Revenue Management Adriana Novaes January 2011
Revenue Management Presentation Scheme 1.Introduction 2.Techniques 3.Industry Applications
Introduction To RM Definition Origins Pricing Segmentation Customer Segmentation Product Segmentation
Revenue Management Origins Airline Industry Airline deregulation Act 1978 Low-cost and charter airline competition Yield Management Robert Crandall (American Airlines) Revenue Management Robert Cross (Delta Airlines)
Revenue Management Definition Application of information systems and pricing to allocate the right capacity to the right customer at the right place at the right time. Kimes 1998 RM has been credited with improving revenues 3-7% in the airline, hotels and car rental industries Queenan, Ferguson, Highie, Kappor 2007
Seven RM Core concepts Consider price first when balancing supply and demand. Replace cost-based pricing with market-based pricing. Sell to segmented micro-markets, not mass markets. Save your products for your most valuable customers. Make decisions based on knowledge, not supposition. Exploit each products' value cycle. Continually re-evaluate your revenue opportunities. Cross 1997
Common pricing practice Unitary Cost Margin Price Discounts
Pricing Segmentation Selling to Micro-markets ONE PRICE FOUR PRICES Price Price Demand Demand REVENUE GROWTH
Pricing segmentation A strategy aimed at differentiating customers in order to charge different prices to different customers based primarily on differences in willingness to pay. Two different non-exclusive strategies: Customer Segmentation : offering different customers exactly the same product at different prices. Product Segmentation: offering slightly different products targeted at different customer segments.
Pure Customer Segmentation Difficult to implement Unpopular with consumers Sometimes illegal Exceptions: student and senior citizen discounts Group membership discount Geographical (zone pricing) couponing
Product Segmentation Inferior product to be sold to price-sensitive customers Es. restricted early booking, branded gasoline versus generic, national brand label versus store label versus generic. Geographical Es. zone pricing, international pricing, neighbourhood pricing for services Time-based segmentation es. time of purchase(airline,hotels), time of use (fashion,usage-time), delivery time
Buying Behaviour Matrix Attribute Prices Sensitive Price Insensitive Characteristics Characteristics Product Features Limited Comprehensive Physical Ordinary Luxurious Service Adequate Abundant Warranty Minimal Extensive Perception Brand Preference Indifferent Loyal Status Unconcerned Conscious Reliability Inconsequential Critical Circumstances Urgency Casual need Acute need Alternatives Multiple Few Purchase Size Large Small
Common segment bases Time of purchase Time of reservation Day of the week Cancellation likelihood Senior and youth Options and premptability Channel Delivery time Spend amount Loyalty Size of business Frequency Business and individuals Package Group discounts Saturday night stay Trip length or length of stay
Stepwise RM Approach Segment the market Predict Customer Demand Optimize price Recalibrate Dynamically Segmentation based on buying behaviour, not just current and past classifications Forecast of demand and capacity at product/price level Mathematically determine capacity availability and price that maximizes expected profit Continually monitor performance and update market response Cross 2009
Introduction References 6 th Annual Revenue Management & Price Optimization Conference, Workshop Principles of Revenue Management and Price Optimization Robert G. Cross, Atlanta 2010 6 th Annual Revenue Management & Price Optimization Conference, Workshop Introduction to the science of Revenue Management an Price Analytics Mark Ferguson, Atlanta 2010 Revenue Management Robert Cross, Broadway Books NY 1997 Theory and practice of Revenue Management Talluri & Van Ryzin Springer 2004 Restaurant Revenue Management, Kimes etall Cornell Hotel and Administration Quarterly June 1998 p.32-39 A comparison of Unconstraining Methods to Improve Revenue Management Systems Queenan et all, Production and Operation management, Vol16 No6 pp. 729-746 November December 2007
Revenue Management Techniques Capacity-based RM Price-based RM Common elements: Customer behaviour and market-response models Economics of RM Estimation and forecasting
Capacity-based RM BOOKING LIMIT LOW VALUE CUSTOMERS Single Resource Capacity Control Network Capacity Control HIGH VALUE CUSTOMERS CAPACITY OVERBOOKING TIME
Single Resource Capacity Control n classes Assigning booking limit for each class Mutually exclusive segments (customers in each segment can only afford one class) Es. different classes of single flight leg of hotel rooms for a given date CLASS 1 CLASS 2 CLASS 3 12 12 10 10 8 8 $100 $75 $50
Capacity-based RM Algorithms Adaptive or regular Static or Dynamic Arrival assumption Static: lower value consumers comes first Dynamic: arbitrary arrival order Discrete or Continuous Demand and capacity profile Discrete: probability distribution increases in jumps Continuous: probability distribution increases continuously Updating Adaptive: updates booking policy parameters Regular: no up-dating Optimization Or Heuristics Resolution approach Optimization: maximizes or minimizes a revenue function Heuristics: solve a problem by the application of interaction rule cycles.
Types of control Booking Limit Protection Level Bid price amount of capacity that can be sold to any particular class j at a given point of time (b j ) specifies an amount of capacity to reserve (protect) for a particular class j or set of classes (y j ) Threshold prices are defined in a bib price table (based on the remaining capacity or time). A request is accepted just if the revenue exceed the threshold price (x), x = remaining capacity
Control classification Partitioned divides the available capacity into separate blocks (buckets) that can be sold only to the designed class. Nested C = capacity x=amount of capacity available j=classes index b=booking limit y=protection level =bid price b j = y j, j = 1,..., n the capacity available to different classes overlaps in a hierarchical manner with higher-ranked classes having access to all capacity reserved to lower ranked classes b j = C - y j, j = 2,...,n
CLASS 1 CLASS 2 CLASS 3 12 12 10 10 8 8 $100 $75 $50 b 1 = 30 y 1 = 12 b 2 = 18 y 2 = 22 b 3 = 8 y 3 = 30 (x) $100 $75 $50 12 22 30 C = capacity x=amount of capacity available j=classes index b=booking limit y=protection level =bid price x
Displacement cost Or opportunity cost is the expected loss in the future revenue from using the capacity now rather than reserving it to future use. Value function V(x) measures the optimal expected revenue as a function of remaining capacity.
Single resource capacity Models Optimization models Heuristics Adaptive methods Dynamic Programming (Brumelle & McGill 1993) Montecarlo method (Robinson 1995) EMSR-A EMSR-B + Buy factor (Belobaba 1987, 1992) Robbins-Monro Algorithm (Van Ryzin&McGill 2000)
Network Capacity Control Products are sold as bundles A lack of availability of any one resource in the bundle limits sales Es. ODIF (origin-destination itinerary fare class) combination problem, room capacity on consecutive days when the customer stays multiple nights 1 night stay 3 night stay 2 night stay FRA Airline hub GIG 03/04 04/04 05/04 06/04 CDG MXP FJK
Network capacity types of control Partitioned Booking Limit Virtual Nesting Airline and hospitality legacy Bid-price Current practice allocate a fixed amount of capacity on each resource for every product that is offered. provide bounds on optimal network revenue uses single-resource nested-allocation controls for each resource of the network. Indexing assigns products to virtual classes sets a threshold price for each resource in the network. a revenue request is compared with the sum of the bid prices of the resources required by the product
Network-based capacity Models Expected values/ Demand distribution for single itineraries Most used Heuristics: EMSR-B Demand sample generation Linear and non-linear programming Single Capacity Problem Decomposition Simulation with Stochastic Gradient
Overbooking Increasing capacity utilization in a reservation based system when there are significant cancellations(airline cancellations and no-shows: 50%) Smith et all 1992 Customer relation policies: denied-services compensation strategies, selection criteria and oversale auctions Simon 1993 Legal and regulatory issues Most common application : Airlines, Hotels, Car Rental, Restaurants New application: Sporting Venues, Manufacturing, Professional Services
Overbooking costs Underage cost Overage cost Cost Airline: lost fare (but which fare?) Hotel/Casino: lost room rate + incidental profits Total cost Overage cost Compensation: often free future ticket or stay Provision Cost: meals, drinks, gifts Reaccomodation: sometimes list price at competitor ) Goodwill: ( ways to reduce this? ) Underage cost Number of overbooking
Overbooking over time Reservations Overbooking limit Reservations with Overbooking Show demand C Reservations without Overbooking T Time
Overbooking Models Static models Overbooking Criteria Service level Ecomomic Beckmann 1958, Taylor 1962, Thompson 1961, Rothstein& Stone 1967, Bierman & Thomas 1975, Martinez & Sanchez 1970 Show demand Distribution approximation Binomial Deterministic Normal Customer Class Mix Group Cancellations Gram-Charlier (Taylor 1962)
Overbooking Models Dynamic Models (Chatwin 1999) Exact approach Heuristic approach Combined Capacity Control (Subramanian et all 1999) No-shows Cancellations Exact method Class-dependent refunds Substitutable Capacity (Karaesmen and Van Ryzin 2001) Network Overbooking (Karaesmen and Van Ryzin 2001)
Dynamic Pricing Elasticity of Price Price versus quantity-based RM Applications: Retail markdown pricing Manufacturing E-business Promotion Optimization Auctions
Price Response Function Demand Unit sensitive Slope of tangent line = d (p) p Price The slope, d'(p) measures the local rate of change of the price response function
Elasticity of price ε = percentage change in demand for a 1%change in price ε < 1 (inelastic): raising price will increase revenue (price insensitive) ε = 1: Revenue is independent of price ε > 1: Raising price will decrease revenue (price sensitive) Depends on time period of measurement Unit free level of measurement (industry elasticity may be low, Individual product elasticity is always higher)
Price RM vs Quantity RM Price-based RM Quantity based RM Classic Retail applications Classic Airlines, hotels, cruise ships and rent-a-car applications Ability to vary price/quantity: Firms commitments Cost of making price changes Flexibility in supplying products or services (re-order stock or reallocate inventory) (Galego&Van Ryzin 1997)
Elasticity of price Toilet tissue 0.6 Shampoo 0.84 Cake Mix 0.21 Cat food 0.49 Frozen entrèe 0.60 Gelatin 0.97 Soups 1.05 Automobiles 1.2 Chevrolets 4.0 Chevy Caprice 6.1 More specific means higher elasticity
Retailing Markdown pricing 5%-15% gross margins improvement by the usage of model-based pricing software (Friend &Walker 2001) Style-Goods Markdown pricing Consumer Package-goods Promotion Apparel, sporting goods, high-tech and perishable foods High initial price, mark-down low reservation items Markdown on peak periods: more sensitive customer behaviour Lazear 1986 Pashigan 1988 Warner &Birsky 1995 Soap, diapers, coffee, yogurt,... Customer awareness of past prices/promotions reference price stockpile behaviour
Retailing Markdown prospective Manufacture Retail Trade promotions: manufacture discounts to retailers (with or without retailer pass-thru) Increase sales or profit for its brand Kopalle et all1999, Silva- Russo et all 1999, Tellis & Zufryden 1995 Retailer /consumer promotions: retail discount to customers Overall sales of profits for a category of multiple brands and multiple manufacturers products. Incentive compatibility constraints Greenleaf 1995, Kopalle et all 1996
Discount Airline pricing Price change in a limit set of values: 1.99, 9.99, 19.99 Price rises over time depending on capacity and demand for a specific departure Fare 250 200 150 100 50 0 1 2 3 4 5 6 7 8 9 10 11 12 Week prior
Dynamic Pricing Algorithms Function of Demand variation over price and time Mathematical Programming Probability Distribution of Demand variation over price and time Dynamic Stochastic Programming
Techniques References Theory and practice of Revenue Management Talluri & Van Ryzin Springer 2004 Algoritmi per il Single Resource Capacity Problem Algoritimi per il Network Resource Capacity Problem Price-Based Revenue management, Fabio Colombo, DTI-UNIMI 6 th Annual Revenue Management & Price Optimization Conference, Workshop Introduction to the science of Revenue Management an Price Analytics Mark Ferguson, Atlanta 2010
Industry Applications Airlines Hotels Restaurants Retailing Manufacturing
Tipology of Revenue Mangement Price Predictable Fixed Movies Stadiums and arenas Convention centers Variable Hotels Airlines Rental cars Cruise lines Duration Unpredictable Restaurants Golf courses Internet service providers Continuing care Hospitals Kimmes1998
Hotels Room occupancy Out-house food and beverage Events, conventions conferences Out-house food and beverage Reservations Controls Property Management Systems (PMS) GDS Just 20% Applied techniques: Overbooking (Hadjinicola&Panayi 1997) Traditional nested allocation and bid price minimum/ maximum length of stay control intradays/hour forecast updates and optimization Hanks, Cross Noland 1992 Kimes1989 Orkin1988 Varini et al 2002 Burns 200 Bitran and Modschein 1995
Airlines Reservations Controls (Allocation Overbooking) Airline host reservation systems GDS Sales/ CRM Barnhart Talluri 1997 Call center Applied techniques: Overbooking Web Fare classes server Public fare x private fares Internet-only fares Capacity controls by availability fare classes posted on GDS
Manufacturing Few current applications of RM (Gray 1994). SCM and ERP vendors are starting to offer pricing optimization systems. Service Perishable inventory Order has to be denied RM Applications opportunities Manufacturing Inventory can be stored for future use Order can be delayed Perishable products like high-tech products and concrete. (Kalyan 2002, Elimam &Dodin 2001) Demand smoothing: delaying production of low-value demand to off-peak times, while ensuring prompt production of high-value demand during peak times.
Restaurants Revenue drivers Price Meal duration Maximize Revenue per available seat hour (RevPASH) Possible research area: Menu engineering Main author: Sheryl E. Kimes - Cornell University
Retailing Retail RM system POS ERP/SCM Dynamic price bottleneck Menu costs (Cost on changing prices) Store's historical data Retail RM systems functions: Price Optimization Automate routine price changes by location and channel Monitoring profit and sales targets for items and categories Tracking performance of promotions and advertising campaigns Maintaining consistent pricing and rounding rules Automating price matching based on competitors prices Supporting price-sensitivity experiments Generating reports ans statistics to track pricing performances ESLs (electronic shelf labels)not widely deployed Girad 2000 Johnson 2001 Mantrala&Rao 2001
Applications References Theory and practice of Revenue Management Talluri & Van Ryzin Springer 2004 6 th Annual Revenue Management & Price Optimization Conference, A Tag Approach to Effective Communication Randy Light and Bill Dudziak, The Home Depot, Atlanta 2010