Numerical simulation exercises to analyze the impacts of potential changes in non-tariff policies commonly use ad valorem equivalent tariff treatment even though estimated impacts using explicit model representation and ad valorem equivalent treatments will differ. The difficulty for modellers is that the detail and subtlety embodied in a wide array of policy interventions means that some simplification is appealing, but no meaningful general propositions exist in the theoretical literature as to the sign or size of the differences in predicted effects. All that can seemingly be done is to investigate the differences case by case, but even here the findings are sensitive both to the particular form of model used as well as the model parameterization employed. As a result, there is relatively little in the literature that provides guidance as to how serious the pitfalls may be, and how misleading ad valorem tariff equivalent treatment is. Here I draw on three examples of numerical modelling where explicit representation of policy interventions are used. The picture that emerges is one of large quantitative and even qualitative differences in predicted impacts. These examples suggest that where interventions differ from a tariff, ad valorem representation should be undertaken in numerical trade modelling only with substantial caveats.