We assessed the performance of a transfer function model for sea-level studies using salt-marsh foraminifera from two estuaries of northern Portugal. An independent data set of 12 samples and 13 sub-fossil samples from a core were used to evaluate if reconstructions and errors derived from current models are adequate. Initial transfer function models provided very strong results as indicated by cross-validation (component 2; r2 = 0.80–0.82; RMSEP ranged from 10.7 to 12.3 cm) and improved its performance by ca. 10% when sample size reached ca. 50. Results derived using an independent test data set indicate that cross-validation is a very effective approach and produces conservative errors when compared to observed errors. We additionally explored the possible effect of transforming the concentration data into percent in the error estimations by comparing the results obtained based on the use of both concentration and compositional data. Results indicate that this type of transformation does not affect the performance of the transfer function. Results derived from a reconstruction of sub-fossil samples from a core indicate that high-resolution sea-level reconstructions are possible, but show that depositional environments have to be selected carefully in order to minimize the impact of possible taphonomical loss.