This paper exploits a rich longitudinal survey on firms’ characteristics and business practices to examine the comparability and determinants of productivity estimates. As Productive Development Policies are aimed at improving firms’ productivity, validating the accuracy of productivity proxies is critical since the selection of beneficiaries and assessment of effectiveness are often based on potentially biased estimates. Our findings suggest that traditional proxies show significant biases, especially in services sectors and smaller firms. The most significant observable determinants of permanent positive transitions toward higher quintiles in the TFP distribution are the educational levels of owners and workers, and the presence of an innovation department. The permanency of unexpected TFP shocks varies depending on sector, size, age, and workforce education, being more permanent in smaller firms of primary sectors with a less educated workforce and dominantly transitory in larger, younger firms with a highly educated labor force.