What is a natural kind? This old yet lasting philosophical question has recently received new competing answers (e.g., Chakravartty, 2007; Magnus, 2014; Khalidi, 2013; Slater, 2015; Ereshefsky & Reydon, 2015). We show that the main ingredients of an encompassing and coherent account of natural kinds are actually on the table, but in need of the right articulation. It is by adopting a non-reductionist, naturalistic and non-conceptualist approach that, in this paper, we elaborate a new synthesis of all these ingredients. Our resulting proposition is a multiple-compartment theory of natural kinds that defines them in purely ontological terms, clearly distinguishes and relates ontological and epistemological issues —more precisely, two grains of ontological descriptions and two grains of explanatory success of natural kinds—, and which sheds light on why natural kinds play an epistemic role both within science and in everyday life.