-
-
Marketplace
-
Channel Resources
Articles from this Site
-
One brand, one Web site! DM Review is now the home of all the content you're used to at BIReview.com and much more. If you are registered at BIReview.com, you're already registered at DM Review. If not, take just a moment to sign up for all the free services we have for you at the new DMReview.com.
The Big Picture -- A CEO Allegory
Imagine a Global 500 company in which the board of directors has just appointed a new CEO. Among his many priorities, he wants to address the board's perennial interest concerning the ROI of technology, especially the new area of business intelligence, or BI.To understand business issues and to prepare vignettes for the board to help them visualize how BI investments are being used, the new CEO invites each division president to prepare a presentation on a way they used BI to drive effective decisions in the last six to 12 months. He asks them to share: (1) the key data as it was presented, (2) the decision made, and (3) the results of the decision implemented. Furthermore, he asks them to come with live access to the data and their BI analyst to "drive" the technology.
Responding to such a challenge is a double-edged sword. While it clearly gives the division presidents high visibility, the exposure may be less than flattering. The new CEO comes from a background in scientific research, he is known to move smoothly between details and the big picture, and his reputation has preceded him for being searingly incisive in his questioning and analysis. Revealing errors during this process may not ruin a career, but is an embarrassing, short-term blow to one's stature. Years after such an experience, though, it is said that many people have thanked him for the lessons they learned.
As we review one of these presentations, can you spot any sensitive issues before the CEO does? Further, can you predict the plan that our CEO will propose after this process plays out?
The Division President's Presentation
"This graph, prepared by our BI staff, is a powerful example of how technology helps support our decision-making and how it helps us measure the impact of management decisions we have made. A little less than a year ago, our BI manager showed me an earlier version of this graph. It was similar to this except without the data for this past year, shown on the far left of the chart."
Figure 1: The Employee Population Profile (As presented)
"The chart shows the number of employees currently working for us, and the number of years they have been with the company since they were hired. We noticed clearly the trough you see on the right side. Our BI team suggested that the trough probably indicated the departure of employees who had been with us for seven to nine years, and implied that this trough was caused by a lack of promotional and career enhancement opportunities. Using this vivid graphic in support of our management decision-making, we sent a directive to our management, worldwide.
"We asked our management to defer making external new-hires and to promote from within.
- First, this approach would help address the issues identified by our BI team.
- Second, we told them how this deferral of new hiring would provide a major financial boost to the corporation in difficult economic times. Third, we suggested that, if voluntary action was not sufficient during this fiscal year, we would need to consider more formal actions next fiscal year.
"While this fiscal year is not quite over, I am pleased, and proud, to show how our management has responded. Notice the dramatic drop-off in external hiring within the past year, as shown on the left side of this graphic prepared by our BI team.
"Clearly, employees of all educational backgrounds now have improved opportunities for promotion and career enhancement. We can be proud of how our management, worldwide, has responded to this important problem. Moreover, we can be proud of having a talented and dedicated BI staff with the skills and ability to identify an important problem like this, to support our decision-making, and then to help us clearly measure, and present, the results of the management decisions we have made."
The Non-Existent Data
On his way to understanding "the big picture," the CEO began asking detailed questions. "The employees represented in the space between the first two labels on the horizontal axis have been with us more than 'Less than One Year' yet less than 'One Year'; how does one get in that category?"
Realizing the impossibility of this, the division president's pride turned to chagrin as he turned to his senior BI analyst for an explanation. "Well, sir," began the analyst, "our BI tools are very good and we can aggregate and disaggregate our data any way we wish. From a management perspective, it seemed important to understand five major categories of employees, those who have been with us 'in' each of the labeled groupings on the chart [<1, 1-3, 4-6, 7-9, 10+]. So, actually sir, there is no data 'between' those labels on the chart."
Nodding in agreement, the CEO went on: "Then why do you show an inflection point [a point on a curve where it changes direction from going up to going down, and vice versa] just to the right of '1-3 Years' and to the right of '7-9 Years'? How can there be an inflection point where there is no data?"
The Inappropriate Curve
The BI analyst, stumped for a moment, recovered quickly: "Ah, yes, the first version of this chart was with straight lines, like dot-to-dot, but the BI team did not think that was nice enough to bring to the management committee, so we chose the option in our BI software to run a spline curve over the data. Let me show you how bad it looked before the spline curve." Figure 2 is the chart he displayed, which also highlights the fact that only five data points exist for each of the five 'curves.'
Figure 2: The Employee Population Profile (As originally produced)
The CEO winced. He knew his mathematics and he knew that five data points, and other factors in this data set, made it inappropriate to invoke a 'polynomial interpolation algorithm' such as this spline. When asked about this, the analyst replied: "Well, sir, the BI software allowed us to do it, so we assumed it was okay." The CEO jotted down a note: "We must train our employees to be smarter than their software!"
The "Black" Mnemonic
Moving to the whiteboard, our CEO felt impelled to take the first brief step in this training. Down the left side of the board, one bold letter above the other, he wrote: N O I R. He knew this was important for each of his division presidents and staff to understand. He felt that by taking the time to discuss this foundation for effective data visualization he would set an example that would soon spread through the grapevine around the company. "NOIR is important!"
The CEO proceeded to use NOIR to make some powerful points about this presentation, points that no one in the room would ever forget.
"While noir means 'black' in French, in this context NOIR characterizes all data in our corporate databases. Since we invariably make data-based decisions, this is very important to the 'big picture' success of our company. Every number in our company's databases represents a 'measure' on one of four different types of measurement scales. NOIR is a mnemonic for those scales.
- Nominal: A scale where each number represents a name (e.g., divisions in our company).
- Ordinal: A scale where each number represents an "ordering" (e.g., platinum, gold, or silver customers, indicating a customer doing a large, medium, or small amount of business with us).
- Interval: A scale such as degrees Celsius or Fahrenheit with equal intervals but no discrete zero point, as zero on those scales does not mean the absence of heat. We create many interval scales in business.
- Ratio: A scale with equal intervals and a discrete zero, such as: counts of widgets, percent, dollars, pounds, yen, degrees Kelvin. We can divide one number on these scales by another to form a ratio."
The CEO as Corporate Leader
The best leaders know they must lead by setting an example. Taking the time to present this foundation for effective data visualization, and being conversant in all of its nuances, clearly set an example that understanding NOIR was important in this company. It's not often that one sees a CEO acting more like a statistics professor, but, as he will clarify and embellish shortly, he was delivering the message that the use of NOIR is strategic and critical to 'the big picture' success of this company.
The Employee Population Chart in the Context of NOIR
When we look back at Figure 1, we see that three dimensions are presented. The "Number of Employees" is measured on a ratio scale, while "Employee Education" and "Years with Company" are ordinal scales. Yet, there is no redeeming value for "Years with Company" to have only five ordinal "buckets."
"Longevity with the company" is in the BI database to the nearest day. Over a period of 20 years, we would thus have a ratio scale of over 7,300 days. Since there are not enough pixels across the screen to plot over 7,300 days, we do need to "bucketize" the data. If we bucketize into equal intervals of one bucket per month we will have 240 buckets, while one bucket per year gives 20 buckets. After we "intervalize" the scale this way, we observe that there is a discrete zero. We can say that 20 months or years is twice 10 months or years, so it may still be considered a ratio scale. Furthermore, we also have an appropriate scale, and enough data points, to invoke a spline curve, if desired.
The CEO asked the BI analyst to produce such a chart, showing "Years with Company" with a value for each of the past 20 years. While the analyst worked deftly at his keyboard, the CEO pointed out that "ratio" data is very "rich" and while we may often have to intervalize it, or more rarely ordinalize it, we may lose something with each translation. As the following analysis of the intervalized data in Figure 3 illustrates, with the ordinalized data of Figure 1 we lost any ability to compare annual levels of hiring.
Figure 3: The 20-Year Employee Population Profile (On an Interval Scale)
The "Beached Whale" is a Fluke
Based on the same raw data, Figures 1 and 3 do not at all look similar. Figure 1 has the silhouette of a "beached whale" because of the vagaries of its arbitrary ordinal buckets of "Years with Company." While the central three buckets in Figure 1 are equal intervals (1-3, 4-6, 7-9) the much smaller bucket on the left and much larger bucket on the right, are ordinal. The fluke of the whale on the far right is caused by adding all the values for years 10 through 20 and plotting them at just one point. The sum of the employees hired in the 11 years from 10 to 20 is much greater than the sum of those hired in the three years from 7 to 9. Thus, the "trough" noticed in the initial analysis is a "fluke" -- a completely accidental creation, caused by the capricious aggregation of the data in to ordinal buckets, with a much larger bucket on the right.
The Management Problem is Fictitious; the Cause is Spurious; and the Result is Bogus
The ordinal bucket at the right of Figure 1 caused management to chase a completely fictitious problem. In an earnest effort to come up with a solution for this non-existent problem a plausible yet completely spurious, cause was identified. Thus, the problem and its supposed cause have no relation to reality.
Perhaps it can be argued that the "solution" of holding down external hiring and promoting from within was useful to the company anyway. If so, the visual representation of that successful outcome would be to see the line representing the number of employees drop down very low for the past year, just as Figure 1 had supported that wishful thinking.
Figure 4 shows how management actually executed the proposed solution, comparing part of Figure 1 on the left with part of Figure 3 on the right. The white arrow on the lower right of Figure 4 shows how the number of employees hired 1, 2, and 3 years ago add up to 8,170 employees. The red arrow at the top shows how this agrees with Figure 1 where only the three-year aggregation is shown. The precipitous decrease is an artifact. It is due solely to the fact that it links the sum of three years' hiring with less than one year's hiring. Understanding NOIR, we realize this is an artifact of visually linking two ordinal buckets, which should not be done. In reality, as shown in Figure 3, during the past year hiring was 25 percent greater than in the previous year (3,750 versus 3,000).
Figure 4: The Solution: As Presented [Fig.1 on left] vs. Reality [Fig 3. on right]
The approach that the CEO took to understand this data presentation has shown that the management problem was entirely fictitious; the supposed cause was spurious; and the claimed result was bogus. Yet, it revealed a more strategic flaw: the organization did not have a common language of visual communication. The situation existed because the division president and staff did not have a mastery of graphicacy skills. Such skills would have prevented this situation in the first place, or at least enabled the management team to effectively spot the issues before they caused trouble. Without a common language, they might continue to misunderstand important data, or waste more time chasing issues that were not real.
Building a Shared Vision for the Future
The CEO wanted to end this meeting on a high note with a positive vision for the future. "This is not the first division in which I've seen this problem. We have made a substantial investment in business intelligence tools, and their potential to drive our success is enormous. These tools, though, are only as powerful as the people using them. Now, we are going to invest in our people -- our human capital -- to turn the power of our BI technology into a clear competitive advantage in the marketplace.
"With some training, starting from the top of the organization and moving on down, we will all be as fluent in our graphicacy skills as we are in our literacy skills -- for graphicacy is a critical component of business intelligence. Today's discussion of NOIR set a foundation for concepts in graphicacy, and was the first step to enhancing our ability to understand the reality in our data. We are going to be number one in our industry in the way we use data visualization for effectively reaping a great return on our investments in BI technology."
Howard A. Spielman, M.B.A., Ph.D., President of Management Semiotics International Inc., can be reached at HASpielman@ManagementSemiotics.com.
For more information on related topics, visit the following channels:
