At a recent Business Intelligence trade show, the conversation turned to the origins of data visualization and I asked when people thought the discipline first took root. A vendor marketing representative in her 20s responded "Oh, a long time ago, probably back around '85." Standing next to her, the vendor's much older CFO also remembered charts being made back then. Both were surprised when I told them they had the last two years right, but had missed by a couple of centuries!
William Playfair, the unsung hero of bar, line, and pie charts, engraved the first of them in 1785 for his Commercial and Political Atlas, published the following year. That first edition was mostly line charts, with one bar chart, and in 1801 he published the first pie chart. Most people who use 'chart wizards' on their computers today are astounded to learn of the long history of data visualization. There are lessons to be learned that go back just as far.
A 220 Year-Old Data Visualization Tells a Powerful Story for Today
For the 12 years from 1770 to 1782, the imports and exports of England to and from all of North America are shown in Figure 1. The copper plate engraving was hand-painted 220 years ago to show the line of Exports to America in red and the line of Imports from America in yellow. Where the exports exceed the imports, there is a "Balance in Favour of England" colored blue, and where the imports are larger there is a "Balance Against England" colored pink.
Figure 1
The layout of the axes of this chart, the labeling, and the use of color, are paradigms of best practice for today. If it wasn't for the 18th century engraved text font, it might almost seem as though it came out of a contemporary computer.
The modern annotation identifying the Boston Tea Party shows the impact of that event. With a time lag for 'real-time data-transfer' back across the ocean, England began to stop exporting to America and the red line tumbles. (Why ship cargo to America and get it dumped in the harbor?) Yet English businessmen began stocking up on raw materials, and the yellow line picks up quickly. Thus, the previously favorable balance of trade now goes against England.
After Paul Revere's ride, and the outbreak of the Revolutionary War, the chart shows this trend reversing. The yellow line of imports from America drops to a base level of trade with Canada, while the red line of exports to America gets higher. This raises the question of what the English could have been shipping, for which the colonists in America would have been paying.
It seems the data came from the Custom House records and reflect the value of all goods on board a ship heading westward across the ocean. If this had included gunpowder and cannonballs, then the colonists would surely not have paid for this. The line of exports may show a fair value for the exports, but since they were being paid for by the National Debt of England - and not the colonists - there is no "Balance [of Trade] in Favour of England." This is an excellent example of what today we call a problem with the 'data dictionary' - a clear definition of the meaning of all data terms. These types of data definition problems are not due to 21st century data warehousing. They have been plaguing businesses, and nations, for centuries. Such a bold example should inspire us to seek best practices to address them.
High-Tech 'Rollover'
William Playfair gathered the data for the Expense of the Ordnance as displayed in Figure 2. Careful examination indicates two features that one might think are recent innovations. His horizontal lines near the top of the chart are annotations to indicate periods when England was at war. Such a graphic annotation may be used today on time-series charts to indicate extraordinary events such as the duration of a labor dispute, natural disaster, or material shortage.
Figure 2
It is even more surprising to see what he does for 1782. The expense of the ordnance is 'off the chart,' so he runs it up outside his chart border (see modern arrow at top) and rolls over to rise up from the bottom of his virtual screen. When shown to today's programmers, they roar with delight at this example of what they call 'rollover'-on the virtual computer screen of nearly 220 years ago.
Data Visualization Cannot Make Some Problems Go Away
William Playfair published five editions of these books between 1786 and 1801. One of his purposes was to focus on the ever-increasing National Debt of England-amazingly, over a time horizon of 100 years. Figure 3 shows charts from his first edition and last edition side-by-side illustrating how the National Debt continued to climb. The double-headed arrow compares the annotated points when the "American War Ended." The horizontal scale has shrunk by about 15 percent to fit the 15 extra years, but the vertical scale has shrunk even more-to fit a percent increase in the National Debt!
Figure 3
Looking back over 200 years, we can see that good data visualizations require clean data with a good data dictionary. Furthermore, while data visualizations cannot solve problems alone, they certainly can highlight the issues to encourage effective solutions.
Howard A. Spielman, M.B.A., Ph.D., President of Management Semiotics International Inc., can be reached at HASpielman@ManagementSemiotics.com.
