Trawl performance was studied based on net spread and variability in the contact of the footrope with the seabed and their relationship with vessel operations, catch weight, and environmental conditions. Artificial neural networks (ANNs) and generalized additive models (GAMs) were used to model the response of each variable. For the variables net spread and variability in footrope contact (VFC), the ANN models were more accurate than the GAMs, with greater generalization capacity in the validation phase. In the best ANN model for net spread, all variables were significant. The relationship between tow direction and wind direction (tactic) was most important in the sensitivity analysis. Net spread increased with increasing towing speed and windspeed, and decreased with increasing wave height. In the ANN model for VFC, there were differences among vessels, and both scope ratio and catch size were not significant. VFC increased with increasing depth and decreased with increasing towing speed and windspeed. The results demonstrate that both operational variables and environmental conditions affect trawl performance, and suggest that survey protocols designed based on this information might help to improve the precision of biomass estimates.