THE METHOD of validating proving ground test routines as a sample of customer operations is described. Comparisons of field survey and proving ground failure statistics are analyzed to establish sample reliability and severity ratios. Alternative means of accelerating the test program are considered, with examples indicating the effect on correlation.Use of proving ground routes as a link in establishing laboratory test to customer service correlation is treated. Procedures and instrumentation for developing load cycle information on the road, and programming random load cycles into laboratory test fixtures for simulated service life testing are discussed with some examples of applications to automotive components.In the performance test area, several performance parameters representative of customer requirements are developed. Techniques of improving comparative evaluations by eliminating the effect of extraneous variables such as the driver and the weather are presented. Accuracy, sample size and confidence level of performance test data are revealed, with indications of the significance which can be attached to differences in comparative test data.*THE METHOD of validating proving ground test routines as a sample of customer operations is described. Comparisons of field survey and proving ground failure statistics are analyzed to establish sample reliability and severity ratios. Alternative means of accelerating the test program are considered, with examples indicating the effect on correlation.Use of proving ground routes as a link in establishing laboratory test to customer service correlation is treated. Procedures and instrumentation for developing load cycle information on the road, and programming random load cycles into laboratory test fixtures for simulated service life testing are discussed with some examples of applications to automotive components.In the performance test area, several performance parameters representative of customer requirements are developed. Techniques of improving comparative evaluations by eliminating the effect of extraneous variables such as the driver and the weather are presented. Accuracy, sample size and confidence level of performance test data are revealed, with indications of the significance which can be attached to differences in comparative test data.*