Early in their study of one-dimensional kinematics, my students build an algebraic model that describes the effects of a rolling ball's (perpendicular) collision with a wall. The goal is for the model to predict the ball's velocity when it returns to a fixed point approximately 50-100 cm from the wall as a function of its velocity as it passes this point initially. They are told to assume that the ball's velocity does not change while it rolls to or from the wall--that the velocity change all happens very quickly and only at the wall. In order to evaluate this assumption following the data collection, I have the students analyze one such collision using video analysis. The results uncover an excellent teachable moment about assumptions and their impact on models and error analysis.