Even though the term 'Business Intelligence' is barely 15 years old, BI, as a function, has been practiced for well over a century. By looking back 100 years, we gain a very important perspective on what we do today. Furthermore, we can ask: How much better (or worse?) is our BI today?
A robust definition of BI ... relevant for today and at least the last century:
Since many definitions of BI rely on 21st century technical terminology, we need an understanding that is robust enough to apply to the business environment of 100 years ago. Thus, we could plausibly assert the following definition:
BI is an interactive process for exploring and analyzing information to discern trends or patterns, thereby deriving insights and drawing conclusions. The BI process includes communicating findings and effecting change. The domains of BI may include customers, suppliers, products, services, and competitors.
Achieving this goal was just as important for management in the 19th century as it is today. By the turn of the 20th century, effective 'BI' was considered a critical competitive advantage, just as it is today. Beginning in the railroad and steel industries, it quickly spread through engineering consultancies to the nascent automobile and electrical power industries. By 1908, even dairies were plotting milk production curves of individual cows, and by 1913, the term "Executive Control Curves" was in common use.
The development of a functional 'BI' system around 1900
As accounting became more standardized in the early 1900s, the data for a functional BI system became available. Management found, though, that they needed more than what they perceived to be offered by the accounting data-only the bird's-eye view of the whole business at any one period of time. They needed a cross-index of that accountant's information, so they could see over a longer period of time the whole history of the business across different financial metrics. In response to this need, graphic methods for presenting these facts were developed, and increasingly adopted, in the first decade of the 20th century.
By 1914, the Westinghouse Electric and Manufacturing Company of Pittsburgh, then an up-and-coming industrial powerhouse in the American economy, regularly plotted about 4,000 of these graphic 'BI' curves which recorded the activities of all departments of the business. The majority of the curves had one more point added each month, but some of the curves were updated on a weekly basis. Another company, then an engineering consultancy (and still in business today), was plotting 8,000 curves in 1914, and had a major influence in bringing the use of this functional 'BI' approach to many American factories and to the construction of the Panama Canal.
Figure 1 shows a 4 x 6 card specially prepared for these 'BI' curves. At first glance this may look quaint, by today's multi-color/multi-font PowerPoint standards. It actually has some features, though, that are not often found in today's BI systems. The BI features in this card system are highlighted below.
Figure 1
The features of a functional 'BI' system in the early 1900s
Customized for long-term archival storage and the then-current copying technology:
The quadrille was pre-printed on the lower right of the Curve Card with 11 horizontal lines and 13 vertical lines. A special card-stock was selected to be "guaranteed absolutely against any deterioration for twenty years." Furthermore, it was transparent enough to allow the later copying of the card, after data and labels had been added with India ink, with the 1910 copying technology of blue-printing.- In today's BI: Is your data store "guaranteed absolutely against any deterioration for 20 years?"
Numeric data table with sub-totals presented:
The sales, valued in dollars, of automobiles produced at one plant, are plotted on the Curve Card in Figure 1 for fiscal year 1911. Each month is numbered across the bottom, and the data for each month is written at the top. The quarterly sums are also shown for the convenience of management.- In today's BI: When viewing a data visualization, can you easily get at the underlying numbers and also cyclical totals?
Thirteen data points per fiscal year:
In Figure 1, while the fiscal year began in August 1910 (month 8), the data point for the previous month is also plotted, to give a context for whether August is beginning to trend up or down. This previous-month's data is redundant to what was presented as the last data point on the previous-year's chart. This is why 13 vertical lines were specifically included. With 21st century technology, such redundancy is rarely tolerated. Thus, management's visual perception today is often handicapped, and in this way the 100-year-old format has much to commend it.- In today's BI: Do your data visualizations give a context for how initial data points are trending?
Desired trend indicator:
On the right side of the chart there is an upward pointing arrow. The specially prepared Curve Card only had the shank of the arrow pre-printed. When setting up the card for a particular business metric, thought was given to the direction the curve should appropriately trend, and then an arrow-head was drawn on the shank. For example, management would desire 'total sales' to trend up, and 'expenses per unit of output' to trend down. Fluctuations in some metrics, such as 'payroll,' mean nothing unless the underlying conditions were known. For that metric, an increase in productivity would lower payroll, a desirable goal; yet, a decrease in product demand might also lower payroll, an undesirable goal. Thus, there would be no arrow-head on a payroll chart.- In today's BI: At one quick glance can you see a confirming indicator of the desired trend?
Consistent chart format for a particular business metric:
There is a serial number in the upper left corner of each Curve Card. This unique code identifies the metric being presented. From month-to-month, quarter-to-quarter, or year-to-year, management would always be looking at the same data presentation format for the same metric. They would get to be comfortable with what to expect in these consistent chart formats, and what not to expect. They would not be exposed, as is often the case today, to the creative presentation techniques of every analyst who wished to show how well they could tweak the chart engine of their BI software.
