Travel & Hospitality

GDS (Global Distribution System) is a computerized network connecting travel suppliers (airlines, hotels, car rentals) with travel agents and OTAs. Major GDSs: Amadeus, Sabre, Travelport. Airlines file their fares and inventory to GDS. Travel agents query GDS to search, book, and issue tickets. GDS charges booking fees to airlines and subscription fees to agents. Revenue model is under pressure from NDC which enables direct airline-to-OTA connections.

NDC (New Distribution Capability) is IATA's XML/JSON-based standard for airline retailing. Unlike GDS which shows basic fares, NDC enables rich content (images, ancillaries, branded fares), dynamic pricing, and personalization. Airlines adopt NDC to: bypass GDS fees, control their retail experience, sell ancillaries better, and offer personalized prices. OTAs integrate NDC to get better content. Challenge: Each airline has slightly different NDC implementation.

Hotels distribute inventory through multiple channels: direct (hotel website), OTAs (Booking.com, MakeMyTrip), GDS (for corporate/TMC), wholesalers/bed banks, and meta-search. Channel Manager connects PMS to all channels, maintains rate parity, and syncs availability in real-time. When booking comes from any channel, availability updates everywhere. Challenge: managing rate parity, channel mix optimization, and commission costs (OTA fees 15-25%).

Revenue management optimizes pricing to maximize revenue given perishable inventory (airline seats, hotel rooms). Core concepts: demand forecasting, dynamic pricing based on demand/competition, fare class inventory control. Airlines use booking curves, historical data, and ML to forecast. Hotels track pickup pace, competitor rates, and events. System recommends prices, but revenue manager often has override. Key metrics: RevPAR (hotels), RASM (airlines), load factor.

Flight search is compute-intensive: query multiple suppliers (GDS, NDC, direct), handle millions of fare combinations, respond in <3 seconds. Architecture: Cache static data (schedules, fare rules), use distributed search (parallel supplier queries), implement result caching (Redis), pre-compute popular routes, use smart truncation for long-haul multi-city. Aggregation layer dedupes and ranks results. Consider latency vs completeness trade-off.

When airline cancels flight: 1) PSS marks flight cancelled, inventory released. 2) Affected PNRs identified, passengers notified. 3) Rebooking options generated (next available flights). 4) Passengers can rebook online or via call center. 5) If no rebooking, full refund processed. 6) If OTA booking, OTA system syncs with airline for status. Challenges: downstream hotel/car cancellations, communicating with passengers, handling surge in call center. Airlines increasingly automate rebooking.