Bad Charts... and Insights From Better Ones
I first met Charley Kyd in the late 1990s when we were both enamored with TM/1, the (still excellent) OLAP software tool now owned by IBM and known as Cognos TM1. Charley is definitely an Excel master and writes frequently about spreadsheet reporting technology and issues. Recently he released a piece called Bad Chart! Bad, Bad Chart! about a particularly poor graphic the Wall Street Journal published. The offending chart in full was this:
The bottom part of the chart is the egregious bit. Here it is, a bit larger:
Charley correctly nails the essence of the problem with this chart:
We use charts to help our readers understand numeric data as quickly and easily as possible. Designed properly, charts offer the closest way we have of transmitting business data directly to our readers’ brains.
But these charts fail to do that. The primary reason for this failure is that the human mind can’t compare areas easily. To illustrate, look at the red square above. When you glance at the chart—not the numbers—do you immediately notice that the dark square is about three times the area of the light square? I certainly don’t.
The WSJ editors must have agreed with this concern, because they felt it was necessary to plaster the actual number of employees on the face of each square. In short, the charts merely decorate the numbers shown; they're junk charts!
Charley then goes on to suggest this as an improved version of the chart:
I certainly agree that this is far better than the WSJ’s original chart. Still, something was missing for me. Consider this chart:
If our goal is to transmit information to the reader’s brain, as Charley suggests, this transmits more information more quickly in my opinion. It says:
- IBM is by far the largest employer, though HP is catching up. I mean, it’s literally above all the others, the first one you see in the chart.
- Oracle and Microsoft are about the same size, and much smaller than HP and IBM.
When I first looked at the chart I noticed that Oracle started below Microsoft and finished above it, so it must have a much higher growth rate. When I plotted out the growth rates in a single chart, I found this:
These previous charts have focused on employee counts and how those have changed over time. The original Wall Street Journal chart also noted the annual revenue of each company. It’s interesting that IBM, HP, and Microsoft are all within range of each other in revenue terms (around $100B/year now) and Oracle is much smaller.
I did a little research and found the revenue back in 2003 as well. And if we have employees and revenue, then we’re just one operation away from one of my favorite metrics: revenue per headcount. Let’s graph that:
While we’re at it, I also graphed the change in revenue per headcount:
I’ve written about revenue per headcount before, back in March 2003 and December 2011. In the more recent newsletter of the two we looked at a histogram rev/HC in large organizations and concluded that it’s rare to see $1M rev/HC in a non-capital intensive business. I am extremely surprised to see Microsoft growing at a healthy rate while increasing revenue per headcount by 50%. Normally I expect growth vs. margin tradeoffs where you get more revenue albeit at a lower margin, and my speculation is that’s what has happened at HP and IBM since 2003.
In the end, though, revenue per headcount is only one metric of many. Note that we have not touched on EBITDA margins or cash flow or the stock price or dividends, so we’re pretty far from having a comprehensive view of value drivers. Nonetheless, at least we teased some good learnings and further questions by starting with a truly horrible chart.
|