When new wireless technologies are deployed and subjected to real usage patterns, unforeseen performance problems inevitably seem to arise, to be fixed only in later generations. Why do these performance issues fail to appear in experimental settings before the technology is deployed? We believe that one of the major reasons behind the discrepancies found between experimental performance evaluations and real-world experience lies in the unrealistic workload patterns typically used in experiments. One of the significant contributions of this work is to rigorously demonstrate that common synthetic traffic models for wireless local-area networks induce drastically distorted performance metrics at every layer of the protocol stack. In order to show this, we present a testable definition of “sufficient realism” for traffic models, and develop the theoretical methodology necessary to interpret experimental results using this definition. Finally, we show by example that this distortion can completely invert the relative performance of protocols. The greater overall contribution of this paper, however, is the complete collection of ideas, techniques and analytical tools that will allow the development of more realistic synthetic traffic models in the future.