We propose a new Vector Autoregression (VAR) identification strategy to study the impact of noise, in the early releases of output growth figures, which exploits the informational advantage of the econometrician. Economic agents, uncertain about the underlying state of the economy, respond to noisy early data releases. Econometricians, with the benefit of hindsight, have access to data revisions as well, which we use to identify noise shocks. A surprising report of output growth produces qualitatively similar but quantitatively smaller effects than a demand shock. We also illustrate how a noise shock cannot be identified unless ex-post information is used.