Recent OECD research has utilised harmonised cross-country firm level data to explore the contribution of public policies to cross-country differences in productivity, innovation and resource allocation. This paper describes the steps taken to and the trade-offs involved in constructing firm-level total factor productivity (TFP) measures using ORBIS, a cross-country longitudinal firm-level database available from Bureau van Dijk, an electronic publishing firm. First, it shows that not all productivity measures can be calculated using readily available variables for all countries, and presents possible solutions to this problem by using imputations for certain variables. Second, it assesses the accuracy of these imputations on a set of countries where the available data in ORBIS provides a good coverage, for a wide range of TFP measures. Indeed, an extensive comparison of the actual and the imputed values of TFP for those countries suggests that TFP measures using imputations provide a reasonable approximation for the "true" values. Furthermore, to improve representativeness, resampling weights are constructed - which help correcting for the underrepresentation of small firms - while for the sake of international comparability, industry-level PPP conversions are also applied. Finally, as a plausibility check and to illustrate the potential of the database, the paper explores the country-composition of the globally most productive firms, the forces of convergence to the productivity frontier and the impact of regulation on productivity growth, in a sample of 18 OECD countries.