Balancing multiple performance properties while minimizing experimental effort is a common challenge in industrial R&D. Bayesian optimization provides a powerful framework for adaptive, sequential experimentation in complex, nonlinear response spaces. This presentation evaluates the Bayesian Optimization platform in JMP Pro using a realistic formulation simulation developed for PPG’s Sigma Logic™ experimental design training program.

Sigma Logic, launched in 1998, trains approximately 130 global research associates annually through beginner- and advanced‑level courses built on JMP. A team‑based competition using a simulated system is a hallmark of the advanced course, and the simulation evaluated here served as the focus for 16 teams across three courses. The current formulation challenge includes 12 controllable factors and four critical performance metrics, with the objective of meeting performance targets at Six Sigma quality and minimal cost within a constrained experimental budget.

Multiple optimization campaigns were conducted in JMP Pro, each initialized with a standard screening design and followed by successive rounds of Bayesian‑recommended experimental runs. Optimization progress was tracked across campaigns and compared with traditional sequential DOE strategies employed by course participants. Results show that Bayesian optimization efficiently navigated the nonlinear response surface and identified acceptable solutions in substantially fewer runs. The presentation concludes with practical learnings on integrating Bayesian optimization into experimentation workflows based on JMP to accelerate formulation development in industrial research.