Benchmarking is often used in establishment surveys to adjust sample weights to match the current distribution of the population of interest. In the National Compensation Survey (NCS), an establishment survey of employer compensation costs conducted by the Bureau of Labor Statistics, the sample reference period is several quarters prior to the reference period for the estimates. Hence, the weight of each sampled establishment is adjusted to match the distribution of current employment by industry from the Quarterly Census of Employment and Wages program. NCS data indicate that compensation costs also are correlated with establishment size. This research studies the effect of splitting each industry cell further by establishment size class for benchmarking purposes. Several size class partitions are being studied. First, compensation costs for the existing sample are derived using benchmarking by industry and size class, and then compared to the current estimates. Second, a multi-sample simulation is being used to assess the effect of benchmarking by industry and size class on the mean squared error of compensation estimates. Results and analysis are presented in the paper.