Readers of this blog will know that I frequently find fault with published infographics, but today it is the turn of Microsoft’s Power BI team to be the subject of my criticism.
A colleague alerted me to the online demo: http://www.microsoft.com/en-us/powerBI/solutions/demo/business-demos.aspx and in particular the one in the section labelled “Data Visualization”, which contains three charts collectively labelled “Profit Margin Analysis”. I will comment on the two lower charts in later posts because there is quite enough to say about the topmost chart.
This chart is a scatter plot, showing “Profit” on the vertical axis and “Cost” on the horizontal axis for a portfolio of “Departments”. The only way to determine the “Sales” for each department is to hover your mouse over each label and then read off the value from the tooltip. Coincidentally, the Sales figure for each department has exactly the same value (the extremely over-precise 3,097,642.34). Some of the bubbles are labelled with their department number and others are not – it is very difficult to see which label relates to each bubble. The axes are labelled with the word “(Thousands)” and the numbers also have the suffix k, an undesirable tautology.
I am afraid that I cannot see any analytical value in this chart at all as it stands, because I cannot determine the Sales or Margin of each department. So I extracted the data into the Metapraxis visualization software tool (Empower) to try to do a better job. I hope that most people know that the broad definition of Profit is Sales minus Costs, but this relationship does not hold true in the data set provided. I therefore ignored the Sales value (the aforementioned 3,097,642.34 for each department) and calculated a new Sales value for each department, as Profit + Costs. I then calculated Margin for each department (Profit over Sales, as a percentage). The scatter plot I constructed is below.
Sales is plotted on the horizontal axis and margin on the vertical axis. Multiplying these two items together gives us Profit, and we can join points on the graph where Profit is equal with a series of “isoquants” (similar to contour lines on a topographical map or pressure lines on a weather map). Each bubble is clearly labelled and the axis labelling is without redundancy. Numbers are not stated to an unnecessary level of precision. The new chart provides us with some useful insights into the data, including the following:
- D15 has the highest Margin % and the highest Profit (but not the highest Sales)
- D18 has the lowest Sales and the lowest Profit (but not the lowest Margin %)
- D4 and D12 both generate about $200k of profit, but D4 does so with higher Sales and lower Margin % than D12
- D14 is ranked 12th in Sales but has the lowest Gross Margin % and the second lowest Profit
I would be interested to know whether readers share my criticisms of the original graph and my enthusiasm for my own rendition of the data.