Accurate estimates of in place thickness of early age (3 to 28 days after casting) concrete pavements are needed, where a thickness accuracy of ${pm}6mm$ is desired. The impact-echo method is a standardized non-destructive technique that has been applied for this task. However, the ability of impact-echo to achieve this precision goal is affected by Vp (measured) and ${eta}$ (assumed) values that are applied in the computation. A deeper understanding of the effects of these parameters on the accuracy of impact-echo should allow the technique to be improved to meet the desired accuracy goal. In this paper, the results of experimental tests carried out on a range of concrete slabs are reported. Impact-echo thickness estimation errors caused by material property gradients and sensor type are identified. Correction factors to the standard analysis method are proposed to correct the identified errors and to increase the accuracy of the standard method. Results show that improved accuracy can be obtained in the field by applying these recommendations with the standard impact-echo method.