We reconcile seemingly conflicting hypotheses and evidence that surfac
e in the principal-agent literature. Specifically, we examine the lite
rature that deals with the effect of costly monitoring on retail-organ
izational form. Our principal-agent model of the optimal relationship
between up and downstream firms allows the principal to garner two typ
es of imperfect signals of agent effort: sales data and behavior data.
The model yields predictions that we confront with the econometric ev
idence, which comes from both franchising and sales-force-compensation
literatures. We find that, when variation in the informativeness acid
in the cost of increasing the informativeness of both signals is cons
idered, the evidence is consistent with the theory.