Plant and pollinator communities are paradigmatic examples of mutualistic ecosystems, characterized by the fact that both sets of agents interact mutually benefiting each other by the increase of reproductive fitness of plants and the access to nutrients of pollinators. Such ecosystems exhibit a marked temporal signature, called phenology, showing significant year-to-year variability as well as seasonal patterns of activity and flowering. Despite recent efforts in quantifying the temporal variation of the structural features of the network of interactions, the dynamics of the community is still typically investigated using the aggregated network, thus, entirely overlooking phenological aspects. In our work, we address the question of the stability of plant-pollinator mutualistic communities when incorporating the information about phenology. We modify a recently proposed model that integrates, in a multilayer framework, both inter-guild mutualism and intra-guild competition for mutualistic resources. The actual temporal overlaps between species are accounted for in an effective manner through the interaction coefficients. As a result, the worst-case scenario assumption that all species sharing a mutualistic partner will be competing, implied when using an aggregated network, is now tempered, and the amount of mutualistic benefit obtained is modulated according to the overlap. We parametrize this model using phenological data obtained for a real ecosystem. This dataset contains both the empirical network of interactions and the initial times and periods of activity of all the species. We observe that, when the competition is limited to those species that coexist in time, the persistence of species in the stationary regime increases significantly (see Fig. 1). In order to decide whether this effect trivially arises from an average competition that is smaller than in the aggregated network, we build a null model by reshuffling the overlaps while preserving all the observed mutualistic interactions. In this case, we find that, while the persistence and biodiversity are larger than in the aggregated system, the null model does not explain the persistence found when using the empirical phenology. In particular, species with shorter periods are more prone to extinction when the initial times of activity are shifted. On the whole, our results show that phenology is a key factor that leads to considerations on the community stability that, otherwise, cannot be predicted by the network’s topology alone.