Many techniques for improving latency in the Internet trade trade off some amount of extra bandwidth consumption for reduced latency. But when is the latency improvement achieved worth the cost of the added overhead? We develop a benchmark for answering this question.