Life cycle assessment (LCA) has been widely used in swine production. However, some uncertainties are associated with LCA results, due to the diversity of production systems and agricultural practices, raising questions if the predicted data represents what actually happens in the pig system. The objective of this study was to evaluate the use of predicted and/or observed data as inputs of LCA of piglets (15 to 30 kg) in different dietary crude protein (CP) scenarios. Two animal profiles were created: feed intake and weight gain data representing the Brazilian national average (yp-AVG); and representing experimental conditions (yp-SPC). Performance and excretion of pigs were measured (yobs) and predicted (yp) for each CP scenario (19.2, 18.2, 17.2 and 16.2%) by using InraPorc (R), and thus, LCA was calculated. Predicted (yp) and yobs data of performance, nutrient excretion and LCA were compared as paired way by using a Bayesian approach. The feed intake and body weight gain were the same (P>0.05) for yp-SPC and yobs, while yp-AVG and yobs did not differ only for body weight gain, being the feed intake obtained for yp-AVG 94 g/d lower than yobs. No differences (P>0.05) between yp-SPC and yobs were observed for nitrogen intake, phosphorus intake and excreted. Conversely, yp-AVG and yobs did not differ (P>0.05) only for phosphorus excretion; nitrogen intake and excretion were 7% lower and 29% higher than yobs, respectively. Considering the LCA, yp-SPC did not differ (P>0.05) from yobs for global warming potential, energy demand (CED) and land occupation, while yp-AVG did not differ from yobs only for CED. Using models to predict data from specific farm performance indicators, to estimate the LCA of pig farming, clearly appears more appropriate than using national average indicators. The farm-characterization through a specific animal profile could provide robust results about the environmental impact of pig farming.