Simulation study demonstrates asymptotic optimality for integrated pricing, advertising, and inventory policies in new ventures, indicating the necessity of cross-functional learning.
Key Points
To develop an optimal dynamic learning policy that jointly determines pricing, advertising, and inventory levels for firms lacking historical customer demand data.
Formulated a multi-period decision model over T periods to simultaneously optimize pricing, advertising spend, and inventory ordering while learning unknown demand parameters.
Derived sufficient mathematical conditions to achieve exponentially fast learning rates across interconnected operational and marketing functions.
Evaluated the proposed policy analytically and through numerical experiments against a clairvoyant benchmark possessing perfect model information.
Established an asymptotically optimal policy that yields a profit regret bound of order log T relative to the clairvoyant benchmark.
Demonstrated that joint learning across both advertising and demand models significantly outperforms decoupled or isolated functional optimization.
Observed that policies deviating from the specified learning conditions to avoid active experimentation suffer substantial profit degradation.