An efficient and safe hyperthermia cancer treatment requires the accurate control of the heating performance of magnetic nanoparticles, which is directly related to their size. However, in any particle system the existence of some size polydispersity is experimentally unavoidable, which results in a different local heating output and consequently a different hyperthermia performance depending on the size of each particle. With the aim to shed some light on this significant issue, we have used a Monte Carlo technique to study the role of size polydispersity in heat dissipation at both the local (single particle) and global (macroscopic average) levels. We have systematically varied size polydispersity, temperature and interparticle dipolar interaction conditions, and evaluated local heating as a function of these parameters. Our results provide a simple guide on how to choose, for a given polydispersity degree, the more adequate average particle size so that the local variation in the released heat is kept within some limits that correspond to safety boundaries for the average-system hyperthermia performance. All together we believe that our results may help in the design of more effective magnetic hyperthermia applications.