Improving Risk Matrices: Colors, Not Numbers (Tip #2)

The source of those great 1-to-5 and 1-to-9 scales we use in our risk matrices are traced to Rensis Likert. In the early 1900s, he and other psychologists were seeking “the solution of a technical problem which has arisen in relation to the quantitative aspects of the study of social attitudes.” Likert explored different statistical methods of analyzing his scales and measures. He concluded that the “clustering method,” based on ordering and ranks, performed just as well as the “sigma method”, based on parametric statistics assuming continuous data.

Likert scales – originally 1= strongly approve, 2 = approve, 3 = undecided, 4 = disapprove, 5 = strongly disapprove – appear in various 5-, 7-, 9-, and 10-point versions. Likert scales are used today in everything from risk assessments to whether you liked your meal or hotel.

SS Stevens in the early 1940s would clarify, after seven years of committee study on how to quantify human sensations, that there were four basic types of scales – nominal, ordinal, interval, and ratio. For ordinal (Likert) scales, the appropriate empirical operation is “greater or less than” and the permissible statistics are “median and percentile.” Only ratio scales, based on continuous data, are suitable for classic (parametric) statistics.

In layman’s terms, standard deviation and reporting Likert scales to decimal places are not appropriate. “Poor”, “Red”, and “5” mean the same thing.  “Good”, “Green”, and “1” mean the same thing. There is no 1.2, just as there is no “green and a tad less” or “Good and a little less.” There is no 3.59 just as there is no “yellow.59.”

Tip #2 is to stop referring to numbers when communicating with risk matrices. Call both the consequences and the likelihood components by their descriptive condition or call them by their color. If not calling them by numbers makes the whole exercise look a little less quantitative, it is because it is not quantitative. Use labels, not numbers.