In high-precision optics manufacturing, dual-component module assembly faces significant yield challenges due to the complex interplay between component tolerances and final performance. Current inefficiencies stem from poor pairing and uncertain specification limits, necessitating a resource-intensive "mix-and-match" bottleneck that results in substantial labor and material losses.

This paper presents a data-driven solution using multivariable regression (MVR) modeling to establish a transfer function. By defining module key performance output variables (KPOV) as a function of component key performance input variables (KPIV), the model identifies critical to quality (CTQ) drivers and their specific sensitivities, enabling precise performance prediction prior to physical assembly.

The implementation of  virtual component selection facilitates strategic digital pairing to ensure compliance with performance targets. This approach successfully resolved the physical "mix-and-match" bottleneck, yielding an operational efficiency gain of 30 labor hours and a financial reduction of $11,000 per week. Furthermore, the model directly improved both component and module yield while enhancing overall module performance. It also enabled the refinement of specification limits based on empirical sensitivity analysis rather than theoretical estimates.

By transitioning from reactive testing to proactive virtual selection, this scalable framework provides a robust methodology for improving assembly yields and performance. This approach effectively reduces operational overhead and offers a template for optimizing complex multicomponent processes to module products across various technical sectors.