Fisher-design experiments rely on random assignment to uncover regularities, but without a valid induction method, their findings cannot be generalized. In theory-driven experiments, models design the experiment. The process is analogous to engineers building structures from blueprints: a theory’s models act as blueprints from which an experiment’s replicas are built. Replicas are set in motion when an experimenter introduces the study’s initial conditions. For example, to test a coercion model, an experimenter builds a lab-based replica, then introduces strategically rational actors into the structure’s positions. Tested theory is applied to impose order on chaotic, empirical observations. Building an experimental replica is difficult, if not impossible, without a modeling procedure.