This paper aims to build and empirically evaluate a discrete choice model of merger remedies as a basis for policy analysis. The database consists of 229 merger cases accepted in Phase I or Phase II of the European merger process between 1990 and 2005. We focus on the following question: Which merging firms’ characteristics lead the European Commission to decide whether to require conditional acceptance? Although a lot of empirical studies have been carried out these last years, ours is distinguished by at least two original features. First, we explore determinant factors of the Commission’s decisions with a neural network model differentiating cases accepted with or without remedies (either structural or behavioral). Secondly, we implement three multinomial logit models. We find that variables related to high market power lead more frequently to a remedy outcome, no matter the phase. Innovative industries such as energy, transportation and communications positively affect the probability of a behavioral remedy. Lastly, former Competition Commissioner Mario Monti’s policy appears to be pro-remedy, i.e. seeking concessions from merging parties.