Machine strength grading of sawn timber is based on the relationship between indicating properties (IPs) and so-called grade determining properties (GDPs). The former are calculated using board properties measured non-destructively whereas the latter are determined by destructive tests. For T-classes, applied GDPs are tensile strength, density and modulus of elasticity in tension. Efficient utilisation of sawn timber thus require IPs predicting the GDPs with high accuracy. This study aims at deriving an IP for tensile strength in the board direction using multiple linear regression based on board properties and at applying this new IP in a grading method and calculating corresponding yield in T-classes. The nondestructive and destructive testing were performed for more than 900 boards of Norway spruce originating from Finland, Norway and Sweden. Board properties necessary for calculating this new IP were density, axial resonance frequency and in-plane fibre directions at the board’s longitudinal surfaces, and these properties were determined using X-ray scanning, dynamic excitation and optical scanning, respectively. A high coefficient of determination between IP and tensile strength was obtained, implying that this new IP predicts tensile strength better than IPs used in industry today. Furthermore, since only a part of the board is destructively tested, settings can be determined using either IPs calculated for the tested part of the boards or IPs calculated for the entire length of the boards. Employing the latter resulted in higher yields but lower coefficients of determination. Several grading machines on the market are based on this approach.