We study optimal contracting for innovation projects involving initial exploration followed by exploitation upon success, integrating the experimentation literature with dynamic corporate finance. The shape of the exploitation payoff frontier — not just its level — determines the optimal contract for experimentation and its duration. When the exploitation stage involves moral hazard and limited liability, the principal's continuation profit is inverted U-shaped in the agent's continuation utility. This weakens the principal’s rent-extraction motive and may induce more efficient experimentation, even when the exploitation payoff frontier is always lower. Applications to corporate governance and innovation subsidy design yield sharp policy prescriptions.