Estimations of simple monetary policy rules are often very rigid. Standard practice requires that a decision is made as to which indicators the central bank is assumed to respond to, ignoring the data-rich environment in which policy-makers typically form their decisions. However, the choice of the feedback variables in the estimations of simple rules bears non-trivial implications for the prescriptions borne from these rules. This paper addresses this issue for the euro area using a new comprehensive real-time database for the euro area and examines the ECB’s past interest-rate setting behaviour in two complementary ways that are designed to deal with both model and data uncertainty. In a first step we follow the “thick-modelling” approach suggested by Granger and Jeon (2004) and estimate a series of 3,330 policy rules. In a second step we employ a factor-model approach similar to Bernanke and Boivin (2003) for the US Fed, but with structurally interpretable factors à la Belviso and Milani (2006). Taken together, we find a strong justification for the need of adopting robust approaches to describe the historical evolution of euro area monetary policy. We also find that the ECB is neither purely backward nor forward-looking, but reacts to a synthesis of the available information on the current and future state of the economy.