This work is an example of how to adapt a classification method, in this case a classification tree, to the present standardized method for the development of settings for strength grading machines. Data from commercially available industrial strength grading equipment were used on a large sample (approximately 1440 pieces) of Norway spruce (Picea abies (L. Karsten)) in various sawn dimensions. The equipment is a multisensor scanning device combining planar X-ray and resonance frequency measurement. Destructive testing was done according to European standard EN408. The goal was to make the classification, based on machine data, as close as possible to the optimum grading, which was done according to standard. Two different approaches for classification by cost-sensitive decision trees were applied to the data and compared to classification accredited according to EN14081. Classification accuracy increased from 64% correctly classified to 73%, and a reduction from 33% False Negative to 23% was achieved. False Positive increased from 3% to 4%. The outcome was an increase in value for the producer by 0.9%–2.1% at 2007 average price level. The improvement came mainly from an in-yield increase in C30 by 10%.