We compare the performance of organizational forms (M‐form and U‐form) in experimenting with uncertain projects. In our framework, organizational forms affect the information structure of an organization and thus the way to coordinate changes. Compared to the U‐form, the M‐form organization achieves better coordination in “attribute matching” but suffers from coordination in “attribute compatibility” and less gains in specialization. The distinctive advantage of the M‐form is its flexibility in choosing between small‐scale and full‐scale experimentation.