Using Impact Analysis to Drive Process Assessment and Improvement

• Main Authors: Beale, Michael A., Franklin, Paul H.
• Format: Conference Proceeding
• Language: English
• Subjects: Big Data; Business; Current measurement; Decision Support; Decision support systems; Impact Analysis; Lessons learned; Machine learning; Reliability engineering
• ISSN: 2577-0993
• DOI: 10.1109/RAM.2018.8462998

Impact Analysis is a method to measure and illustrate the drivers of complex failure and resulting impact on decisions for decision support. Before decision makers can enjoy highly correlated illustrations confirming their decisions the bands of performance based on failure must be confidently defined. Defining failure is the first step of achieving confident decision support for engineers ensuring the resulting visuals will be impacted as designed. The confirmation of a failure is achieved within the final step of Impact Analysis called impact reporting. Impact reporting results in simple statements easily understood and acted upon by decision makers and engineers. As an example, illustrating that 1 % of a given universe is responsible for 15% of failure. Providing the ability to show how it grew over time and the other related elements to the performance. The variance between 15 and 1 % is what we call "impact", In this example it would be overstated by 14% and ordered higher than the lesser impacted universe. The many attributes of failure can be captured with a single related criterion or leaf level optimized criterion. As an example, a single complex failure criterion could be a combination of low or high performance thresholds being exceeded across many metrics and behaviors over time. As failure is defined business rules and complex logic are mitigated independently by engineers and decision owners once and then collectively resulting in a "validated" single encapsulated failure. By encapsulating the many thresholds of failure into a single measure allows complex business rules to naturally move efficiently out into new processes without the risk of redefining business rules and overall failure. Once failure is defined its data signature is applied to past data histories with the help of new data shaping technologies that split the data into bands of performance, optimal/prefailure/failurel/critical across the optimal time period increments and data granularity. Impact Analysis compares two or more performance bands over time that are independently measured and plotted horizontally (over time, X-axis to itself per period) and vertically (to other performance bands within the same time period, Y-axis). With the ability of big data and machine learning techniques to better predict and define failure reliability engineering can use Impact Analysis to encapsulate their quality metrics and knowledge and greatly aid in the machine learning process and ult