Sparseness and Completeness: Simplifying Bowties to Improve Understanding and Predictive Power

Abstract

Abstract The paper discusses the use of mathematical and methodological tools to simplify bowties and create understanding of which barriers will be effective for unidentified threats. This paper further develops Bow-Tie concepts by applying the FAME methodology (SPE-190633) to the bowtie ‘levels analysis’ (SPE 127180). The ‘levels analysis’ defines the barriers at the various levels of a bowtie as the workplace (Level 1, L1), organisational (Level 2, L2), and cultural and regulatory factors (Level 3, L3). Sparseness: When L1 contains only the controls appropriate it becomes clear that though the number of threats acting on the top event appear to be very high, threats can be bundled into generalised threats that are controlled by the same L1 controls. Completeness: For management the most important information in a bowtie are the escalation controls in L2. In practice the number of L2 controls contained within a specific bowtie are a small set. This means that once sufficiently many L1 controls is identified, analysis of escalation factors and associated controls uncovers the full set of L2 controls. Sparseness: While it might appear that generating a large enough set of threats covering the entire risk space to which an operation is exposed will demand the investment of significant time and effort, the actual number required is much lower and thus much more efficient. This sparseness approach requires the practitioner to question whether the threats being added to the bowtie analysis are truly separate areas in the risk space or if they are actually overlapping areas that are covered by the same controls. Completeness: The completeness of understanding of the associated L2 controls does not require completeness of understanding of all the L2 controls. The General Failure Types of the original Tripod model, that led to the Swiss Cheese model, now called Basic Risk Factors, had 11 members intended to cover the whole range of organizational latent conditions. These can serve as categories, but other taxonomies appropriate to the organization and activities as long as they cover the whole range of organizational activities will be just as good. This paper develops two tools to improve the usability and manageability of bowties. It provides mathematical tools to provide understanding of the remaining uncertainty in the bowtie creation process.