Monitoring the anchovy fishery is of great importance for assuring the sustainability of the Humboldt ecosystem. The on-board observers program constitutes a rich platform for data collection for monitoring, as it consists of the collection of data from vessels during their fishing trips, regarding catches, size distribution, discards, catch per unit of effort, among others. To get robust indicators from these data, it is necessary to use a solid statistical procedure for computing the number of fishing trips to sample. This work presents a method with those characteristics, for computing each indicator independently and at different time scales. We show that the optimum size varies depending on the objective (indicator) and the time scale. Based on these results, practical recommendations for fixing the sample size are given. We finally discuss the aplicability of this methodology for other pelagic resources.