When should principals acquire verifiable information?

We analyze a principal-agent model in which a principal has two possibilities to improve his knowledge about the quality of an investment project. First, he has access to an informationtechnology that provides a \textit{verifiable}, unbiased signal. Second, he can hire an agent who detects bad projects with some probability depending on his unobservable effort, and who reports his findings opportunistically. We analyze whether the principal should check the signal before or after he offers a contract. The first policy has the advantage that the agent''s effort can be adjusted to the signal, whereas the second policy allows areduction in the agent''s rent. We show that checking the signal afterwards is always superior if the signal is sufficiently uninformative.