Characterizing the temperature-dependent development rate requires empirical data acquired by rearing individuals at different temperatures. Many mathematical models can be fitted to empirical data, making model comparison a mandatory step, yet model selection practices widely vary.We present guidelines for model selection using statistical criteria and the assessment of biological relevance of fits, exemplified throughout a Lepidoptera pest dataset. We also used in silico experiments to explore how experimental design and species attributes impact estimation accuracy of biological traits. Our results suggested that the uncertainty in model predictions was mostly determined by the rearing effort and the variance in development times of individuals. We found that a higher number of tested temperatures instead of a higher sample size per temperature may lead to more accurate estimations of model parameters. Our simulations suggested that an inappropriate model choice can lead to biased estimated values of biological traits (defined as attributes of temperature dependent development rate, i.e., optimal temperature for development and critical thresholds), highlighting the need for standardized model selection methods. Therefore, our results have direct implications for future studies on the temperature-dependent development rate of insects.