Identification of holistic patterns is crucial when studying several pertinent Operations and Supply Chain Management (O&SCM) issues. If we view each business practice as a single node, then a firm’s overall operation can be conceived as a system of interconnected nodes. The abundance of business practices and the complex interactions among them present challenges in detecting ways by which firms can attain competitive advantage. This challenge is further accentuated by the fact that it may not be a single business practice, but the configuration of a set of practices that drives firm performance. Firms grapple with ways to transform themselves into high-performing organizations. How do high-performing firms differ from low-performing firms? Specifically, how do these firms differ in the amount of connectivity among practices? Which practices are central to a high performing firm? What configurations of practices are salient in high-performing firms but not in low-performing firms? Answers to these questions pinpoint the core strengths of a high-performing firm and provide guidelines to align a firm’s investment with the target areas for improvement. Finding empirical answers to the above questions is a challenging task. One difficulty is caused by the lack of observations to properly model the complex interactions among a large list of variables. Because firms operate as complex systems made up of copious interconnected business practices, to properly model these business practices we need to analyze datasets that have high dimensions (i.e., containing a large number of variables). However, in order to analyze a high-dimensional dataset, the conventional data analysis methods require a large sample size. Ironically, O&SCM research has historically faced the issue of low response rate, leading to low number of empirical observations. This lack of sufficient sample size is especially problematic when examining complex relationships among a wide range of variables in searching for the right configurations of capabilities. Thus, it is critical to find a technique that can work with a relatively small sample size and still be able to derive meaningful business insights from a high-dimensional dataset. Another problem with small sample size is that it is often sensitive to violations of the normal distribution assumption. This violation of normality assumption is common in empirical datasets and is more problematic in small samples. Therefore, it is important to find a tool that is robust against violation of normality, or to provide a remedy that overcomes this challenge. In a study we use the Sparse Inverse Covariance Estimation (SICE) method, in conjunction with bootstrapping technique, to address the above problems.