Think of your online store or booking site as a hotel front desk clerk who sees all reservations coming in. This “clerk” watches how early people book, what they add to their cart, and how full the inventory is getting, then decides in real time which extras (upsells) to offer and at what price to maximize total revenue without scaring customers away.
Most companies treat upsells (add‑ons, bundles, upgrades) as static checkboxes or generic recommendations. They ignore rich booking data (timing, lead time, demand level, remaining capacity, customer type), so they either leave money on the table or push the wrong offers at the wrong time. This work shows how to use booking and demand data systematically to make better upsell decisions in revenue management.
If deployed commercially, the main moat would be proprietary booking histories and demand signals (first-party data) combined with embedded placement in core revenue-management workflows. The underlying optimization logic itself is replicable by capable competitors.
Classical-ML (Scikit/XGBoost)
Structured SQL
Medium (Integration logic)
Real-time decision-making at scale requires low-latency access to current booking status and demand forecasts; the main bottleneck is likely integration with transactional systems and ensuring optimization runs within tight response-time budgets on high-traffic sites.
Early Adopters
Focuses specifically on quantifying the incremental value of booking data for upsell decisions within revenue management, rather than just dynamic pricing of the core product. That narrower focus on upsell—timing, offer selection, and capacity/demand interplay—goes beyond generic recommendation engines or standard RM pricing models.