Andrei Kirilenko Massachusetts Institute of Technology We measure message processing time or latency inside an automated trading platform. We show that latency is a random variable that has a strong predictive power over both volatility and the volatility of volatility of a highly liquid asset over and above changes in message traffic. We argue that in automated markets, processing time contains valuable nontrade information about the price formation process. We recommend that automated trading platforms improve pre-trade price transparency by reporting characteristics of latency to market participants on an ongoing basis along with order book events, transaction prices, and trading volume.