Home > Articles > Security > General Security and Privacy

  • Print
  • + Share This
This chapter is from the book

Waterfall Charts

Popularized by McKinsey and Company, the "waterfall chart" provides a flashier alternative to the stacked bar chart. It is best used in relatively simple exhibits where the analyst is trying to show the relative contributions of different factors to a larger total. For example, in the past I have used waterfall charts with executive audiences to illustrate how particular categories of security vulnerabilities contribute to an overall risk score.

A typical waterfall chart (see Figure 6-6) contains the total number at the top, represented as a horizontal bar. Bars arrayed underneath "explode" the component numbers onto separate rows. Numbers for each row appear to the right of the bar. A dashed line typically separates each bar.

Figure 6-6

Figure 6-6 Waterfall Chart

Waterfall charts tend to be more readable and less claustrophobic (or space-efficient, depending on your point of view) than bar charts. Waterfall charts look neater, and facilitate comparisons better, than the equivalent stacked bar chart.

A side benefit of waterfall charts is that they can be used in small-multiple exhibits, but only when done with care, and with multiples of perhaps two at most. Figure 6-7 shows a "side-by-side" small-multiple variation of the waterfall chart that looks good and tells its story well.

Figure 6-7

Figure 6-7 Small-Multiple Waterfall Chart

Few software packages exist that can create waterfall charts; in most cases, analysts must hand-draw them using a graphics package like Visio, ConceptDraw, Adobe Illustrator, or (in a pinch) a presentation package such as PowerPoint. As an alternative to hand drawing, waterfall charts are also good candidates for automation. A few lines of Perl or Java, for example, can easily generate vector graphics for waterfall charts.

Waterfall charts are good for high-end management presentations, but that is about it. Their legibility degrades quickly after about a dozen rows. For data sets that are highly dense, consider treemaps instead. I discuss treemaps later in this chapter.

  • + Share This
  • 🔖 Save To Your Account