Airlines & Booking

Revenue management (RM) is the science of selling the right seat to the right customer at the right price. How it works: 1) Booking classes: Each flight has 180 seats but 10+ 'virtual' fare classes (Y, B, M, H, Q, V, etc.) each with different prices and rules. A seat in class V costs ₹2,500, same seat in Y costs ₹12,000. 2) Demand forecasting: ML models predict demand for each flight based on historical data, day of week, seasonality, events, competitor pricing. 3) Dynamic availability: RM system opens/closes booking classes based on forecast vs. actual bookings. If bookings are ahead of forecast, close cheap classes early. If behind, keep them open longer. 4) Price discrimination: Business travellers book last-minute (willing to pay ₹12K). Leisure travellers book early (price-sensitive, pay ₹3K). RM captures both by offering cheap fares early and expensive fares later. 5) Overbooking: Airlines overbook by 5-15% because 5-10% typically don't show up. ML predicts no-show rate. Example: IndiGo DEL-BOM might sell 185 seats for 180-seat aircraft, expecting 8 no-shows.

GDS (Global Distribution System) is a network that connects airlines' inventory with travel sellers worldwide. The big three: Amadeus (Spain), Sabre (USA), Travelport (UK). Ecosystem flow: 1) Airlines load inventory and fares into GDS (and directly via NDC). 2) OTAs (MakeMyTrip, Booking.com) connect to GDS to search availability and book. 3) Travel agents use GDS terminals (Cryptic commands like AN15MARDELBOM) to search and book. 4) When a booking is made, GDS sends a request to the airline's reservation system, creates a PNR, and confirms. 5) Airline pays GDS a booking fee (typically $2-5 per segment). Revenue model: Airlines pay GDS per booking. GDS pays travel agents incentives to book through them. OTAs earn from markup on fares or service fees. NDC disruption: IATA's NDC enables airlines to sell directly to OTAs via modern REST/XML APIs — bypassing GDS. Airlines get richer content (images, bundles), more control, and save GDS fees. But GDS still handles 40%+ of bookings.

Challenges: Searching across 500+ airlines, millions of fare combinations, connecting flights, and real-time availability. Architecture: 1) Pre-computation: For popular routes (DEL-BOM, BLR-DEL), pre-compute and cache results every few minutes. 80% of searches hit cached results. Redis or in-memory store. 2) Route graph: Build a graph of all airport connections. Use graph algorithms to find valid routings (direct, 1-stop, 2-stop) — limit to reasonable connection times and total duration. 3) Parallel search: Query multiple sources in parallel — Amadeus GDS, Sabre GDS, IndiGo NDC, SpiceJet API. Merge results. 4) Fare combination engine: For connecting flights, combine fare components using IATA fare construction rules. Use ITA-style algorithms (QPX) for optimal fare search. 5) Caching layers: L1 cache (in-memory, 30-second TTL for availability), L2 cache (Redis, 5-minute TTL for search results), L3 (pre-computed for popular routes). 6) Progressive loading: Show partial results immediately (cached), then append real-time results. User sees results in <1 second for popular routes. MakeMyTrip approach: Pre-cached popular routes + parallel GDS + NDC queries + aggressive result caching.

NDC (New Distribution Capability) is an IATA standard that enables airlines to distribute rich content directly to travel sellers via modern APIs. Before NDC (legacy): Airlines sold via GDS using EDIFACT (1980s teleprinter format). Limitations: plain text, no images, limited product differentiation. An economy seat on IndiGo and Air India looked the same in GDS. With NDC: Airlines expose APIs (REST/XML) with rich content — branded fares (Economy Light, Economy Flex), product images, bundled offers (seat + meal + baggage), dynamic pricing. Travel sellers (OTAs, agents) connect directly to airlines. Key changes: 1) Rich content: Airline can show branded fares, images, ancillary bundles. Better shopping experience. 2) Dynamic pricing: Airline can offer personalised prices based on customer profile (vs. fixed filed fares). 3) Lower distribution cost: Airlines save GDS booking fees ($2-5 per segment). 4) Direct relationship: Airline has more control over the customer experience. Challenges: Every airline's NDC API is different (vs. standardised GDS). OTAs must integrate individually. Not all airlines have NDC ready. Hybrid distribution (GDS + NDC) is the current reality.

Disruption management is critical — a single cancelled flight affects 150-200 passengers who need rebooking. Process: 1) Detection: Operations centre detects disruption (weather, technical, crew issue). Flight status updated to delayed/cancelled in system. 2) Auto re-accommodation: System automatically rebooks passengers on next available flight on same airline. Priority order: connecting passengers (miss connection), premium cabin, loyalty tier (Platinum → Gold → Silver), then by booking class. 3) Interline rebooking: If own flights are full, rebook on partner airlines (interline agreements). IndiGo can rebook on Air India for certain routes. 4) Communication: Instant SMS/email/push to affected passengers with new itinerary. Self-service portal for alternatives. 5) Compensation: Per DGCA (India) regulations — meal vouchers for long delays, full refund option for cancellations, monetary compensation for denied boarding. Tech challenges: Real-time optimisation across entire network. One cancellation can cascade — passengers on connecting flights, crew scheduling, aircraft rotation all affected. Airlines use Operations Research algorithms (min-cost flow, constraint optimisation) to minimise total disruption cost.