This paper evaluates various theoretical failure criteria pertaining to workability in cold forging reported in the published literature for their reliability and sensitivity in predicting the occurrence of ductile fracture in metalworking. Finite element (FE) simulation of the published upsetting experiments on cylindrical test specimens was performed to determine the threshold values attained by various criteria at the fractured locations, for a wide variety of materials. A comparison of the experimental threshold values of different criteria, with those obtained through FEA of complex metalworking processes at fracture was also made. A statistical analysis of the results revealed that none of the criteria are truly friction and geometry-independent for universal application. Nevertheless, within a family of processes such as upsetting, the criteria depending on cumulative specific plastic energy adjusted suitably with the maximum tensile stress, are the most reliable ones in the estimation of workability limits. © 2003 Elsevier Science B.V. All rights reserved.