Empirical testing in corporate finance often proceeds from qualitative theory to hypothesis tests by way of verbal plausibility arguments or via static models that do not actually make predictions regarding the quantities measured in the data such as investment rates and leverage ratios. Examples are provided regarding how this methodology has resulted in faulty inference. It is proposed that the formulation of hypothesis tests should instead be informed by operational models that explicitly map the primitive theories to measured variables. Examples are provided for how a particular class of operational models, Dynamic Quantitative Models, can be used to: inform empirical tests; estimate parameters; extract information from observed response elasticities; and perform a broad range of quantitative policy experiments. Recent criticisms of such models are addressed, and future directions discussed.