Unpredicted sheet forming failures of dual-phase (DP) steels can occur in regions of high curvature and with little apparent necking. Such failures are often referred to as “shear fractures”. In order to reproduce such fractures in a laboratory setting, and to understand their origin and the inability to predict them, a novel draw-bend formability (DBF) test was devised using dual displacement rate control. DP steels from several suppliers, with tensile strengths ranging from 590 to 980 MPa, were tested over a range of rates and bend ratios (R/t) along with a TRIP (transformation induced plasticity) steel for comparison. The new test reliably reproduced three kinds of failures identified as types 1, 2, and 3, corresponding to tensile failure, transitional failure, and shear fracture, respectively. The type of failure depends on R/t and strain rate, and presumably on the initial specimen width, which was constant in this study. Two critical factors influencing the lack of accurate failure prediction were identified. The dominant one is deformation-induced heating, which is particularly significant for advanced high strength steels because of their high energy product. Temperature rises of up to 100 deg. C were observed. This factor reduces formability at higher strain rates, and promotes a transition from types 1 to 3. The second factor is related to microstructural features. It was significant in only one material in one test direction (of 11 tested) and only for this case was the local fracture strain different from that in a tensile failure. Alternate measures for assessing draw-bend formability were introduced and compared. They can be used to rank the formability of competing materials and to detect processing problems that lead to unsuitable microstructures.