Rule Cubes for Causal Investigations

With the complexity of modern vehicles tremendously increasing, quality engineers play a key role within today´s automotive industry. Field data analysis supports corrective actions in development, production and after sales support. We decompose the requirements and show that association rules, being a popular approach to generating ex- planative models, still exhibit shortcomings. Recently proposed interactive rule cubes are a promising alternative. We extend this work by introducing a way of intuitively visualizing and meaningfully ranking them. Moreover, we present methods to interactively factorize a problem and validate hypotheses by ranking patterns based on expectations, and by browsing a cube-based network of related influences. All this is currently in use as an interactive tool for warranty data analysis in the automotive industry. A real-world case study shows how engineers successfully use it in identifying root causes of quality issues.