AI-Driven Dynamic Pricing and Promotion Engine for Retail and Hospitality

Abstract

In the fast-paced retail and hospitality sectors, static pricing strategies increasingly fail to capture rapid shifts in market dynamics and consumer behavior. Implementing automated, data-driven pricing models within software-as-a-service (SaaS) platforms has therefore become essential for maintaining market competitiveness. However, existing dynamic pricing solutions often struggle to balance pure predictive accuracy with complex, real-world operational constraints and highly volatile localized external factors. This dissertation addresses this gap by developing and evaluating an AI-driven dynamic pricing and promotion recommendation engine integrated as an external API within the StoresAce SaaS platform. The system implements a two-stage pipeline: a demand modeling component benchmarking gradient-boosted trees against deep learning architectures, followed by an optimization module that translates forecasts into actionable recommendations under strict business constraints, market competition, weather sensitivity, and localized events. The engine's software architecture emphasizes modularity, continuous integration (CI/CD), and latency-oriented performance validation. Comprehensive evaluation via historical replay simulations, benchmarking against static baselines, and domain-expert reviews demonstrates that the engine successfully enforces operational constraints while simultaneously optimizing revenue and volume objectives. Furthermore, compliance auditing confirms that the system meets strict enterprise latency and feasibility standards under production-like conditions. Ultimately, these findings reveal that viable real-world dynamic pricing requires the explicit synthesis of operational decision criteria rather than relying on predictive accuracy alone, providing a scalable blueprint for production-ready decision support systems.