The study of variation transmission in multistage machining processes is a strategic task. It enables us to understand how variation is added and transmitted across the process stages and, therefore, to identify opportunities to reduce variation in key characteristics at the final stage. In this paper, a data-driven technique, based on a first-order autoregressive model, is applied to a multistage machining process of an aero-engine component. Data on 15 key characteristics of 42-tracked components are taken at each of eight sequential process stages. The statistical analysis of data relative to critical key characteristics permits determining the quantity of variation that is added at each stage and the quantity that is transmitted from upstream. This differentiation between added and transmitted variation allows the discovery of which stages contribute most to variation of key characteristics at the final stage. The statistical and technological interpretation of results enables identifying three typical technological scenarios that affect multistage machining process. The knowledge of the variance transmission modalities related to each scenario is a winning factor in achieving quality improvement and cost reduction. It has a direct influence on the effectiveness of variation reduction efforts and may provide useful information to define manufacturing cycles or to select machining tolerances. The application proposed has been developed in the AVIO industry, an international aerospace company at the leading edge of propulsion technology. Copyright © 2011 John Wiley and Sons Ltd.