph. Here, it's obvious that your company ranks second, a few percentage points better than Company B, with a market share of 20.23 percent. Here, the design works with the data.

A simple example, to be sure.
But the issue goes beyond aesthetics. Poorly designed reports can lead to poor business decisions, because they fail to communicate a clear picture of what's really going on. Without a clear understanding of the data, people must choose from an array of undesirable options: act on what they think it means; delay a decision, or avoid making one altogether.

"Most presentations of quantitative business data are poorly designed – painfully so, often to the point of misinformation," writes consultant and data visualization expert Stephen Few. "This problem, however, is rarely noticed."¹

Why is this so? And with margins so thin and competition so fierce, why is it allowed to continue?

Problem? What problem?

Free White Paper

Learn core principles
for displaying data

The problem exists largely because we're not aware there's a problem to begin with. Instead, writes Few, we confuse computer skills with communication skills. "Knowing how to use Excel does not make you a data analyst and knowing how to use PowerPoint does not necessarily make you a communicator."²

Data analysis and communication requires a set of skills that must be learned.³ Yet less than one percent of those who prepare tables and graphs have been trained to design them for effective communication.⁴ And with the array of charting options available in most business intelligence software, it's just as easy to design a good report as a bad one.

"The 'intelligence' of business intelligence resides in people, not machines." writes Few. "What is meaningful in the data must be recognizable to the eyes."⁵

Perception and understanding are linked

Report authors know what is meaningful – trends, exceptions, cause-and-effect relationships. And in most reports the data is there. But the story it needs to tell is not. In building their reports, authors may unknowingly fail to take into account the critical link between perception and understanding: between the way the data is organized on a computer screen and the way it's understood by the people viewing it.

Perception and understanding are inextricably linked. When presented with disparate visual items (in a BI report or elsewhere), the brain automatically tries to make sense of them. It looks for patterns, contrasts, repetitions, and similarities. These shape the conclusions that people make about what they're looking at.

Report authors need to understand this process. Otherwise, they risk building reports that – however accurate the source data – lead their users to the wrong conclusions. "Accidental commonalities in design can easily induce false groupings in the eyes of viewers," writes author and reknowned data visualization expert Edward Tufte. "Viewers [can] mistake decorative tints for real information."⁶

Two quick fixes: keep in simple, show the data

The solution to this problem is better education. But how to begin? By following two simple principles of data presentation: Keep it simple. Show the data. The two famous examples below demonstrate these principles. It should also be noted that both date from the 19^th century – well ahead of any data presentation software, BI or otherwise.

Keep it simple: Dr. Snow's "cholera map"

In 1854, London was hit with an outbreak of cholera. In 10 days, the disease had killed 500 people and officials were at a loss for answers. Six years earlier, London doctor John Snow had proposed that cholera was spread through contaminated water. But he had yet to find proof.

Dr. Snow's Cholera Map

A section of Dr. Snow's famous "cholera map" showing the concentration of cholera incidents near the contaminated pump. Modern health officials use Dr. Snow's innovative technique to understand the spread of diseases within communities and around the world.

With the death toll rising, Dr. Snow took an innovative approach. He plotted the location of cholera deaths on a map of central London, using dots to indicate deaths and crosses to indicate the area's water pumps. With this technique, Dr. Snow observed that cholera deaths were concentrated in the area surrounding the Broad Street water pump. He ordered the handle of the pump removed, thus containing the outbreak and proving his theory.

Show the data: Minard's Napoleon map

In June 1812, Napoleon began his invasion of Russia with 422,000 men. In September, defeated, he began his retreat from Moscow. By December, the army finally left Russian territory. The disastrous campaign had cost more than 400,000 lives. Charles Joseph Minard told the story of the campaign with brutal efficiency with his 1861 map.

Minard's Napoleon map

Minard's map plots six variables on a two-dimensional surface: (1) the size of the army as indicated by the line width; (2,3) the latitude and longitude of its various locations; (4) the direction of its movement; and (5,6) the temperature and dates during the retreat from Moscow.

The width of the band indicates the size of the army at each location. The army's march east begins on the left-hand side at the Polish-Russian border in June 1812. The path of the army's retreat from Moscow is depicted by the darker, lower band, which is linked to a temperature scale and dates at the bottom of the chart. At a mere glance, one can see the path of Napoleon’s retreat and the scope of the losses sustained by his army in the bitter Russian winter.

What history can teach us

What can report authors learn from hundred-year-old maps? By keeping things simple, Dr. Snow's cholera map revealed a previously unknown relationship (cholera and water) that led to direct corrective action (removing the pump handle). By showing a complex interplay of multivariate data, Minard's Napoleon map tells a clear and compelling story.

Cause-and-effect relationships. Direct corrective action. The complex interplay of multivariate data. A clear and compelling story. These are precisely the qualities that people need from their business intelligence.

In a future issue, we'll look at how to balance the principles of good design with IT realities.