Torques from a mutually inclined perturber can change a transiting planet's impact parameter, resulting in variations in the transit shape and duration. Detection of and upper limits on changes in impact parameter yield valuable constraints on a planetary system's three-dimensional architecture. Constraints for warm Jupiters are particularly interesting because they allow us to test origins theories that invoke a mutually inclined perturber. Because of warm Jupiters' high signal-to-noise ratio transits, it is feasible to detect changes in impact parameter. However, here we show that allowing the impact parameter to vary uniformly and independently from transit to transit leads to incorrect inferences about the change, propagating to incorrect inferences about the perturber. We demonstrate that an appropriate prior on the change in impact parameter mitigates this problem. We apply our approach to eight systems from the literature and find evidence for changes in impact parameter for warm Jupiter Kepler-46b. We conclude with our recommendations for light-curve fitting, including when to fit impact parameters versus transit durations.