This study was established to test whether the fit between the overall log demand distributions required by mills and the cumulative log output distributions could be improved by localizing the demand matrices controlling the bucking-to-order process on modern cut-to-length harvesters. Fifteen mature Norway spruce (Picea abies (L.) Karst.) stands were cut with a bucking simulator under the control of both the stand-specific and uncontrolled reference demand matrices. The test simulations involved three Norway spruce log products: sawlogs, veneer logs, and pulpwood logs. The stand-specific demand matrices for these products were generated in parallel using a genetic algorithm (GA) based search system initiated by the taper of each tree to be harvested and the overall demand and price matrices of each log product. The GA system was run with both the real stem data (i.e., stem profiles recorded by harvester) and the stem data compiled from preharvest forest inventory data. The reference matrices were the overall demand matrices adopted from two Finnish sawmilling companies. The test results showed that compared to the uncontrolled reference matrices the GA-controlled demand matrices produced a 22%–103% higher total fit between the overall log demand distributions and the cumulative log output distributions at the forest level.