This paper evaluates polymer fluid cleanup in general and more specifically why Cotton Valley fracture treatments recover less than one half of the injected polymer. The inefficient cleanup is expected to explain why post-fracture well testing often indicates a smaller fracture length and conductivity than anticipated. Numerical simulations were performed on a typical fracture treatment using a compositional reservoir model and considered a range of yield values for the rheology of the fluid remaining after leakoff. The yield value range was based on that provided by an independent laboratory. Simulation results compare favorably to actual field production and well tests. The yield value for the polymer residue was introduced into the simulator by newly derived flow relations based on the Herschel-Bulkley yield power-law model. The paper presents these relations and the method for inclusion into a compositional reservoir model. Simulations indicate that the fracture only cleans up to a length governed by the yield stress of the fluid and multiphase effects. Without a yield value, the complete length would eventually clean-up if nominal values of dimensionless conductivity are provided. This study represents the first attempt to investigate the effect of yield stress in fracture fluid cleanup. The results suggest that yield stress and relative permeability effects are the dominant mechanisms controlling the cleanup of the fracture and that the cleanup length is determined before gas breakthrough. A curve is presented which relates effective fracture length to polymer yield stress. This type of relation, developed for specific reservoir conditions, can be used to adjust the predicted fracture length and permit more accurate project evaluations without the need for detailed simulation of every fracture treatment.