In this paper we present the use of two procedures for monitoring an electoral survey: the classical sampling approach commonly adopted and the Bayesian model for forecasting electoral results introduced by Bernardo (1984), together with the methodology for selecting the number of sites described in Brazil & Pego e Silva (1994); in these a sample of the same size is selected in the sites of behavior most similar to the entire region being researched, in this case, the municipality. In the one-year period preceding the municipal elections of 10/3/92, five surveys were carried out in the city of Guarapari-ES, including the "mouth-urn" survey, four of which were via classical methodology and three through methodology bayesian In two occasions, August and September of 92,
