Abstract Numerous interactions between plants and animals vary in their outcome between antagonism and mutualism. Interactions between plants and scatter‐hoarding animals provide a prime example of this phenomenon. Scatterhoarders consume large quantities of seeds (potentially reducing plant establishment), yet also disperse seeds and bury them in shallow caches (potentially improving recruitment). Despite intense work on mechanisms that cause these interactions to shift along an antagonism–mutualism continuum, it remains difficult to quantify their final outcomes. We demonstrate how readily available field data can be used to reach this goal, with interactions between rodents and two oaks species (sessile oak Quercus petraea, and red oak Q. rubra) as an empirical example. Our approach consists of quantifying the net outcome of the interaction through collecting data on different vital rates (e.g. probability that cached seeds survive to germination, probability of seedling recruitment with and without rodents; near and far from conspecific trees; with and without seed pilferage) and assembling them in a simple mathematical model. We found that during the period of the study, interactions between scatter‐hoarding rodents and both focal oaks were antagonistic. Even though caching increased the likelihood of seedling establishment, this effect was not strong enough to compensate for the costs of seed predation. Furthermore, there was no evidence that the short‐distance transportation that is usually provided by small mammals benefited early oak recruitment. Synthesis. Our empirical results demonstrated that certain common assumptions – that caching by rodents invariably benefits plant recruitment; that improved seedling establishment after seed burial is sufficient to make plant–scatterhoarder interactions mutualistic; that transportation away from maternal plants is highly beneficial—do not always hold and should be tested rather than taken for granted.