Airline Revenue Management: Impacts of Fare Simplification on RM SystemsChanging Fare Structures WorldwideFare Simplification Reduces SegmentationBOS-SEA Fare Structure American Airlines, October 1, 2001BOS-SEA Simplified Fare Structure Alaska Airlines, May 1, 2004In Retrospect, the Airline RM Problem was Relatively Simple“Spiral-Down” in Traditional RM SystemsRevenue Impacts of Fare Simplification with Traditional RM ModelsTraditional RM Models “Spiral Down” without Product DifferentiationAirline RM Systems Are Struggling Under Fare SimplificationUS Network Carrier Yields and Load Factors 1995-2006Can Existing RM Systems be Saved? Current RM Challenge: Changing and Different Fare StructuresNew Developments in RM ModelingQ-Forecasting of Price-Oriented DemandHybrid Forecasting For Simplified Fare StructuresChange in Fare Class Mix – EMSRb+HFFare Adjustment MethodsLeg-Based Fare Adjustment PrincipleEMSRb Controls with Fare AdjustmentNetwork RM with Hybrid Forecasting and Fare AdjustmentO-D Control Fare AdjustmentHybrid Forecasting and Fare AdjustmentExisting RM Systems Are Inadequate for Changing Fare StructuresSell-up Rates Must Be Estimated from Historical ObservationsWillingness to Pay Relative to Lowest Fare Changes over the Booking ProcessBringing OR Back to Airline RMCan Existing RM Systems Be Saved?M I T I n t e r nM I T I n t e r n a t i o n a t i o n a l a l C e n t e r f o C e n t e r f o r A r A i r Ti r T r a n s p o r a n s p o r t a t i o r t a t i o n nICAT ICAT MIT MIT Airline Revenue Management: Impacts of Fare Simplification on RM Systems1.201 Transportation Systems Analysis: Demand & EconomicsDr. Peter P. BelobabaMIT MIT ICAT ICAT 2Changing Fare Structures Worldwide Major shifts in airline pricing strategies since 2000 Growth of low-fare airlines with relatively unrestricted fares Matching by legacy carriers to protect market share and stimulate demand Increased consumer use of internet search engines to find lowest available fare options Greater consumer resistance to complex fare structures and huge differentials between highest and lowest fares offered Recent moves to “simplified” fares overlook the fact that pricing segmentation contributes to revenues: Fare simplification removes restrictions, resulting in reduced segmentation of demandMIT MIT ICAT ICAT 3Fare Simplification Reduces SegmentationFare Code Dollar Price Advance PurchaseRound Trip? Sat. Night Min. Stay Percent Non-Refundable Y $500 -- -- -- -- B $375 7 day -- -- 50 % M $250 14 day -- -- 100 % Q $190 21 day -- -- 100 % • With fewer restrictions on lower fares, some Y (business) passengers are able to buy B, M and Q. • Keeping B, M, Q classes open results in “spiral down” of high fare passengers and total revenues.MIT MIT ICAT ICAT 4BOS-SEA Fare Structure American Airlines, October 1, 2001Roundtrip Fare ($) Cls Advance PurchaseMinimum Stay Change Fee? Comment 458 N 21 days Sat. Night Yes Tue/Wed/Sat 707 M 21 days Sat. Night Yes Tue/Wed 760 M 21 days Sat. Night Yes Thu-Mon 927 H 14 days Sat. Night Yes Tue/Wed 1001 H 14 days Sat. Night Yes Thu-Mon 2083 B 3 days none No 2 X OW Fare 2262 Y none none No 2 X OW Fare 2783 F none none No First ClassMIT MIT ICAT ICAT 5BOS-SEA Simplified Fare Structure Alaska Airlines, May 1, 2004Roundtrip Fare ($) Cls Advance PurchaseMinimum Stay Change Fee? Comment 374 V 21 days 1 day Yes Non-refundable 456 L 14 days 1 day Yes Non-refundable 559 Q 14 days 1 day Yes Non-refundable 683 H 7 days 1 day Yes Non-refundable 827 B 3 days none No 2 X OW Fare 929 Y none none No 2 X OW Fare 1135 F none none No First ClassMIT MIT ICAT ICAT 6In Retrospect, the Airline RM Problem was Relatively Simple Fundamental assumptions of traditional RM models: Multiple fare levels offered on same flight, same itinerary Each has different restrictions and characteristics Demand for each fare class is independent and identifiable Passengers will only buy their preferred fare product Implications for forecasting: Future demand can be predicted based on historical bookings in each fare class Time series statistical methods used by most RM systems Implications for optimization: Given independent demand forecasts and remaining capacity, optimize booking limits for each fare class by flight or networkMIT MIT ICAT ICAT 7“Spiral-Down” in Traditional RM Systems Simplified fare structures characterized by One-way fares with little or no product differentiation, priced at different fare levels  Without segmentation, passengers buy the lowest available fare Fare class forecasts based on historical bookings will under-estimate demand for higher fare levels Previous “buy-down” is recorded as lower fare demand EMSRb under-protects based on under-forecasts of high-fare demands Allowing more buy-down to occur, and the cycle continuesMIT MIT ICAT ICAT 830,00035,00040,00045,00050,00055,00060,00065,000FullyRestrictedRemove AP Remove SatNight Min StayRemove AllRestr, KeepAPRemove AllRestr and APRevenue Impacts of Fare Simplification with Traditional RM Models-0.5%-29.6%-45%-16.8%MIT MIT ICAT ICAT 9Traditional RM Models “Spiral Down” without Product Differentiation0102030405060708090100Fully Restricted Remove AP Remove SatNight Min StayRemove AllRestr, Keep APRemove All Restrand APFC 6FC 5FC 4FC 3FC 2FC 181.687.879.882.788.1MIT MIT ICAT ICAT 10Airline RM Systems Are Struggling Under Fare Simplification Primary responsibility for revenue maximization has shifted from pricing to RM  Simplified fares still offer just as many price levels, but segmentation restrictions have been removed Existing RM systems still employed to control number of seats sold at each fare level Current RM system limitations are negatively affecting airline revenues Existing systems, left unadjusted, generate higher load factors but lower yields Many legacy carriers are using “rule-based” RM practices, for lack of a systematic approach to revenue maximizationMIT MIT ICAT ICAT 11US Network Carrier Yields and Load Factors 1995-200611.011.512.012.513.013.51995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006US Cents/RPM65%67%69%71%73%75%77%79%81%83%LOAD FACTORPAX YIELDLOAD FACTORMIT MIT ICAT