<p>We study dynamic pricing for perishable goods in q-commerce, where short shelf life, stochastic arrivals, and cross-item interactions make pricing difficult. Our objective is to develop a tractable policy that balances near-term revenue with the future value of inventory while accounting for perishability and basket-level demand. We formulate a finite-horizon stochastic dynamic program that integrates a basket-based MNL choice model with item-specific expiry and platform commissions. To address the intractable nature of the exact solution, we develop an Approximate Dynamic Programming (ADP) policy with a linear value function approximation (VFA). A two-stage estimation initializes interaction effects from co-purchase counts and item utilities from a likelihood estimator. The simulations show that the proposed policy outperforms fixed, linear decay, and myopic pricing policies in total profit, primarily by reducing spoilage and stockouts. However, a strong sell-down heuristic is competitive and can outperform ADP under misspecified VFA weights; we analyze these conditions and suggest suitable remedies. Unlike item-level models that ignore item-level interactions, our basket-MNL captures complementarity/substitution via co-purchase-derived interaction terms and embeds them into a dynamic pricing framework tailored to perishable inventory. The resulting policy is grounded in data, adaptive, and suitable for use in real-world quick‑commerce platforms.</p>