We estimated cost/benefit ratios for different quality control programs of radiation output measurements of medical linear accelerators. The cost/benefit ratios of quality control (QC) programs (a combination of output measurement time interval and measurement action levels) were defined as workload divided by achievable dose accuracy. Dose accuracy was assumed to be inversely proportional to the 99% confidence limit of shifts of total treatment doses and workload as inversely proportional to the output measurement time interval. Our previously reported method was used to estimate the distribution of shifts of total treatment doses due to changes in accelerator radiation output (Gy/MU). The confidence limits of dose shifts were estimated for different QC programs and for different levels of output measurement reproducibility. Output shifts used in the estimations had previously been observed for four linear accelerators over 5 years. We observed that the cost/benefit ratio increases remarkably when the output measurement time interval is less than 1 month. The ratio depends strongly on the action levels and reproducibility of the QC measurements. Improvement of these factors optimizes the cost/benefit ratio by a factor of several times. The most cost-effective output measurement time interval to achieve 99% confidence limits of ±2, ±2.5 or ±3% for dose shifts ranged from 0.25 month to as much as 6 months depending on the factors given above and the intended accuracy level. It is several times more cost effective to increase dose accuracy by lowering the action levels of the QC measurements and by attempting to improve their reproducibility than by simply shortening the time interval of the output measurements. Methods improving utilization and interpretation of the results of the QC measurements play a key role in further optimization of cost/benefit ratios in dosimetric QC.