In a general agency model with a risk-averse principal, we compare signals on two dimensions–their efficiencies at providing incentives for the agent's effort and efficiencies at conveying inference of the outcome. In order to tackle the moral hazard problem, the standard first order approach...