All industries need to understand the demand for their product and how that demand may be distributed over time. They need to adequately meet that demand - and possibly enhance or adjust it to their advantage. The electric power industry is in the unusual position of being expected to supply all that is demanded of it, on a practically instantaneous basis, yet it often has no easy way to store excess supply.
This problem is not new, but we can learn much from observing they way it was addressed by two major power companies in 1935 using data visualizations. With over 70 years of hindsight, we can gain insights into how they may, or may not, have strategically evaluated the technology options available to them.
The "Day-by-Day" Method
On January 2, 1935, a craftsman in the Commonwealth Edison Company in Chicago, Illinois, went down to the workshop and took a one-eighth inch thick sheet of three-ply basswood that was 11 inches wide and 17 inches long. This was the technology for building a visual model of business data. The width represented the 24 hours in the day. Midnight, December 31 - New Year's Eve - was on the left side and midnight, January 1, was on the right side. At 10-minute intervals across this 11-inch horizontal axis, the number of kilowatts of load was plotted on the vertical 17-inch axis. This "Power Demand Curve" for January 1, 1935, was then cut out of the basswood on a scroll saw. This may have been 11×17 inches to match a standard double-sheet of quadrille-ruled graph paper of that dimension, which was glued onto it.
Every day this same process was repeated and the basswood curve for the succeeding day was layered behind the previous curves. Figure 1 shows the 3-D data visualization that was produced at the end of the year. This visualization, in its glass case, was nearly five feet long and weighed about 300 pounds. The 24 hours in the day are labeled across the short end of the case and the 12 months are labeled across the long side of the display, with dark vertical lines indicating the weeks. The horizontal lines across the long side indicate the number of kilowatts of load.
Figure 1: Commonwealth Edison Co. Chicago, IL, 1935 System Load
Two basic dimensions are represented here: time and kilowatt-hours. Since the time dimension is cyclical, with 24 hours repeating every day, it is here split in to two dimensions: hours and days. As few other dimensions are similarly cyclical, the "time" dimension offers interesting problems for data warehousing, yet very interesting opportunities for data visualization. Now, this 3-D model represents three distinct dimensions: hours, days (which are easily labeled in weeks and months), and kilowatt-hours.
Today, in the early 21st century, we often take visualizations like this for granted; but in 1935 this was a powerful representation of a "virtual reality" space. Over 70 years ago, science and engineering books were replete with three-dimensional visualizations of physical phenomena (e.g., the relationship between the pressure, temperature, and volume of a gas). But, placing "24-hours-in-a-day" orthogonal, at right angles, to the "365-days-in-a-year," as on the base-plane of Figure 1, creates a non-physical space ... a "virtual reality" space with "time" on two of these dimensions. Such a 3-D virtual reality space was very useful, and many companies invested in such models.
To collect and plot this data by hand, and then to carefully cut the curve on the scroll saw, may well have taken a skilled employee about a quarter of a workday - every day of the year. This "technology" was an expensive process and other companies - in an effort to harness the conceptual power of this visualization model - tried to find ways to save money and get a similar product for business analytics.
The "Annual Retrospective" Method
Figure 2 presents the model created by Detroit Edison for their system load in 1935. It looks very similar, at first glance, but it is constructed differently. After collecting the data for the entire year, each day was divided in to 48 periods, each one-half hour long. Then, 48 curves were cut out, each representing the power load for that half-hour period across all 365 days of that year. While the Commonwealth Edison model was created by sandwiching 365 curves across the short axis, the Detroit Edison model was created by sandwiching only 48 curves across the long axis.
Figure 2: Detroit Edison Co. Detroit, MI, 1935 System Load
The Strategic Value
It is important for management to identify the strategic value of this visualization model in the first place, as well as what additional value it may offer or what unintended consequences may arise from its creation. The "technology" decision here was the construction technique for a visualization model. The limiting factor was the cost of the craftsman. Cutting 48 curves per year is 13 percent of the work involved in cutting 365 curves per year. While it may seem like a no-brainer to go from .25 of a person-year of labor to .03 of a person-year, we can ask what strategic opportunities may be missed in this decision on the use of technology.
Once one understands the "virtual reality" space of these models, they both offer a general overview of the power load over the course of a year. Both present the annual production of the enterprise and would make a great conversation piece in the corporate boardroom. The cyclicality of the demand over each day and week are evident as is the seasonality. This is fine for a retrospective view; but how can the enterprise strategically prepare for the future?
In both of the electricity load models for 1935, it is clear that there is a higher demand for electricity in the winter, for heating and lighting in the long evenings, than in summer. Yet, major changes were afoot. In 1931, 1 million household refrigerators were manufactured in the U.S. and this number was almost doubling every two years thereafter. In 1924, the first department store ever equipped with air conditioning was Detroit's J.L. Hudson's Department Store, and more stores and office buildings soon followed. These trends would surely cause major strategic changes in this industry, and these load models offer different capabilities to support this strategic decision-making.
