(167c) Addressing Issues in the Design and Use of Risk Matrices in Process Safety

Risk matrices are widely used in process safety to rank risks posed by processes. For example, commonly they are used in process hazard analysis to rank the risk of hazard scenarios and determine the need for risk reduction. There are no industry standards for risk matrices and most companies develop their own. Risk matrices are deceptively simple but their design and use are susceptible to pitfalls for unwary users. Some of these pitfalls are not obvious and invalid risk rankings can be produced without being recognized by users.

Flaws in the theoretical framework of risk matrices and mathematical inconsistencies in their use have been identified in the risk analysis literature. These issues pose potentially serious problems for the application of risk matrices in process safety. The issues are described in this paper together with approaches to address them.

Risk matrices use estimates of event severity and likelihood levels to identify risk levels. In order to construct a risk matrix, it is necessary to specify the number of severity and likelihood levels, the range of severity or likelihood covered by each level, and the number and assignments of risk levels. Care must be exercised to produce risk matrices that facilitate and not complicate risk ranking. For example, they must allow for ready risk discrimination. Too few or too many levels create difficulties. Also, risk reduction measures specified for different risk levels must be consistent with the underlying risk values. Guidelines for constructing risk matrices that address these and other issues are provided in this paper.

Many companies calibrate risk matrices with reference to process safety target levels for facilities. Pitfalls exist in calibration, for example, the need to calibrate matrices for both individual and group risk, and for each process, rather than incorrectly using a single matrix in all cases. These pitfalls are described and guidance is provided to ensure calibration is performed correctly. This paper also provides guidelines for using risk matrices that cover issues such as consistency, subjectivity, and bias.

The correct and incorrect use of risk matrices are illustrated with examples.