Forty beef carcasses were classified for conformation and fatness. Besides, carcass weight, fat thickness (FT), carcass dimension, marbling by computer image analysis and ultrasound readings was recorded to complement grading. For predicting intramuscular fat (IMF) content, FT, number of intramuscular flecks and conformation increased R-2-Value from 0.19 to 0.64 compared to conformation alone. For visual marbling, ultrasound readings and thoracic depth (TD) increased the R-2-Value from 0.24 to 0.57 compared to fatness score (FS). The best variables for predicting weight of fabricated subprimals were carcass weight or compactness which is a function of carcass weight (R 2 between 0.94 and 0.63). Fatness score was poorer than FT for predicting yield of subprimals cuts from round (R-2 = 0.16 vs. 0.50) and ultrasound readings for less valuable subprimals (R 2 = 0.31 vs. 0.39). These results showed that other variables could be used in combination with carcass fatness or conformation to achieve a more accurate estimation of fat and carcass yield. (c) 2009 Elsevier Ltd. All rights reserved.