In conservation biology and population dynamic studies, accounting for the effect of imperfect detection of animals in the wild is crucial for correct inference of demographic parameters. However, articles published in South American herpetological journals still report estimates without considering how detection can influence estimates of vital rates, assuming that detection probability is perfect (i.e., 100%) and constant. Using data from a population of the torrent frog, Hylodes asper, we calculate return rates and compare them with survival probability estimates adjusted for detection probability to highlight the discrepancies between the two metrics. Then, using power analysis, we also explore how survival is underestimated, considering different scenarios and sampling efforts, given low detectability. Finally, we provide information on the optimal number of surveys to achieve a reasonable precision, assuming a fixed number of individuals initially captured for a series of parameter values. Ignoring potential bias of uncorrected estimates may lead to weak inference and erroneous decisions for management and conservation. We recommend that researchers consider detection probability in their studies to improve the accuracy of population estimates.