Demystifying Price Optimization: A Revenue Manager’s Guide WHITE PAPER SAS White Paper Table of Contents Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 What Is Revenue Management? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 What Is Revenue Management Science?. . . . . . . . . . . . . . . . . . . . . . . 2 What Has Changed?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Taking Revenue Maximization Beyond the Travel Industry. . . . . . . . . 4 Price Optimization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Revenue Management in the Future. . . . . . . . . . . . . . . . . . . . . . . . . . . 6 How SAS® Is Different. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 For More Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Bibliography. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Content for this paper was provided by Alex Dietz, Principal Industry Consultant for the SAS Hospitality and Travel practice. He is a 25-year veteran of pricing and revenue management solutions’ development and consulting in hospitality, travel and retail. Prior to joining SAS, Dietz was Vice President of Revenue Management and Marketing for Raleigh-Durham based Midway Airlines. He began his career with American Airlines and SABRE, where he oversaw development of industry-leading revenue management systems for the airline industry. Dietz has a bachelor’s degree in industrial engineering and operations research and a master’s degree in business administration – both from Syracuse University. Demystifying Price Optimization: A Revenue Manager’s Guide Introduction Revenue management has evolved into a very different science from what it was when airlines first started using it in the mid-1970s. Today, revenue optimization is common throughout travel and hospitality sub-industries including cruise lines, rental cars, hotels, ferries and passenger railways. As the practice has expanded, the science behind revenue management has been influenced by trends that invalidate important assumptions that were built into the original models. As analytic approaches to pricing have evolved, we at SAS have often heard the question: “How is price optimization different from revenue management?” It’s a simple question that has become increasingly difficult to answer – largely due to the evolution of revenue management practices and the definition of revenue management. As this paper explains, the question itself reflects a mindset based on historical business practices. SAS recognizes that today’s travel and hospitality marketplace encompasses aspects of both revenue management and price optimization. To give customers the best possible results, we recommend an analytical approach that takes into account both revenue management and price optimization techniques. What Is Revenue Management? To truly understand revenue management, it’s important to start with today’s definition and review how it has changed over time. Wikipedia defines revenue management as “the application of disciplined analytics that predict consumer behavior at the micromarket level and optimize product availability and price to maximize revenue growth.” This definition is very broad, but probably well in line with the thinking of most travel revenue managers and executives. Practitioners and industry experts alike think of revenue management not as a science but as a wide set of business practices with a common goal. By this definition, price optimization is no different from revenue management, and might be better described as a specific methodology for applying revenue management. To get a better understanding of how price optimization evolved from being a separate discipline into being a tool in the revenue management toolbox, we need to explore the development of revenue management as a science. 1 SAS White Paper What Is Revenue Management Science? The application of mathematical models, specifically to travel pricing, can be traced to the mid-1970s. Before then, the use of analytics in travel sales was limited to approaches surrounding overbooking. But in the mid-’70s – after deregulation of the US airline industry – a number of airlines introduced discounted fares that were sold into the same compartment as the already-available full fares. These new discounted fares generally required significant advance purchase: 21 days. The new discounts allowed the airlines to fill seats that would have otherwise gone unused on low-demand flights. But as a result, the airlines needed a method to control the number of reservations taken on these reduced fares when demand for late-booking, full-fare passengers was expected and when total demand from both discounted and full fares would exceed capacity. Managing these sales required airline systems’ inventory control capabilities to evolve so that analysts could control the number of discounted sales on a day-by-day and flight-by-flight basis. Next, the airlines formed yield management groups. The yield manager’s role was to maximize revenue on each flight by predicting the sale of full-fare reservations and controlling the number of discounted seats sold accordingly. The amount of space to reserve for these future reservations was (and still is) referred to as protection. As the industry evolved this practice, airlines developed increasingly complex fences (restrictions placed on discounted fares) to distinguish these fares from full fares (and each other). The fences reduced dilution (reduced revenue caused by a customer purchasing a discounted fare that was not intended for his market segment). Airlines enhanced their selling systems to enforce these restrictions as they were developed and adapted. Examples of fences include: • Advance reservations. • Nonrefundable fares. • Minimum length of stay (including the infamous “Saturday night stay” requirement). • Maximum length of stay. • Day-of-week applicability. As this business process evolved, mathematicians began to model the problem based on the manner in which the business was being run at that time (ca. 1985). Their goal was to find the mathematically optimal protection levels that would maximize revenue. The business had evolved to the point where multiple protections would need to be calculated for any given flight departure – because there were now many different discounted fares available, each with a different value and customer target segment. In effect, the mathematical problem involved determining the maximum number of each of these different discounted products that should be sold to maximize revenue. The early systems developed for this purpose were known as yield management systems, and they mark the beginning of revenue management as a science. 2 Demystifying Price Optimization: A Revenue Manager’s Guide The most involved yield management systems attempted to account for complexity due to the network effect: specifically, the problem of forecasting demand for a specific segment (or leg) of an itinerary, when many different kinds of itineraries (all with different values to the airline) also flowed over it. The question was critical to growing hub-andspoke carriers where significant percentages of passenger itineraries included a stop at the carrier’s hub. What Has Changed? The original formulations for revenue management science were based on how the airlines ran their businesses. Fenced fare structures were critical components of this business model, because they clearly separated different customer groups with distinctly different values. And, although these fences differed across disparate customer groups, a key modeling assumption was that these different values were relatively stable over time. (This assumption arose in part because of the fairly small computing power of older systems compared to today’s systems.) This assumption of stability allowed yield management systems to assume a level of independence between customer groups. For example, the systems assumed that a customer for a 21-day, nonrefundable advance purchase (AP) fare with a Saturday night stay requirement would not purchase a full fare if the 21-day AP fare was sold out – and that (similarly) a full-fare customer had no interest in that 21-day AP fare if it was available. In effect, this assumption meant that each different fare type could be treated as a separate product which serviced a separate market of customers, and which happened to utilize the same inventory unit (the airline seat) in doing so. What is revenue management, and how did it evolve? Application of disciplined analytics that predict consumer behavior at the micromarket level and optimize product availability and price to maximize revenue. 1980s Airlines introduce discounted fares & revenue management 1990s Revenue management spreads through travel & hospitality industries 2000s Limitations of inventory optimization solutions result in trend toward price optimization Figure 1: How did revenue management get to where it is today? Three important trends have emerged since these original methodologies were developed: • Low-cost airlines introduced simplified fare structures with fewer fares and significantly reduced fencing. Today, these simplified fare structures dominate most air travel markets – effectively invalidating the assumption of demand independence made in revenue management models. 3 SAS White Paper • Revenue management has been introduced into markets where strict fences on rates or fares never existed. Hotels, rental cars, cruise lines and others have rate or fare structures that do not contain strict fences – so the assumption of demand independence is again problematic. • Rates and fares in these simplified structures have become increasingly dynamic. Revenue managers now frequently manage the price at which rates and fares are sold on a day-to-day basis. These trends have taken travel and hospitality practices far afield from the expectations of the original revenue management scientists who modeled the airline revenue management problem. Of course, revenue management scientists have tried to overcome the underlying limitations in the original construct of the problem – generally, they have done this by introducing changes to forecasting and optimization approaches that allow for some measure of interdependence of demand for different fares or rates. Consequently, we see common references to “sell up” or “buy down” probabilities associated with forecasting and optimization in such systems. There are also other business elements generally required for the application of this original revenue management approach: • Fixed available capacity – capacity cannot be increased to accommodate surplus demand. • Product perishability – the product loses its value to a customer after some point in time and cannot be sold thereafter. • Advance reservations – advance reservations are accepted for the product. These underlying assumptions continue to be valid for travel companies, so they have not caused the same types of concerns as the other trends. Taking Revenue Maximization Beyond the Travel Industry While many travel and hospitality companies were struggling to make revenue management approaches work, other industries were considering how to maximize their own revenue and profitability – and recognizing that the airlines’ original revenue management approach simply would not work for them. Generally, the lack of fit was due to an inability to meet one of the required business elements described previously. Imagine if you were a retailer trying to apply revenue management techniques: • You don’t take advance reservations for your product – your customers come to your store and purchase the product while they are there. • Most of your products aren’t really perishable – if you don’t sell a shirt one day, it will be there to sell the next day. 4 Demystifying Price Optimization: A Revenue Manager’s Guide • You don’t have a fixed capacity – if you start selling more milk, you can buy more milk to sell. • All of your customers pay the same price – they come in, see the price tag on the item or shelf, and that’s what they all pay. The bottom line: Your business fails to meet every one of the elements that are needed for the application of traditional revenue management. But does that mean retailers can’t use analytical modeling and/or operations research techniques to maximize revenues? Of course not! They just need a different approach. Price Optimization Price optimization approaches were developed to suit industries whose characteristics don’t match that of the original revenue management approaches. In these businesses (retail is a good example) products are made available at a single price to a broad market of customers. In this broad market, each customer has a different ability or willingness to pay, and the business can change the product’s price over time. If the price is raised, it begins to exceed the price that most customers are willing to pay – so demand decreases. If the price is lowered, more customers are willing to pay for it – so demand increases. This is simple microeconomic theory. Using a price optimization approach, product demand is estimated as a function of price. So the price becomes the decision variable rather than the protection value that is used in maximizing revenues or margin. In businesses that use price optimization, there are usually no advance reservations or separation by segment, so there’s no one to “protect” sales for. In addition, inventory may not be constrained – so there’s no reason to protect inventory, either. Since the market is not be segmented by price, the question becomes “what is the optimal price to charge the overall market to maximize return, accounting for the fact that demand will change as the price changes?” The retail industry has widely adopted price optimization approaches, but these approaches are also used by other industries. Decision support solutions that utilize price optimization approaches (like the SAS Revenue Optimization Suite for retail1) need to estimate demand and its sensitivity to price (i.e., price elasticity) for a wide variety of products, and often from a wide variety of channels or stores. The calculation of price sensitivity on a large-scale, automated basis introduces new technical challenges that these solutions have had to overcome – much research has been invested in this area over the last decade or so. 1 SAS Revenue Optimization Suite: sas.com/industry/retail/revenue-optimization/index.html 5 SAS White Paper Revenue Management in the Future The trends that have taken hold following the initial development of revenue management methods have undermined traditional revenue management science, because they invalidate important assumptions built into the models. Today, rather than selling a series of clearly fenced fares, many travel companies – including airlines and passenger rail lines – offer different fare families. Fare families are distinguished from each other using a set of purchase restrictions (e.g., refundable) or benefits (e.g., variations in frequent flyer points earned, or standby benefits). In effect, each fare family represents a different fare product, and the price for each fare family can be changed on a day-to-day and flight-by-flight basis. Fare families represent a two-fold problem for the traditional revenue management approach: • There is no independence of demand for fares within a fare family – all of these fares have the same set of rules and restrictions, and so the assumption of independence fails completely for different fares within a fare family. Plus, the choice of higher or lower prices is known to influence demand – and this effect needs to be modeled in some manner. • The differential in restrictions and benefits between fare families is known to be relatively weak, so when price differentials between fare families are too large on a given flight, a customer will “jump” from choosing one fare family to choosing a different fare family. Accommodating this complex type of behavior requires a demand modeling approach that considers demand for each flight as a function of both the availability and pricing of each fare family on the flight. Consider the example of a roadside hotel. This business meets the primary conditions of revenue management (fixed capacity, perishable product and advance reservations); but it probably does not have clearly fenced segments. Rather, this business chooses one of several possible rates to charge on any given date, and that generally available rate services the vast majority of its customers. Once again, there is no independence of demand between these rates – they all have the same set of characteristics, so the hotel guest will always select the least expensive one. These types of “priceable” products represent a significant portion of today’s travel and hospitality market. Most travel and hospitality firms have such a product, and for many of them, it is a significant portion of their business. For priceable products, a price optimization approach seems to be a good fit for the business problem. Because with this approach the selection of an optimal price is the primary output of the model, and price elasticity is directly modeled. 6 Demystifying Price Optimization: A Revenue Manager’s Guide In the airline industry, two new approaches to accounting for these changes have evolved: • Continued investigation into estimating up-sell (within fare families) and cross-sell (between fare families) to inform revenue management models. This methodology continues to represent the current state-of-the-art for revenue management modeling for airlines. • Investigation into the use of choice models as a more robust approach, which accounts for fare family rule and benefit differences – and how these are valued versus price differentials between the fare families. While theoretically attractive, these investigations have not yet borne fruit in terms of actual applications. In the hospitality industry (including hotels and cruise lines), priceable products make up a significant portion of the market – but not all products are priceable. Many customers purchase rates and fares that are controlled not through price, but through availability. For example, many promotions are managed this way, and some segments can be both priced and managed via availability. These changes cause much of the confusion surrounding revenue management and price optimization approaches. In fact, today’s travel and hospitality marketplace has aspects of both revenue management and price optimization problems. Hybrid analytic approaches combining aspects of both revenue management and price optimization approaches are the best solution for today’s marketplace. The hybrid approach also allows for market changes that will continue to take place over time. How SAS® Is Different At SAS, we have developed modeling approaches that allow for optimization of both priceable and yieldable rates and fares in SAS Revenue Management and Price Optimization Analytics. These models can forecast demand for both price-elastic, priceable fares, as well as those that cannot be priced. Our optimization approach calculates both optimal price (for priceable fares) and optimal protection levels (for fares or rates that must be managed via availability). Traditional Revenue Management SAS® Revenue Management and Price Solution Approach Optimization Analytics Limited number of demand models and parameters. Automated forecast model selection from a broad portfolio of models, and automated parameter optimization per product and segment. Price elasticity ignored or treated as a post-demand modeling process. Price elasticity incorporated directly into demand modeling. Decomposed pricing and inventory control optimization. Integrated pricing and inventory control optimization across product classes. Figure 2: A comparison of a traditional revenue management approach and SAS® Revenue Management and Price Optimization Analytics. 7 SAS White Paper Finally, and perhaps most importantly, our optimization approach treats pricing and availability decisions as one integrated problem – and makes calculations accordingly. After all, changes in price can affect demand, which affects optimal protection for yieldable products. Similarly, changes in the availability of yieldable products influences decisions regarding the amount to charge for priceable rates and fares. Revenue management and price optimization – the two must work together. Conclusion The major advantage of SAS’ hybrid approach is that it models the problem as it exists – and does not require significant business shifts or related technology changes to maximize revenues. Optimizing yieldable and priceable products together enables companies to develop pricing and inventory approaches that suit the needs of their market, but are still supported by high-performance analytical modeling. The approach also offers flexibility in supporting new sales and marketing approaches in the future. Other hybrid approaches tackle pricing and yielding as separate decisions – first calculating optimal protection levels, and then separately calculating optimal pricing using the protections output from the first step as input to that decision. At SAS, we have found that this approach of arbitrarily breaking the problem into multiple steps results in suboptimal results. And there are revenue benefits to an approach that considers the entire business model. Our testing shows significant revenue benefits of the integrated analytics approach used in SAS Revenue Management and Price Optimization Analytics versus that used in other solutions – on the order of 2 percentage points in revenue growth. This is the future of revenue management and price optimization: flexible, integrated models that allow businesses to be modeled as they actually exist. No more round pegs in square holes. No more shoehorning our businesses into processes that support analytics, or attempting to utilize analytics that don’t really fit our businesses. Let the analytics come to the business. 8 Demystifying Price Optimization: A Revenue Manager’s Guide For More Information View the webcast titled New Beginnings in Revenue Management: sas.com/reg/web/corp/2138309 Learn more about SAS® Revenue Management and Price Optimization Analytics for hospitality and travel: sas.com/industry/hospitality/revenue-management/index.html Visit The Analytic Hospitality Executive blog: blogs.sas.com/content/hospitality Bibliography McGill, J.I. and Van Ryzin, G.J. (1999). Revenue Management: Research Overview and Prospects. Transportation Science 33: 233–256. Fiig, Isler, Hopperstad and Olsen (2012). Forecasting and Optimization of Fare Families. Journal of Revenue and Pricing Management, Vol. 11, 3, 322-342. Koushik, D. Higbie, J. and Eister, C (2012). Retail Price Optimization at InterContinental Hotels Group. Interfaces, Vol. 42, No. 1, 45-57. 9 About SAS SAS is the leader in business analytics software and services, and the largest independent vendor in the business intelligence market. Through innovative solutions, SAS helps customers at more than 60,000 sites improve performance and deliver value by making better decisions faster. Since 1976 SAS has been giving customers around the world THE POWER TO KNOW®. SAS Institute Inc. World Headquarters +1 919 677 8000 To contact your local SAS office, please visit: sas.com/offices SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc. in the USA and other countries. ® indicates USA registration. Other brand and product names are trademarks of their respective companies. Copyright © 2013, SAS Institute Inc. All rights reserved. 106156_S100361_0113