Concentration-dependent cytotoxicity experiments are frequently used in toxicology. Although it has been reported that an adequate choice of concentrations improves the quality of the statistical inference substantially, a recent literature review of three major toxicological journals has shown that the corresponding methods are rarely used in toxicological practice. In this study the performance of different sets of concentrations, also called designs, are analyzed, while the overall goal is to promote the advantages of optimal design procedures and to present a user-friendly guideline for planning new cytotoxicity concentration-response experiments. We compare the frequently used log-equidistant design to a Bayesian design, which is constructed by methods of optimum design theory. Using both a dense data set of concentration-cytotoxicity data of valproic acid (VPA) and regular assay data of 104 substances, the performance of the different designs is analyzed in two scenarios, where detailed previous knowledge on VPA is available or not. The results show that it is critical to apply a specific design strategy to determine optimal concentrations for cytotoxicity testing. In particular, the Bayesian design technique with and without incorporating pre-existing knowledge of a specific test substance resulted in a more precise statistical inference than the other used designs. Finally, we present a guideline for upcoming experiments and an accessible user-friendly Shiny app (see http://shiny.statistik.tu-dortmund.de:8080/app/occe ).