Theoretical literature predicts that raiders should hold part of the shares – a so-called toehold – before a takeover operation. Whereas theoretical results predict that the toehold should be maximal, empirical results reveals heterogeneity in the observed levels of toeholds. Our study explains this gap. We develop a tender o?er model with atomistic target shareholders and asymmetric information. We find that it is optimal to acquire a toehold prior to the announcement of the takeover and we determine the optimal level of the toehold. Some particular environments may reduce the optimal level of toeholds: strong dilution mechanisms, soft financial constraints for the bidder, low financial frictions, a high legal protection for investors and not all-cash payment methods. We provide empirical evidence that support our predictions.