Despite the many years of widespread use of the BMW Intake Valve Deposit (IVD) vehicle test, relatively little has been published quantifying the variation in the test procedure. This paper presents an analysis of the variability in the BMW test. Though results from 8045 km (8K; 5,000 mile) tests rather than 16090 km (16K; 10,000 mile) are highlighted due to the size of the available database and relative sensitivity of the data, analysis suggests that variation at 8K is representative of 16K variation. A square root transformation of average deposit weight at 8K, though more cumbersome than the more common log transformation, is found to be the most appropriate way to eliminate the dependence of variation on the absolute level of deposits. Within-car variation is found to account for over half of the test-to-test variation, contradicting the notion that car-to-car differences are the dominant source of variability. In a comparison with published results, the CRC Ford IVD test is found to exhibit about half the variability of the BMW dynamometer in the range of most interest. Consistent with previously published IVD literature, unadditized fuel results are found to have lower variability, both between tests and valve-to-valve, than additized fuels and within-test valve-to-valve variation is found to exhibit a statistically significant pattern in additized fuels.