The natural growth of wood leads to significant variation in the mechanical properties of structural lumber among an ensemble of boards and within a single board. As such, the safe and efficient use of wood in structural applications requires a process to assess the probabilistic nature of wood properties. Current processes neglect within-member variation and utilize uniform design properties for a given species and grade. This paper presents a framework for a high-fidelity treatment of within-member strength variation to enable higher precision uncertainty characterization of strength. The model incorporates clearwood strength variation and the strength reduction of knots. The model is adapted to incorporate two distinct functions for knot strength reduction, providing information for the comparison of the two calculation methods and demonstrating the model’s adaptability as knot interaction continues to be studied. Simulated boards are graded by knot frequency and locations, and comparison is made between modelling results by grade and published experimental design values. The modelling method is also compared to experimental data of beams in bending to demonstrate the modelling process. Experimental failure of beams in bending is compared to the modelled failure location. The developed framework can be employed in computational analyses, such as reliability studies, to support improved material efficiency of structural lumber and large engineered wood products, like cross laminated timber. The tensile strength models were calibrated to Eastern hemlock but could be applied to other softwood species of interest. Incorporating a high-fidelity description of material variation in the fabrication of mass timber products could support material efficiency and promote the use of traditionally low-value wood species in high-value structural applications.