Hotel Revenue Management: The Post-Pandemic Evolution to Revenue Strategy PDF

Hotel Revenue Management  Hotel Revenue Management The Post-Pandemic Evolution to Revenue Strategy Dave Roberts Hotel Revenue Management: The Post-Pandemic Evolution to Revenue Strategy Copyright © Business Expert Press, LLC, 2022. Cover design by Charlene Kronstedt Inte...

Hotel Revenue Management  Hotel Revenue Management The Post-Pandemic Evolution to Revenue Strategy Dave Roberts Hotel Revenue Management: The Post-Pandemic Evolution to Revenue Strategy Copyright © Business Expert Press, LLC, 2022. Cover design by Charlene Kronstedt Interior design by Exeter Premedia Services Private Ltd., Chennai, India All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means—electronic, mechanical, photocopy, recording, or any other except for brief quotations, not to exceed 400 words, without the prior permission of the publisher. First published in 2022 by Business Expert Press, LLC 222 East 46th Street, New York, NY 10017 www.businessexpertpress.com ISBN-13: 978-1-63742-191-8 (paperback) ISBN-13: 978-1-63742-192-5 (e-book) Business Expert Press Tourism and Hospitality Management Collection Collection ISSN: 2375-9623 (print) Collection ISSN: 2375-9631 (electronic) First edition: 2022 10 9 8 7 6 5 4 3 2 1 History is often shaped by small groups of forward-looking innovators rather than by the backward-looking masses. —Yuval Harari, Homo Deus  Description This book guides the reader from the building blocks of revenue management, to pricing science and merchandising, and to broader issues of setting objectives in support of a revenue strategy. The discipline is evolving, and that evolution has been accelerated by the COVID-19 pandemic. Leaders in hotel revenue management, and more broadly in sales & marketing, need to understand this evolution, and lead and adapt accordingly. This will require a strong foundation in analytics— not just modeling, but also business analytics in support of a holistic strategy. As more of the tactics of revenue management are executed through automation, and powered by machine learning, revenue managers will become more focused on strategy, and will need to think about revenue management in the larger commercial context of marketing, loyalty, and distribution. As the strategy component of the discipline increases, so too must the breadth of knowledge of revenue managers. Keywords revenue management; hospitality; hotel; pricing; strategy; analytics; business; optimization; goal setting  Contents Foreword Acknowledgments Introduction Chapter 1 Context Chapter 2 Building Blocks of Revenue Management Chapter 3 Forecasting Chapter 4 Inventory Management Chapter 5 Pricing Chapter 6 Discounted Rates Chapter 7 Negotiated Account Rates Chapter 8 Distribution and Loyalty Chapter 9 Merchandising Chapter 10 Total Hotel Revenue Management Chapter 11 Revenue Management in a Downturn Chapter 12 Revenue Management in a Recovery Chapter 13 Machine Learning in Revenue Management Chapter 14 Topline Analytics Chapter 15 Talent Chapter 16 Thoughts for the Future Appendices About the Author Index  Foreword The future ain’t what it used to be. —Yogi Berra Revenue Management is at a watershed. Hospitality Revenue Management had been rapidly evolving for the past decade as technology advanced and hotel firms became more astute about leveraging the vast amount of knowledge and data in Revenue Management practitioners and databases. The discipline was advancing in a steady, linear, and somewhat predictable, path. Then COVID-19 intervened, completely disrupting the longest period of economic expansion the modern world has ever seen. Recovery from that economic devastation is just beginning, but the residual effect on hospitality is completely unknown, perhaps unknowable. The economy will not ‘snap back’ to a pre-COVID state. The only certainty is uncertainty. Business travel demands are shifting as the pandemic has made millions of people in business much more comfortable with video conferencing and collaborating remotely. They may be slow to resume travelling as corporations extend budgets cuts. Discretionary travel, especially international travel, may be dampened by fear or legal restrictions such as certificates of immunity. On the other hand, there is a real possibility that travel will see a resurgence resulting from pent-up demand. In this period of ambiguity, we need guidance. There is no script or playbook for the post-pandemic recovery, but thanks to Dave Roberts, we have assistance. The title of this book, Hotel Revenue Management, the Post-Pandemic Evolution to Revenue Strategy, reveals the secret to success in the coming years. Revenue streams will continue to be unpredictable. It will not be sufficient to manage revenue, no matter how clever one is or how much technology is deployed to the effort. Strategies must be developed that will actively create revenue streams, not just manage the demand. Revenue Management must evolve to Revenue Strategy. Any good book should challenge your thinking. A great read is one that validates your knowledge, but then, unexpectedly, it confronts you, perhaps uncomfortably. Do you really know what you thought you knew? This book provides a penetrating analysis of existing Revenue Management thinking and practice. For me, it was a joy to see an experienced practitioner methodically dissect the fundamentals of Revenue Management—forecasting, pricing, inventory management and distribution. The book has a direct, no BS approach to what works, and what doesn’t work. More important than the analysis of contemporary techniques, this book describes the future of Revenue Management, which ain’t what it used to be. For years, the rigorous analytical approaches of Revenue Management began to find their way into other functions such as sales, marketing and distribution. Vast Revenue Management databases and experienced revenue managers would support these functions. Revenue Strategy flips the script so that the analytical rigor and process of Revenue Management will not just support, but it will drive all decisions involving customer acquisition and retention. Revenue Strategy will be game changing, and it will be essential for success in the post-pandemic economy. Indulge me a few words about the author, Dave Roberts. Dave is one of the unquestioned leaders in Revenue Management. Aside from the unique experience he has gained from his increasing roles and responsibilities at Marriott, he has earned respect from everyone in this discipline. I’m honored to be his friend. His passion for Revenue Management permeates this book. The book is not a stiff and formal treatise. It’s conversational. Reading the book was like being beside Dave as he grew and learned in the space. He shares his thoughts, and he even verbalizes his asides, so that the reader has the full benefit of his thinking. Enjoy the journey with him. There is a lot to learn. —Bob Cross Chairman, Revenue Analytics  Author, Revenue Management, Hard-Core Tactics for Market Domination  Acknowledgments In the course of writing this book, I had the opportunity to work with and interview some outrageously talented business leaders, and top-notch academics, including three winners of the HSMAI Vanguard Lifetime Achievement Award (and likely some future award winners as well). Their wisdom and insights are reflected in this book, and I am enormously grateful to them. In alphabetical order, they are: Chris Anderson, Professor, Cornell University Brian Berry, Executive VP, Commercial Strategy, Pyramid Hotel Group Jason Bryant, Cofounder and CEO, Nor1, an Oracle Company Matt Busch, Senior Vice President, Equifax Bob Cross, Chairman of Revenue Analytics Dax Cross, CEO of Revenue Analytics Sloan Dean, CEO of Remington Hotels Craig Eister, Former SVP of Global Revenue Management & Systems, IHG Cindy Estis Green, CEO and Cofounder of Kalibri Labs Erich Jankowski, VP of Commercial Strategy, Host Hotels & Resorts Pavan Kapur, Chief Commercial Officer, Caesars Entertainment Sherri Kimes, Professor, National University of Singapore Klaus Kohlmayr, Chief Evangelist & Head of Strategy, IDeaS Revenue Management Solutions Mike Lukianoff, Data Science Advisor Kelly McGuire, Managing Principal, Hospitality, ZS Associates Juan Nicolau, Professor of Revenue Management, Virginia Tech Andrew Rubinacci, EVP, Revenue Strategy, Aimbridge Hospitality  Trevor Stuart-Hill, President of Revenue Matters Tim Wiersma, Founder of Revenue Generation, LLC I’d like to thank Glenn Withiam, my supertalented editor, for all of his advice on this book. Thank you also to Business Expert Press, the publisher of this book, and a widely recognized leader in education for students and professionals alike. I am forever grateful to my many outstanding bosses, teams, and colleagues at Marriott, especially those in revenue management, but also to so many others, across several disciplines, at all levels, and all over the world. Please know that if I have ever worked with you, I have learned from you. And a special thank you to Bob Cross, recognized by many (including me) as the foremost expert in the world in revenue management, for his thoughtful foreword to this book, as well as his friendship and mentorship over so many years.  Introduction Thank you for your interest in this book, and in this wonderful discipline. Revenue management is the love of my professional life, and I hope that passion becomes obvious to the reader. The purpose of this book is to share some things I’ve learned along my 25-year journey in the hospitality industry, as well as a vision for the post-pandemic future. I’ll make several references to the pandemic throughout the book. In fact, there is an entire chapter dedicated specifically to revenue management in a downturn, and another dedicated to revenue management in a recovery. While we all hope that COVID-19 will soon be mostly behind us, to far too many, the pandemic has been catastrophic, and often deadly. To a great many more, it has been unsettling in the extreme. To state the obvious, the global pandemic changed many aspects of our personal and professional lives. We will likely see that it changed the nature of revenue management itself. For example, we will certainly see an increased focus on cost containment and automation (true for the industry overall as well). That said, the pandemic has not changed and will not change many concepts and fundamentals, nor will it change the fact that the discipline of revenue management will continue its evolution to revenue strategy, which is the focus of this book. In fact, the COVID-19 pandemic has accelerated that evolution. The book is geared toward revenue management practitioners at all levels and in all functions. Some practitioners may be new to revenue management, while others may have extensive experience. It is my intent that even the most seasoned revenue management experts, many of whom I am fortunate to know personally, will glean value from this book. In addition, my hope is that this book is useful for anyone in the hospitality business, not just those in revenue management. I firmly believe that a grounding in revenue management is essential for any leader in hospitality, regardless of job function or title, and I wrote this book with that in mind. I hope this is also useful for students and teachers, as supplemental reading, as opposed to as a textbook. I’m hopeful also that parts of this book are useful well beyond the travel business, even though hospitality is my focus. Revenue management is a growing discipline, and is applicable to many non-travel industries, especially those with capacity constraints, from broadcast advertising to storage units to golf courses, and more. The pricing component of revenue management is of course applicable to virtually every business in every industry. Although this book functions as a unified whole, it is structured in stand- alone chapters. Thus, it can be read end to end, or you can select specific chapters based on your own background and interest. Given the complexity of this discipline, each topic is inextricably linked to several others, making a logical flow of chapters a rather personal preference (this book reflects my personal preference). In each chapter, I will reference other related chapters, aiming to make it easier for the reader to follow, and connect the themes. The chapters themselves represent concepts in revenue management, though many discipline issues cross multiple concepts. For example, channel distribution and alternative lodging are issues that involve several revenue management concepts, from pricing to forecasting to merchandising, and more. To the extent possible, I’ve kept the bulk of an issue within a given chapter. You’ll also see that the topics of talent and analytics seem to permeate every chapter (that, by itself, is an important theme for revenue management), and each of these also has a dedicated chapter. While I have tried to address the most important parts of this great discipline, there will necessarily be some omissions. As with any book of this nature, I had to balance scope with brevity. The content of this book reflects my own bias, specifically my above-property lens of revenue management, as this has been the focus of my career. My intention is to share a grounding in revenue management, and then to look ahead. I will share a vision for the future, and the steps we need to take to make that vision a reality. I’m a bit of a hoarder by nature. I keep everything, including every notebook I’ve ever used in my professional life. The photo below shows several of them, from my time at Marriott, in chronological order, from 1996 through 2019. Figure I.1 Notebooks from 1996–2019 In preparing this book, I reviewed all of my notebooks, looking for themes to include. In some cases, I was thrilled to see how much progress we have made as a discipline; in others, I was surprised to see that there is still much to be done. Let’s jump in.  CHAPTER 1 Context What Does the Evolution to Revenue Strategy Mean? Figure 1.1 HSMAI ROC, 2019 In June 2019, six months before we learned of the pandemic, I gave a keynote address at the Hotel Sales and Marketing Association International Revenue Optimization Conference (HSMAI ROC) in Minneapolis, Minnesota. The topic was “The Future of Revenue Management,” and the theme of that talk was “revenue management will evolve into revenue strategy.” The pandemic has accelerated this evolution, and that evolution is the guiding theme of this book. “Revenue management will evolve into revenue strategy.” That sounds like a bit of a platitude, so let me explain what I mean. First off, I am always careful about using the word strategy. It may not be the most overused word in the English language, but I believe it to be the most overused word in a business context. By strategy, I do not mean a business plan, a mission statement, or some lofty vision. I mean a path to differentiated results—a set of actions and decision-making guidelines that add value to the organization. A successful strategy is a recipe for success; this certainly applies to revenue strategy. The evolution to revenue strategy, which is already underway, has some important implications. It means that the tactical decision making of revenue management will increasingly be done by technology, leaving the strategy work to us humans. The tactics of forecasting, pricing, and inventory management are quite well suited to modern technology, and we will certainly continue down that path. But the evolution to revenue strategy is more than merely a strategic approach to revenue management. I believe that the evolution underway is: tactical revenue management > > strategic revenue management > > revenue strategy. Revenue strategy is quite different from strategic revenue management, and we’ll revisit this in future chapters. At this point, I’d like to mention what some futurists claim to be the organization of the future: a human, a dog, and a computer. The human’s job is to feed the dog, and the dog’s job is to make sure that the human doesn’t touch the computer! This is certainly an exaggeration, but the theme is real: computers will take on an increasing share of what humans currently do. I believe Andrew Yang has been right about this all along. The portion of a revenue manager’s day that is devoted to tactical decision making will certainly decline. The revenue manager will need to understand how their system works (if they have one), ensure that all inputs are valid, decide when to override the system, and understand the impacts of those overrides. The portion of a revenue manager’s day that is devoted to strategy will increase. This means taking a holistic look at the topline revenue for a hotel or set of hotels, in a way that a computer cannot. It also means figuring out ways to get better at the tactics, including making wise choices for technology investment and decision-support analytics. And it certainly means setting the right goals, developing plans to achieve those goals, and analyzing and communicating progress. There are many aspects of revenue management that I love. One is this: as a discipline, we continue to make meaningful progress each year. One can look back a few years and be quite impressed with how far we’ve come. I’m certain this will be the case many years into the future. For those of you that are now in revenue management, this may be the most exciting time to be in this discipline (pandemic aside for the moment), and I believe the same can be said next year, the year after, and so on. What I’ll share now are a few areas that require some attention for this great discipline of revenue management to evolve to revenue strategy; these were the themes of my talk at the HSMAI conference. The five areas I describe are, by necessity, only a subset of all of the exciting areas of revenue management. Each of these five areas (which I’ve lettered A through E) will be described in much more detail later in the book, along with several other topics that are critical to the evolution to revenue strategy. My intention here is to give a sense of where we are in this discipline and where we are headed. That said, the notion of “where we are as a discipline” is ripe for misinterpretation. To put a fine point on it, we aren’t anywhere. Different organizations, and even parts of organizations, are at very different points in the evolution to revenue strategy. For example, as we’ll discuss later in the book, the great majority of hotels do not yet have a revenue management system (RMS), and yet some organizations are pursuing the use of Artificial Intelligence to provide real- time pricing at the micro-segment level. However, regardless of where your hotel or organization is, my hope is that you will glean some wisdom from this following discussion, as well as from the rest of this book. (A)Forecasting One area that is in need of progress is forecasting. By this, I mean demand and supply forecasting in support of decision making, as opposed to higher-level projections, such as next month’s Revenue Per Available Room (RevPAR). As an industry, we have put a lot of time and money into demand forecasting, and we have some of the brightest minds working on it, and yet... we are not as good at this as we need to be. The tactical decisions of pricing and inventory are based on these forecasts (at least they should be). Why aren’t we better at this? Partly, it has to do with focus. Forecasting is no longer a glamorous part of revenue management, and it can get pretty technical. We have significant opportunity to improve the science of forecasting, from the inputs to the modeling to the measurement. And the COVID-19 pandemic just magnified the importance of forecasting; as demand  became less inherently predictable, the forecasting challenge got harder. Much more on forecasting can be found in Chapter 3. (B) Pricing Another area for improvement on our path to revenue strategy is pricing. There is a lot of opportunity in the field of price optimization, meaning the price recommendation engines that are a key component of today’s RMSs. To be blunt, however, there is quite a lot of price optimization already in place that is being ignored. Really. Most RMSs today will recommend pricing, at least for retail rates (retail in this context refers to the nondiscounted rates for standard room types). Some RMSs go much further than that, as we’ll see later. Based on many discussions with industry experts, as well as my own experience, these recommendations are overridden perhaps one-third of the time, and some of those overrides are quite significant in magnitude. Have we really built sophisticated price optimization software that is only used when it aligns with the user’s intuition? We, as a discipline, can do better than this. We also need to expand price optimization to all segments and revenue streams, recognizing that this is a long-term effort. Much more on pricing can be found in Chapter 5. (C)Total Hotel Revenue Management Total hotel revenue management (THRM) refers to managing demand across multiple revenue streams. In its simplest form, it means managing transient, group, and local catering demand for both sleeping rooms and function space. More advanced THRM involves more revenue streams such as restaurants, outlets, and spas. Revenue management professionals, and others, have been talking about THRM for well over a decade, in some cases, much longer. Several surveys suggest that this has been identified by many as a significant opportunity for many years. THRM makes sense intuitively, and many companies have invested significant time and money into this. And yet... despite some pockets of progress, we as an industry are not very good at this. This suggests some significant impediments. One impediment is objectives. For example, if you believe that the goal of our hotel’s restaurant is to maximize profits, but I believe it is to drive overall satisfaction with the hotel, and we are evaluated and compensated according to those goals, then we will surely not agree on many decisions. THRM is also hampered by lack of decent quality data. While  reservations systems and property management systems (PMSs) can provide a great deal of useful data upon which to make decisions, the same is not at all true for most F&B or Spa outlets, for example. You’ll find much more detail on THRM in Chapter 10. (D)Topline Analytics One important step in the evolution to revenue strategy is for revenue managers to be integrally involved in all revenue decisions. This involves revenue management decisions with aligned-upon objectives, perhaps extending beyond short-term profit, as we’ll discuss later. It also involves revenue management decisions for purposes of customer acquisition and customer retention (those pricing and inventory decisions could be quite different). And as hotels continue to develop more appealing features, revenue management must be involved in decisions of demand capture. For example, as hotels offer the ability to choose your room at the time of booking, this value could be captured in terms of a loyalty benefit, a channel benefit, a price increase, or simply a demand increase, to be captured with a combination of rate and occupancy. The revenue manager of the future will need to help guide decisions of the hotel/organization with a compelling narrative. The foundation, though not the entirety, of this guidance will be analytics, both decision-support analytics and performance management analytics. Much more on topline analytics can be found in Chapter 14. (E) Talent To state the obvious, talent is how you win. There are many facets of this, some of which I’ll touch on throughout this book. Training in revenue management significantly lags the discipline. This is true all across the industry. Training, both revenue management specific as well as more general training for revenue managers, tends to be an after- thought in many cases. For the continued evolution to revenue strategy, we need a mindset of lifelong learning. We also need to bring in external talent into this discipline, with fresh ideas and new perspectives. However, I hear the following far too frequently: “I could never work in revenue management... you all just stare at a computer screen all day.” Clearly, some PR is called for here. In addition to bringing in external talent, we need to send our revenue management talent out and seed our respective organizations. I dream of a day when every leader in every  discipline at every hotel, and every above property leader has a revenue management background! Much more on talent is found in Chapter 15. These five topics are intended to give you a flavor of what’s coming in this book and to provide structure for discussion of a complex discipline. These and many other concepts will be discussed, with some historical context and with a look to the future. So, what exactly is meant by revenue strategy? As noted earlier, a strategy is a recipe for success; if it’s not a recipe for success, it’s not a strategy. Deciding what your organization is going to do is of course vital to success; deciding what you’re not going to do is equally important, and often much more difficult. If you can’t name some potentially worthwhile endeavors that you will not be undertaking, then you haven’t made the tough tradeoff decisions and you don’t have a strategy. I recommend using that litmus test anytime you see or hear the word strategy in any of your business discussions. Revenue strategy, then, means a recipe for success in revenue generation and capture. It does not mean revenue maximization, or even profit maximization (more on that in the next chapter). It is the set of decisions, based on clearly documented and broadly understood objectives, that determine how we spend our time and money. As Cindy Estis Green, CEO of Kalibri Labs, puts it, “strategy is really planning and resource allocation.” While this of course applies to revenue management decisions related to pricing and inventory, it also applies up-funnel to all sales and marketing activities. A coherent strategy is highly dependent on clearly articulated goals... I’ll come back to this point in the final chapter. For now, let’s note that these goals must spell out the desired balance between sometimes competing objectives, such as profit maximization versus customer acquisition versus customer retention and loyalty versus channel preference, and more. And yet, our language often does not support such clarity. In fact, in my view, our language has gotten lazy. How many of you have read an article about revenue management that describes the need to maximize revenues and profits? With my apologies to anyone who has ever written that, this is lazy wording! Maximizing revenues and profits is a nonstarter; you can maximize one or the other, but not both (with hypothetical exceptions that do not exist in the real world). An analogy would be the desire to maximize rate and occupancy. Both are desirable, but maximizing one means not maximizing the other. There are plenty of other examples of lazy wording. I mention this here, with a few examples, because wording reflects thinking, and lazy thinking will not move us toward revenue strategy. I hereby beseech anyone in hospitality sales and marketing (certainly including revenue management) to stop saying: future data (there is no such thing, nor can one be over-reliant on historical data, since all data are historical; if you mean forecast, kindly say “forecast” —details are in Chapter 3), understand market dynamics (a common, but useless, platitude), optimal profitability (too vague), healthy mix of business (I don’t know what to say here), optimize our channels (another useless platitude). While I’m on the topic of lazy wording, I cringe when I hear anyone refer to revenue management data or revenue management numbers, usually in reference to some topline metric such as RevPAR Index (RPI). If you hear this, it is likely from someone who doesn’t really understand revenue management or even revenue generation. In the summer of 2020, I read three textbooks on hospitality, looking for ideas to incorporate into my classes; one book describes the role of a revenue manager as “responsible for making decisions that optimize [emphasis is mine] a hotel’s RevPAR.” Is optimize the same as maximize here? If so, say it; if not, describe why not. Given my background in Operations Research, I take optimize and optimal quite literally. These words are so often misused in business that I think we’d all be well served to avoid saying optimize and optimal unless we have a specific objective function and well-defined constraints. By the way, the same textbook says, “the revenue manager is responsible for maximizing occupancy and rate.” Oh my. In the summer of 2021, I took a few online courses, and learned in one of them that the goal of a revenue manager is to “sell every room every night at the optimal price.” In addition to being lazy wording, this one is also quite misleading. SMH (shaking my head). These particular examples are ones I’ve seen myself very recently and unfortunately fairly frequently, but there are plenty of others. Get your goals figured out first, then clearly articulate them to anyone who can impact them, and then figure out your recipe for success. The role of revenue strategy at large organizations is ultimately the purview of the chief commercial officer (actual title, of course, varies by organization). Revenue management plays an important role, but all other topline disciplines are necessary for success. Operating in discipline silos, which based on my interviews is quite common in this industry, is an impediment to success. As Sloan Dean, CEO of Remington Hotels, noted, as we evolve to revenue strategy, the role of chief commercial officer will be reflected at all levels of the organization. Sloan predicts, and I think he is right, that we will see commercial strategy/commercial services leaders at the region, area, and market level soon—for example, an area director of commercial services could be an enticing job! I’ll revisit this topic again in the final chapter of the book. In conclusion for this section, let me say that the future of revenue management is in no sense predetermined. Rather, it goes in whatever direction we take it. I love this quote (often, but not exclusively, attributed to William Gibson), “The future has already arrived, it’s just not evenly distributed yet.” What does this mean? Every major step forward that we’ve seen for this great discipline was once merely a thought in someone’s head. And many of the innovations and breakthroughs we will see in the next several years, including the steps to revenue strategy, are already in someone’s head today—perhaps yours! As a side note, let me add here that HSMAI puts on some great conferences, across many disciplines. And if you have an interest in revenue management, and haven’t been to a Revenue Optimization Conference (ROC), I highly recommend it. The team at HSMAI is outstanding, and these conferences are really well thought out and structured.  CHAPTER 2 Building Blocks of Revenue Management A Brief Primer on the Fundamentals of the Discipline In the previous chapter, we discussed the context and motivation for this book, and described what the evolution to revenue strategy means. In this chapter, to get grounded, I want to take a step back and cover some basic concepts of revenue management. We’ll describe what we really mean by revenue management, and review some terminology that you’ll see throughout the book. Each of these basic concepts will be covered again in much more detail in later chapters. Let’s start at the beginning. What is revenue management? The discipline used to be called Yield Management, which was first used by the airline industry, and referred to using pricing and inventory levers to maximize the revenue yield of a flight or set of flights. Actually, the term “revenue management” is a bit of a misnomer. What we usually mean is demand management. Trevor Stuart-Hill, President of Revenue Matters, describes revenue management as “a business process designed to drive the financial performance of an asset through all market conditions.” At a tactical level, hospitality revenue management is the science and process of making pricing and inventory decisions for the benefit of a hotel or set of hotels. Consider a “typical” hotel where we expect Sundays to be low demand, Mondays, Thursdays, and Fridays to be moderate demand, and Tuesdays, Wednesdays, and Saturdays to be high demand (this pattern is quite common). For example, a given hotel will charge a retail rate of $199 next Tuesday, and will only accept bookings for that day that also include a stay on the Monday before or the Wednesday after (so called “length of stay restrictions”; if Tuesday is expected to sell out, then these restrictions will prevent a guest from only staying on Tuesday, and will allow guests who also stay the day before or after). Or, the hotel will accept some lower-rated advance-purchase bookings for next Saturday, but no more than a predetermined number of rooms. In the early years of revenue management, such decisions were made to maximize revenue. The role of a revenue manager, then, was to find the combination of price points and inventory controls that maximized revenue. As the discipline evolved, we began thinking more about profitability, recognizing that different revenue streams have different profit implications, and our decisions were (at least in theory) made to maximize profits, and the role of a revenue manager changed accordingly. More recently, our revenue management decisions are made to maximize “benefit”—mostly profitability, but also taking into account loyalty and channel impacts, as well as reactions of competitors. For those readers with a finance bent, you can think of this as maximizing the Net Present Value of all future profits (or future cash flows for you finance purists ). These pricing and inventory decisions are based on a forecast. This forecast is a demand and supply forecast, as opposed to a revenue projection. The demand forecast is often an arrival forecast, broken down by rate “bucket” and length of stay. The demand forecast can also be a stay- night forecast, meaning a forecast of roomnights (RN), as opposed to arrivals. This is often used to determine “hurdles” to control inventory (more in a moment). The demand forecast is paired with a supply forecast, meaning: how many rooms do I have available to sell? This supply forecast is based on the physical size of the hotel, rooms already on the books (noting the specific stay patterns), out of order rooms, expected cancellations across all segments, and group attrition. Together, the demand and supply forecasts are the basis for pricing and inventory decisions. Chapter 3 covers forecasting in much more detail. Pricing means putting “price tags” on the inventory, for each arrival date, length of stay, room type, and segment (and sometimes by channel). Some of this pricing is supported by price optimization software, and some is not (yet). Price optimization refers to mathematical modeling to generate recommended price points, details of which are in Chapter 5. For the segments of business for which we do not yet have price optimization, some combination of analytics and intuition is used. The inventory management component is a bit more straightforward than forecasting and pricing. Inventory management is a math problem, or, more precisely, an optimization problem. With a given demand and supply forecast, and a given set of price points, the optimal mix of business for the hotel to book can be determined using some approaches from the field of Operations Research. Sometimes, a company or a revenue manager will knowingly make inventory management decisions that do not maximize profits, for reasons of loyalty or distribution for example, which we’ll cover in Chapter 8. The nuances of inventory management for any given hotel can get fairly complex, especially in peak demand situations, and even more so in a hotel with multiple types of rooms to sell. An experienced revenue manager with sound judgment is critical. Details on inventory management can be found in Chapter 4. Let me touch on two more fundamentals: Dilution and Displacement. I was at an International Air Transport Association (IATA) conference several years ago, and I remember one of the keynote speakers, from a major U.S. airline, saying something to the effect of “as a revenue manager, the most important thing you can do when you get back to work is to educate your sales and marketing stakeholders on dilution and displacement.” The presentation was great, and the message is timeless. Let’s briefly dig into each here. What the airline industry calls dilution, we in the hospitality industry call, a bit more descriptively, tradedown. One of the fundamental principles of revenue management is to charge different people different prices based on their sensitivity to price. I highly recommend Bob Cross’ seminal book, called “Revenue Management—Hardcore Tactics for Market Domination,” which discusses this in detail. The two charts below describe this effect. Both charts show demand as a function of price. For any economists reading this, you’ll note that these axes are transposed (when plotting price elasticity, price is typically on the vertical axis and quantity is on the horizontal axis); I use this orientation here to more clearly demonstrate demand as a function of price. This orientation is also used later in the book when I discuss group pricing models, where the two major inputs are also a function of price. Figure 2.1 shows a single price point. At a price of $50, we capture 100 customers, and bring in $5,000 in revenue (the shaded area is $5,000). Figure 2.2 shows three price points. Now we can capture 50 customers at $75, another 50 at $50, and another 50 at $25, for a total revenue of $7,500 (represented by the combined shaded areas). Clearly, offering three price points is better than offering just one, as shown by the two shaded areas. BUT... Figure 2.1 Demand as a function of price: Single price point Figure 2.2 Demand as a function of price: Multiple price points Consider the 50 customers who are willing to pay $75. How do we prevent them from paying less? Every customer who is willing to pay an offered rate, but who ends up paying a lower rate, represents dilution (tradedown). In this author’s opinion, this may be the most misunderstood aspect of revenue management, and perhaps of all of sales and marketing. Much more on this topic is in Chapter 6. Displacement can be interpreted as opportunity cost. If my hotel books a certain reservation, it may then forgo the opportunity to book something else. As an extreme example, let’s consider my last available room. I have a customer who wants to pay $200 for it right now. But the value of that booking would not really be $200, because if I don’t sell it now for $200, I can perhaps sell it to someone else, maybe for $200, or something more or less. Whatever I forgo represents the opportunity cost, or displacement. Displacement is dependent on demand levels (e.g., there is usually no displacement for a one-night stay on a low demand night because any booking will not preclude a hotel from taking another booking). Displacement is also dependent on the booking window. For example, a booking made on the day of arrival (meaning the booking is made on the same day as the guest arrives at the hotel) will have less displacement than a booking made a week in advance, all else being equal. The reason for this is that the opportunity cost of filling a room at the last minute is pretty low (such a booking would only displace another booking later that day). Finally, displacement is also a function of stay pattern; for example, a stay on a low occupancy Sunday night may incur a displacement cost if the guest arrived on the day before, because Saturdays tend to have high occupancy at many hotels. I envision a day when sales efforts are evaluated not on booked RN or booked revenue, but rather on revenue net of displacement. The value of a booking to a hotel, in terms of revenue, is the revenue booked minus the opportunity cost of that booking (meaning, revenue net of displacement). Displacement should be a part of every calculation of revenue impact, and one day it will be. The reason that dilution (tradedown) and displacement matter so much, is that both need to be taken into account in any demand generation effort. For example, a promotion that generates demand for New York City in December, or Paris in June, will tend to have high displacement (pandemic aside for the moment), because the demand is already so high, and the opportunity cost of a booking in this case is that it tends to displace another booking. And demand generation efforts that offer discounted rates, as many promotions do, are always at risk of tradedown. Such promotions can be money losers even in low demand times (meaning that there can be significant tradedown even when there is no displacement). More on tradedown can be found in Chapter 6. In future chapters, we will also cover the impacts of loyalty and distribution on revenue management, which are critical and becoming even more so. We’ll also discuss various analytics approaches and performance metrics to assess how we’re doing. So, there we have the fundamentals. If you have a good grasp of forecasting, pricing, inventory management, dilution and displacement, then you have a good grasp of tactical revenue management. The chapters that follow will take a deeper look into these, and other related topics, and will describe how they fit together in the evolution to revenue strategy.  CHAPTER 3 Forecasting What to Measure, How to Interpret, and Why It Matters In this chapter, we’ll cover some key concepts of forecasting, including why we are not better at it. I’ll discuss different measures of forecast error and which are appropriate under what circumstances, and even when forecast error doesn’t matter (yes, I just put that in writing). I’ll briefly discuss forecasting for meetings and events, and then close this chapter with a note on why forecasting is so important to the evolution to revenue strategy. As noted in the introduction, the tactical decisions of pricing and inventory should be based in part on a forecast of demand and a forecast of supply. A strong demand forecast will (should) result in upward pressure on pricing, because with excess demand, I can tolerate a lower price conversion (meaning, a higher percentage of customers can say no to that price, and I can still have a high occupancy). Similarly, it is the forecast of supply and demand that drives inventory decisions, including the extent to which we overbook a given hotel. Virtually, every hotel overbooks to some extent. The reason for this is that, across the industry, perhaps 30 percent of booked reservations end up canceling or not showing up (this figure spiked in the early phases on the COVID-19 pandemic). Without overbooking, we’d end up with a lot of empty rooms; these decisions are based on forecasts. As an aside, a word of caution on overbooking: we must take into account the cost of walking a guest (meaning rebooking them in another hotel). In an oversold situation, walking a guest to the same brand of hotel a few blocks away, and probably making the stay complimentary and offering some extra points or credits, may not be too costly in terms of customer impact. But if the whole market is sold out, a walk becomes significantly more problematic. The most extreme case I’ve worked through was President Obama’s inauguration in 2009. Hotels in and around Washington DC were sold out far in advance. We noted at the time that if we had to walk someone, we’d be walking them to New York City or to Raleigh, NC. As a result, there was effectively no overbooking! We did, however, require prepayment. If a RMS is in use, then a significant part of a revenue manager’s role is to review supply and demand forecasts, and adjust when needed. This is particularly important around special events, when the system-generated forecast based on history may be less relevant. For an example of when a system-generated forecast based on history is less relevant, we needn’t look any further than the impact of the COVID-19 pandemic. Transient booking windows were dramatically shortened, and demand was heavily impacted by travel bans, quarantines, and general health news for a region, each of which could change very quickly. Reviewing, and overriding as needed, a system forecast is also critical for hotels with significant group business. A revenue manager needs to determine how many rooms will actually be occupied for groups that are booked (e.g., it is not atypical for a given group to end up using only 85 percent of the rooms they have blocked). The revenue manager needs to work with their counterparts in sales and event management to determine how much of the room block can be re-released for sale. To make this problem much more complex, these supply and demand forecasts need to be done, and adjusted, by room type, and often by other attributes such as view type or other features. As a notable aside, one fairly common statement after the outbreak of COVID-19 has been “historical data is useless, and we cannot use it to forecast,” or some permutation of that. This is a logical absurdity. The only data that exists is historical data (as noted earlier, there is no such thing as future data). To ignore historical data is to ignore all data. The sentiment of the statement contains some truth however; the immediate future is not likely to look like the past during times of inflection. That said, any projection of any sort, provided it is not merely a wild guess, is based on some data or information. Even in the middle of the pandemic, any credible forecast of the future is based on historical data, albeit with some significant permutations to account for changes in booking windows, segments, source markets, or any other impact on customer behavior. As Kelly McGuire, Managing Principal at ZS Associates, says, forecasting during a pandemic is like “opening a new property in an unknown market.” One question that continues to trouble me, as well as several of the people I interviewed for this book: Why aren’t we, as an industry, better at forecasting? What’s holding us back? Part of the answer is, I believe, that forecasting does not get enough focus from the discipline. Any decent revenue manager can recite off the top of his/her head the RevPAR, RPI, year-over-year changes, and the corresponding figures for average daily rate (ADR) and occupancy as well. They can likely recite the mix of business by segment, perhaps even by channel and source market. They probably know the general stay patterns for their hotel by season. And much more. But if you ask a revenue manager, “what is your forecast error?” you may get a blank stare, partly because the question is not clear. I dream of a future where any revenue manager could immediately reply with something like “well, it depends on what you mean. For my hotel, the absolute transient demand forecast error from the system, weighted by booking activity, is 5 percent, and the system bias has been –1 percent recently, though both metrics have improved this year. The user forecast, however...” I remember being a guest speaker at a well-respected university (name withheld on purpose). I was talking about a particular demand forecast model in place at Marriott. One student, presumably eager to impress his professor, raised his hand and asked, “what is your forecast error?” I replied, “it’s 8 percent.” He said “wow, that’s high,” to which I responded, “but you don’t know what I’m talking about.” I think I said it a bit more politely. My point was that there are many different measures of forecast error. If I was referring to an aggregated demand forecast for the next month, 8 percent is dreadful; if I was referring to a daily arrival forecast by rate and length of stay, then 8 percent may be pretty good. In addition to the level of detail of a forecast, timing matters too. How is the forecast error weighted? It should be weighted by impact, meaning I care most during timeframes when I expect a lot of demand, and therefore, the pricing and inventory decisions that I make based on my forecast have the greatest impact. For example, I don’t particularly care about my forecast error a month from arrival at an airport hotel (some exceptions of course), because any pricing and inventory mistakes will have little impact, given that there is relatively little demand a month before arrival.  Another consideration is absolute versus bias; both have important meaning. For example, my forecast may be too high on some days and too low on other days, but on average, for the month, it is spot-on. The bias here would be zero, meaning that there is no systemic over- or under- forecasting. However, if we count the magnitudes of both positive and negative errors (hence the mathematical term absolute), we get a sense of the true accuracy. Let’s consider an example of a forecast that is fairly inaccurate, but is unbiased. Figure 3.1 Forecast Error: Absolute vs. Bias In Figure 3.1, the black line represents the actual demand by day, and the gray line represents the forecast for that day. For purposes of simplicity, assume each forecast was generated exactly x days before the arrival date. The dots, which line up with the secondary vertical axis on the right, represent the percent forecast error for each day. In this example, the average absolute value of the dots is 36 percent (average absolute percent error). But if we compare the sum of the daily values of the black line [1,758] to the corresponding figure for the gray line [1,754], we get an error of less than 1 percent; this represents the bias. It would be highly misleading to claim that the monthly demand forecast error is less than 1 percent. This is precisely why the distinction between absolute and bias in error measurements is critical. For hotels that have some version of an RMS, we also need to consider the system error versus the user error. How accurate is the forecast from the system, and how accurate is forecast once the user has influenced it? Note that some systems do not let the user influence the demand forecast (which I think is a flaw), but even in these cases, the system versus user error concept still applies to a roomnight or occupancy forecast (sometimes called a constrained demand forecast). In addition to increased focus from practitioners, and other stakeholders, there is opportunity to improve the actual science of forecasting. Evolving this discipline of revenue management will require using new and creative inputs to demand forecast modeling. The options are many. For example, if bookings at my hotel in Miami have unexpectedly spiked up, it may be that there was a corresponding spike in search activity a few days earlier, for my hotel or for the market. That search activity could have been on my own website, or on TripAdvisor, Google, or some other platform. If such a relationship can be established, it can be used to enhance our demand forecast—in this example, by using search activity as an input. Note that there are companies that provide this search activity tracking today; the challenge is how to incorporate this into an RMS to improve forecast accuracy. Search activity is just one example. Inputs to demand forecasts could include flight/travel activity, gas prices, consumer sentiment, changes in local demand generators, demand for nearby hotels, and much more. As an aside, the notion of using nearby hotels is not new; I refer the reader to a paper called Attribute Smoothing—A Pattern Forecasting Technique, on this exact topic, written (by me ) and published in 1998. Another input that has yet to be used explicitly is supply of alternative lodging (alternative in italic because it is hardly alternative anymore). This flexible supply is also referred to as short-term rentals or housing rental services, among other names. How might this input help with demand forecasting? Today’s RMSs use a forecast approach called a time series, in which historical demand is projected forward, with a variety of often complex permutations. An increase in alternative lodging supply will have a downward effect on demand for hotel stays. In this way, the effect of alternative lodging is seen implicitly in a hotel’s demand forecast, meaning that slower booking activity will result in a reduced forecast. However, what if we could account for this explicitly? If an RMS, or an individual user, could directly forecast supply of alternative lodging, this could improve our forecast even before the effect was felt in terms of booking activity.  In addition to expanding the set of inputs, the science of forecasting is now benefitting from machine learning, details of which are in Chapter 13. While there has been some impressive work on the science of forecasting, much opportunity remains. One often overlooked element of forecasting is an assumption inherent to most demand forecasting models out there: the independence of segments. What does this mean? Pricing and inventory decisions are based on a segment-level forecast. But that segment-level forecast is, in turn, influenced by your pricing and inventory decisions. This seems circular. For now, let’s just focus on the inventory piece; the pricing part can be accounted for with a price-sensitive forecast, meaning that the demand will be in part a function of the price charged. The inventory piece can get tricky. For example, based on my forecast, I will take 20 rooms at the U.S. Government per diem rate, and then close it (make that rate unavailable). Once I close it, the remaining government per diem demand may book some other higher rate, perhaps even the retail rate. In this case, my forecast for other segments should change. The degree to which these other segment forecasts should change is directly related to the incrementality of the segment we just closed (please see the section on dilution/tradedown in Chapter 2, and the more detailed discussion of incrementality in Chapter 6). It turns out that the government per diem segment is largely incremental, meaning that if that rate is not available at my hotel, those customers are very likely to look at a different hotel as opposed to book a higher rate at my hotel. The reason for this is that the customer will only be reimbursed for the per diem rate, and anything above that will be at their own expense. One caveat to this is the portion of customers who book the government per diem rate but who are not qualified for it (so called cheaters; please refer to Chapter 7 for more details on this concept). But overall, this government segment is highly incremental. In the context of forecasting, this means that closing this segment will not have a significant effect on demand at other segments. The same is less true for other discounted rates; e.g., closing an advance-purchase rate is likely to increase, perhaps significantly, the observed demand for your retail segment (customers paying the retail rate). Finally, it may not even make sense to track overall error, even at a segment level, weighted by booking activity, considering absolute error and bias. Why? Because these only matter some of the time. For example, let’s assume I am forecasting demand to be 40 percent of capacity for some future date, and when the date arrives, the figure is actually 50 percent. That is a dreadful forecast error. But it likely doesn’t matter because it doesn’t impact the tactics of pricing and inventory. Situations similar to this example tend to occur around many holidays (high demand dates such as New Year’s Eve in New York City, or Christmas week at Disney World, being notable exceptions). All of the preceding discussion about demand forecasting is focused on transient demand. For us as a discipline to continue to evolve toward revenue strategy, we also need to make meaningful progress on group and catering demand forecasting. The opportunity and the challenge are both large. Some companies have made real progress in this space, but no one has nailed it. Partly this is because it is really hard. As noted in the previous section, we need to be concerned with demand as well as net demand (meaning inclusive of cancelations). The group segment adds another layer of complexity; unlike an individual transient booking, a group can partly cancel—this is known as attrition. As an example, consider a group booking for 100 rooms. On the day of the event, only 90 rooms are actually needed. The 10 lost rooms represent attrition. In some industry circles, this is also known as group wash. While we’re on terminology, slippage refers to our estimate of attrition. In the preceding example, if the revenue manager believes that only 90 rooms will be needed, he/she may slip the group from 100 to 90, which impacts the number of rooms then available for sale to other customers. The challenge of group forecasting, then, is one of a demand forecast, a cancel forecast, and an attrition forecast, along with the timing of each (e.g., it matters quite a bit how far in advance a group will cancel, and how far in advance I can predict that happening). Much work remains in this space. Forecasting for meetings and events is a complex problem, and for this reason, it may be well suited to a machine learning approach (please see Chapter 13). There is likely to be some seemingly convoluted combination of factors that can provide some predictive value here. Such combinations of factors don’t even need to make intuitive sense, they just need to help our forecast. In the context of forecast accuracy, we are in the business of prediction, not in the business of explanation. More details on THRM, which includes meetings and events, can be found in Chapter 10. As a final word on forecasting, I’d like to share one of my favorite quotes; this one from Ian Wilson, a former GE executive: “No amount of sophistication is going to allay the fact that all of your knowledge is about the past and all of your decisions are about the future.” To build on that, the connection between one’s knowledge and one’s decisions is forecasting! And the evolution to revenue strategy is dependent, in large part, on forecasting. We need to get better at it, and we need to spend less time doing it. Strategy work requires time and effort and applied brainpower. Reducing the time spent on tactical decisions, such as pricing and inventory, is predicated on automating many of these decisions. Such automation is of course dependent on an accurate, and trusted, forecast. Note that the generation of the forecast can be automated, but, at least for the foreseeable future the communication of that forecast requires a person. To put a fine point on it, an important role of a revenue manager is, and will continue to be, communicating a forecast to, and sharing forecast scenarios with, all key stakeholders.  CHAPTER 4 Inventory Management Restricting What Is Offered for Sale, Why and How In this chapter, I’ll describe how inventory management works and provide a visual example. I’ll then connect inventory management to channel management. I’ll issue a caveat about the concept of optimal mix, and then close the chapter with a note on how this all fits into the evolution to revenue strategy. Once you have your demand and supply forecasts, and you have set your price points across segments and room types, and maybe channels, you now must determine how much of each part of your inventory to sell, and when. Specifically, you restrict certain sales at certain times under certain conditions—this is why inventory controls are sometimes called restrictions. To get grounded, let’s consider a very simple example. My hotel has two rooms available for sale for each of seven days in a given week. There are 17 requests for that inventory, meaning that the arrival demand for that week is 17. Each of those 17 has an associated price point and stay pattern (meaning, which nights they want to stay). For simplicity, let’s assume there are only two price points, $200 and $125, and that all stay patterns are either one or two nights. A visual representation of this type of demand pattern is shown in Figure 4.1. We see that Inquiry 1 (Record# 1) is for a one-night stay on Monday, while Inquiry 4 is for a two-night stay beginning on Tuesday. In Figure 4.1, we see that Inquiries 1 through 7 have a price point of $200, and the remaining inquiries have a price point of $125. Note that this sequence is for viewing simplicity, as opposed to the order in which the hotel sees the demand. This hotel cannot accept all demand in this case; the reader can verify this by adding up each of the columns for SUN-SAT, and seeing that these days sum to a number greater than two, meaning that if we accepted all of the demand, we’d have more customers than available rooms. The inventory management problem, then, is to determine which inquiries to accept and which to turn away, with the added complexity that you don’t see all of the demand at the same time, and you may need to turn away actual demand based on forecasted demand that may or may not materialize. Specifically, the problem is to maximize total revenue for the week, subject to the capacity constraints (in this example, two rooms available on each night). Note: for this highly simplified example, I’m ignoring any associated variable costs like housekeeping and commissions, and therefore aiming to maximize revenues as opposed to profits. I’m also simplifying the problem by not considering the weeks before and after our selected week; in reality, there will be stay-throughs from the prior week, and also from our selected week into the following week. The preceding example is just intended to give the reader a flavor of an inventory management problem. Figure 4.1 Inventory management—Demand pattern example So, what restrictions help us achieve our objective? I won’t go through every step here, but let’s get started. I’ll say yes to Record #1, but no to Record #2 (the order of these two is not important here, but I want to restrict one of them). The reason I only want to accept one of these two inquiries, Record #1 or Record #2, is that I’d really like to accept Record #9 (or Record #10, but not both). Record #9 will bring a Sunday night as well, which I can accommodate; I’d rather have two nights at $125 than one night at $200 (again, for this example, I’m ignoring variable costs). To get to the specific restrictions to accomplish this goal, I’ll need to think ahead. Once I book Record #1, I could immediately implement a length of stay restriction for Monday night; specifically, I will only allow a Monday night stay if it also includes a Sunday night. This restriction would deny availability to Record #2 but provide availability to Record #9. To continue through this inventory problem, I’ll accept Records #3 to #8 as well. Records #10 to #12 will be turned away by simple capacity constraints. I’ll accept Records #13 to #14, and then turn away the rest. This will result in total revenue for the week of $2,275. Note that there are a few different combinations that will get the hotel to $2,275, but none that will get higher (I encourage the reader to verify this ). It is easy to imagine how complex this inventory management problem can become, with many more rooms to sell, frequent forecast changes, forecast errors, varying lengths of stay, more price points, different patterns of stay-throughs, multiple room types, loyalty constraints, cancelations, early/late checkouts, and more. The actual assignment of guests to rooms in a hotel is currently a very manual process. Jason Bryant, Cofounder of Nor1, envisions a future where we will have “real-time inventory management, where a booking immediately triggers a room assignment.” He notes that humans can’t do this, but computers can, provided that there is a clear objective function and corresponding constraints. Some hotels also use price as an inventory control. If a hotel uses day- based pricing, where the price on a given day is independent of your stay pattern, then another way of restricting Saturday one-night stays is to raise the price for Saturday night. A word of caution on inventory management: beware the notion of optimal mix, at least in a broad sense. It is often misunderstood. This concept can apply to segments and channels. Most people who work in and around hotels have heard some form of this concept, but I believe it is often (though not always) misguided. Let’s consider an example: you may hear, “our optimal group mix is 40 percent.” Taken literally, this means that 40 percent group mix is better than 39 percent, and it is also better than 41 percent. This is the nature of optimal. Despite the fact that this use of optimal mix is fairly common, it is absurd. If more group volume drives hotel performance, then we want more group, and if less group volume will allow for other more profitable business to be booked, then we want less group. This determination is (or should be) completely dependent on the level and the price response of demand for each day across each segment! Please refer to my comments on lazy wording in Chapter 1, particularly regarding the use of the word optimal. This is more than just semantics. The more an optimal figure, for group, negotiated account, contract, and so on, is put out there, the more likely it is to become a target, leading to perverse incentives (e.g., what happens if my hotel is exceeding its optimal group mix, and I have an opportunity to book a large profitable group?). The expression optimal channel mix has similar limitations, and yet, the concept is surprisingly common in the industry. One way to think about it: the notion of optimal can apply to specific dates but becomes much less meaningful across dates. With some notable exceptions, including Kalibri Labs, many in the industry still misuse the concept of optimal mix. OK, off the soapbox. To close this chapter, I’d like to note that the evolution to revenue strategy is dependent in part on getting better at tactics. More precisely, this means getting better at tactics with far less effort (meaning more automation). Inventory management can be both tactical and strategic; as noted in Chapters 1 and 2, the former refers to maximizing profits, the latter refers to maximizing benefit.  CHAPTER 5 Pricing Principles, Science, Strategy, and the Work in Front of Us In this chapter, I’ll give an overview of pricing across multiple segments, describe what is meant by price optimization, and provide some graphics to explain these concepts. I’ll discuss how an RMS makes pricing recommendations, as well as the extent to which those recommendations are followed. Then we’ll see how this modeling capability has expanded beyond the retail segment, as well as how much upside remains, and the work in front of us as we evolve to revenue strategy. We’ll end this chapter with a discussion of how demand-based pricing is so often misunderstood. Recall from Chapter 2 that one of the fundamentals of revenue management is pricing, and that this means putting price tags on your inventory. In practice, this means determining how much to charge for retail business, negotiated corporate business, weekend breakfast packages, wholesalers, advance-purchase, promotions, and many other rate categories. These decisions are made with some combination of art and science. If a revenue manager has access to an RMS that recommends price points, using price optimization, then he/she will use that as a starting point. Many hotels around the world, in fact the great majority (over 80 percent, based on some estimates from industry experts), don’t have a system that recommends pricing. For those hotels, some combination of analytics and judgment is used to set pricing for every segment of business. As an aside, this represents an enormous opportunity for the discipline and the industry. A vendor that can figure out how to develop and sell a low-cost, easy-to- implement, simple-to-use pricing tool will meaningfully advance this discipline and industry, and also stands to make a great deal of money. Much progress has been made in this space recently. And such a system could be extremely beneficial to more complex hotels as well, even those that currently use a sophisticated RMS. As Bob Cross, Chairman of Revenue Analytics, notes, many RMSs of the past were designed and built with a user base in mind that had both the inclination and the time to do their own supporting analysis. Even for hotels that do have access to a sophisticated RMS, some judgement is still required. For example, an RMS may recommend retail price points for a given hotel, but it is the role of a revenue manager to validate these recommendations and to determine under what conditions to override these recommendations. Furthermore, even the most sophisticated revenue management software will give price recommendations for only some segments of business, in many cases, only the retail segment; for all of the other segments, pricing decisions are made based on analytics and judgment. While there is a great deal of analytics in place in support of pricing decisions across multiple segments, I believe that the future of pricing lies in price modeling and price optimization, for every segment of business. The RMS of the future will be much more sophisticated (more comprehensive models, more automation, more self- learning), and yet easier to use (more alerts and notifications, more health checks, less intervention). To get grounded on terminology, I’m using the term price modeling to mean modeling the consumer response to a given price at a given time for a given hotel (think of price elasticity modeling—e.g., at $159, I’ll get three bookings, but at $129, I’ll get seven bookings). I’ll use the term price optimization to refer to the recommendation engine that lives in an RMS; this engine uses price modeling along with demand forecasts and supply forecasts to actually recommend the price to be charged. Price modeling capability has exploded in recent years, and for good reason. Many companies, certainly including Marriott, but also including other chains and several vendors too, have made great strides in price modeling. The concept is simple: if I charge $x, how many bookings will I get? The answer to that is based on modeling. Specifically, this means the relationship between a given hotel’s price and booking activity, accounting for other factors such as overall demand level and competitors’ pricing. Most of the focus thus far has been on retail rates, which may be the most straightforward. Let’s look at an example: Figure 5.1 shows a simplified view of a price response model (again with transposed axis orientation, as described in Chapter 2). For a given retail price, the model will predict the level of demand. Figure 5.1 Demand as a function of price A sound model will take into account any significant predictor of demand, the two most important being overall demand levels (based on season, day of week, recent booking trends) and competitive pricing (like it or not, demand for your hotel can be significantly impacted by the pricing of your direct competitors). This is depicted in Figure 5.2. Figure 5.2 Demand as a function of price (input change) The solid line is the same as the solid line in Figure 5.1. If there is some other impact to demand, for example, if a direct competitor significantly reduces its price, then the demand facing your hotel may look like the dashed line. Anytime that any meaningful input changes, the graph will change. Any model will need to be tested before it is implemented, of course. While it doesn’t need to be perfect, it does need to provide predictive value, and this should be constantly validated. As the industry evolves, current models will lose some predictive power. For example, as more hotels offer some form of member rates, or channel pricing, the retail pricing models in place today will need to be updated. Similarly, cancelation/change policies and other fees will certainly impact a customer’s response to a given price, and these too will need to be explicitly modeled, meaning taken as inputs to a pricing model. There is a fairly common misperception about price response modeling and competitor rates; it goes like this: “why would I base my pricing on what my competitors are doing, when I don’t think my competitors really know what they’re doing?” At first, this may elicit an aha. However, that aha is misguided. Let me put this clearly: the prices offered by your direct competitors influence the customer response to your own pricing, regardless of how those competitors’ prices were derived. In fact, the very definition of a competitor means that a change in their pricing will have an impact on the demand for your hotel, independent of whether the competitive prices were the result of sophisticated price optimization or wild guesses. To state the obvious, such a price response model by itself doesn’t bring any value. Revenue management software doesn’t just model reality, it makes recommendations based on that model. What price should I be charging right now for a given date in the future? We want to consider the price response model for sure, but also underlying market demand, as well as available supply. Price optimization software does exactly this. Any time any of the inputs change, including how many rooms I have left to sell, the optimal price may change, and the price optimization software will reflect that. Adoption As noted in Chapter 1, adoption of system recommendations is an ongoing concern; across the industry, price recommendations are over-ridden perhaps one-third of the time. Why? The simple answer is that the pricing recommendations are not trusted. I believe that the crux of this is education and training, certainly for revenue managers, but also for key stakeholders. It is not uncommon for owners, general managers, and other important stakeholders to, knowingly or not, put pressure on revenue managers to make certain pricing decisions. Note: while in this chapter I’m referring specifically to overrides of pricing recommendations; the same concepts are quite applicable to overrides of forecasts or inventory controls. Klaus Kohlmayr, Chief Evangelist and Head of Strategy at IDeaS, notes that “it is human nature to be skeptical of systems,” but systems “can help take emotions out of decision making,” which is key to revenue management. He believes that anyone involved in revenue management should have a grounding, formal or not, in micro-economics, in order to have the context to challenge system recommendations. That said, Klaus also warns of the “risk of over-dependence on automation,” meaning that revenue managers will need to override the system on occasion, and must do so with confidence. He uses an analogy of airline pilots to emphasize this point: “pilots don’t need to intervene... until they need to intervene.” There is a great deal of pressure on revenue managers, from numerous stakeholders (any current or past revenue managers reading this are now smirking at this understated truth). A focused effort on explaining how these pricing models work may alleviate some of this pressure. Revenue managers need to understand the software they are using, and the resulting recommendations. The system cannot be just a black box built by some technologists and statisticians. As an industry, we need to upgrade our training about the systems we use. And we need to make this training efficient; every revenue manager I’ve ever met (numbering well over 1,000) is extremely busy, and cannot devote hours per day over a matter of weeks to train on a system. With this time constraint in mind, customized training is in order. In my opinion, despite recent progress across the industry, this is an enormous need for this discipline. Every revenue manager should be able to explain how their system works, and understand under what conditions it should be overridden. And every system should have a supporting dashboard or reporting capability that demonstrates the frequency and impact of any overrides. I’m certainly not suggesting blindly following every recommendation coming from the price recommendation engine. But I am suggesting that whenever a revenue manager overrides a price recommendation, they have a reason for doing so, and that reason needs to be more than “it doesn’t feel right.” The reason should include some knowledge that the revenue manager has that the computer does not have. To state the obvious, this presupposes that the revenue manager knows what the system does and does not know. And the stakeholders (owners, GMs, and others) would be wise to ask their revenue managers about the frequency, direction, magnitude, and, of course, rationale for overrides. At the same time, improving the models themselves will also lead to more trust, and therefore more adoption. This modeling effort cannot just be making existing models more accurate. In the broadest sense, a model is just a representation of reality. Reality in our industry is changing rapidly, due to changes in consumer behavior, growth in disruptors, changes in distribution strategy, and many other factors, and our modeling needs to reflect that. Clearly, we have much work in front of us. Dax Cross, CEO of Revenue Analytics, noted “the first waves of Waze also had lots of skeptics.” Dax’s suggestion for users: “Flip the script. Instead of focusing on evaluation of outputs, focus on the quality of inputs. Focus on situations for which there isn’t good data.” For a hotel revenue manager, this suggests a higher incidence of overrides at inflection points in supply or demand. Such situations could include hotel renovations, new hotel supply in the market, new nontraditional supply, new rate types/offerings, or even inflection points in the economic cycle. Dax’s point is more relevant now than ever; the COVID-19 pandemic presented an inflection, the magnitude of which was unknowable at the time. Overrides should have spiked up in the early phases of the pandemic; once the recovery becomes more stable and predictable, overrides should steadily decline. Tim Wiersma, founder of Revenue Generation, emphasized the need for safeguards, noting that “a system needs safeguards just as a human needs safeguards, and human intervention is critical to system safeguards.” So, what should the adoption percentage be? Overriding one-third of the recommendations seems too high, but the override percentage certainly should not be zero. The answer is, at the risk of sounding like an academic: it depends. It depends largely on the demand forecast accuracy of the system in place, but also on the accuracy of any price response models (including the quality of any shopped rates), both of which are hotel and date specific. For illustrative purposes, a suggested goal for override percentage could be taken from Figure 5.3. Figure 5.3 Override percent as a function of forecast error The purpose of this graph is to demonstrate visually that hotels with lower demand forecast error should be the hotels that have a lower override percentage. The variation (the gray shaded area in Figure 5.3) is to account for other relevant factors, such as the quality, consistency, and frequency of rate shops. I believe that the creation of, and use of, a graph like this can lend some much-needed discipline to override decisions, and I recommend that revenue managers and their stakeholders use such an approach as a sanity-check for system adoption. Bob Cross, Chairman of Revenue Analytics, points out that “overrides should increase during economic inflections, when the user may know more than the computer” (referring specifically to the ability to forecast). All of that said, in my view, an aggregated adoption rate of 90 percent should be a goal for the discipline overall, except for times of inflection. And in 2 to 3 years, when our modeling is better, that figure should be 95 percent. For some historical context, the following photo (Figure 5.4) is from a 2015 notebook (proprietary figures blacked out), when we at Marriott were keenly focused on system adoption. We were interested in the adoption of our system pricing recommendations, but also on the direction and magnitude of overrides. Some of these metrics have been shared with owners and operators over the years as well. In addition to driving system adoption, these stakeholder discussions proved extremely valuable for the discipline. Figure 5.4 Notes—Example of system overrides Beyond Retail So far, this chapter has been discussing modeling for retail rates. Price response modeling should and will also be expanded to other segments beyond retail. Some companies, including Marriott, have made significant progress here. For example, Marriott’s One Yield system provides pricing recommendations on premium rooms as well as for groups. I won’t share anything proprietary here (sorry), but I can share a few thoughts. The price response model for groups isn’t a demand as a function of price model like the retail model shown earlier in this chapter. Rather, it is a probability of a given group saying yes to my proposed price model, as shown in Figure 5.5. Figure 5.5 Group price response The reason for this type of model is that we can record historical yes and no for every group inquiry (lots of data quality caveats here), as a function of price offered; this is possible, but quite difficult, for retail demand, as explained in more detail in Appendix 3. You may see these types of models called win–loss models; by contrast, the retail model described earlier is sometimes called a win-only model. Marriott’s Group Pricing Optimizer (GPO) is described in the public domain. In fact, the team that designed and built it won the prestigious Edelman Award, given by the Institute for Operations Research and Management Sciences (INFORMS). If you Google Marriott Edelman, you’ll see more details about this. Please note that Marriott’s modeling work on premium rooms is not (yet) publicly available, so I will not go into any details here. The evolution from revenue management to revenue strategy means that we will have price modeling on all revenue streams, including other segments of business such as negotiated rates and advance-purchase rates, as well as on F&B and outlet revenue. Getting there will require a commitment to this goal, and a plan to develop these new pricing models, as well as the recognition that this is likely to take several years. While I cannot share any proprietary details about Marriott’s modeling work, I will share an important learning about price response modeling: the models are generally not transferable across segments, an example of which I’ll share in a moment. This is a really key point, and one that took me a while to fully appreciate. By this, I mean that, to the extent that buying and selling behavior is different across different segments, our modeling needs to reflect that. Without this context, the expansion of this modeling to other revenue streams will fail. And we need to appreciate that this modeling work is not primarily a math problem. The math part, while certainly not trivial, is solvable, especially with some trial and error, and smart people working on it. The harder part is taking a business problem and framing it as a math problem that can then be solved. This is really hard and requires a strong collaboration between revenue management practitioners and the model-builders, along with some creative problem- solving and stubborn persistence. As an example of these models being nontransferrable, let’s consider price modeling for negotiated account rates, a problem that, as of the writing of this book, no one has solved (for more context on negotiated account rates, please see the introduction to Chapter 7). A given hotel will want to know what price point to offer to a given account for the following calendar year; for simplicity of this example, let’s assume it is a fixed rate all year. How could we model that? Can we just apply existing retail price modeling to this segment? Most assuredly not. The buying behavior is very different than it is for the retail segment. With negotiated rates, the first step is to get the travel manager approval, meaning that your rate will be preferred. But then you need to get individual travelers to actually book it. Any model will need to take into account the opportunity for rebids, as well as any changes by competitors, such as mid-year repricing. A model will need to account for the impact of last room availability, and the possibility of offering it or not, and pricing that option appropriately, details of which can be found in Chapter 7. A model for negotiated rates must also take in account the projected stay pattern of the account. For example, a more appealing stay pattern may warrant a lower price to increase the likelihood of a hotel becoming preferred. Such a model must also account for so- called squatter rates; these are rates that are account specific but not preferred or even negotiated. All of this modeling effort will require assessing the predictive value of past performance; for example, do some accounts tend to exhibit different sensitivity to relative price? Finally, the long-term relationship with an account may warrant pricing that is not profit maximizing in the short term; this could be because the account is also an important vendor, or because the account’s business is critical during a downturn. The point here is that any expansion of price modeling is likely to require a significant effort and commitment, as well as a reasonable expectation of the timelines involved. In fact, it may be the case that the expansion of price modeling will not address negotiated rates in the near future. Revenue management leaders of the future will need to weigh the costs and benefits of price modeling for each segment of business, and prioritize efforts accordingly. Dax Cross, CEO of Revenue Analytics, points out that “negotiated rates may not be dynamic enough to even warrant a model,” so it may be wiser to focus our efforts on segments that are more amenable to a model, such as extended stay. Craig Eister, former IHG executive, suggests that near-term modeling efforts should focus on defining the market rate more precisely (most retail models were built before the advent of so-called member rates). Craig also highlights the need for modeling work for room type differentials, view type differentials, as well as other attributes of the room. Objectives The further development of price optimization will also require a clarification of objectives. As noted in Chapter 2, the objective of revenue management has evolved from revenue maximization to profit maximization to something approaching long-term value maximization. For example, member or channel benefits certainly do not maximize profits, at least in the short term. For example, why would we upgrade an elite loyalty member when we have the chance to sell that room at a premium? Because we believe that this loyalty play will generate more value in the long term. The same can be said for member rates, meaning loyalty members booking on direct channels. These do not maximize profit in the short term. To complicate matters, the loyalty benefit may not go to the hotel that is generating this benefit. For example, an upgrade at one hotel may increase a customer’s loyalty to the brand or portfolio, the benefit of which may well accrue to a different hotel. Knowing that, a given hotel may prefer to sell a premium room rather than upgrade a customer; the hotel clearly benefits from an upsell, while it may or may not benefit from an upgrade. Economists call this situation the free rider problem, and this is why chains and brands have upgrade policies in place, usually backed up by audits to drive compliance. Benchmarking and Pricing A word of caution on pricing: do not fall into the fallacy of over- interpreting benchmarking data as an explicit pricing diagnostic. This is frightfully common. In fact, I’ve even seen it in other published books and articles on revenue management. For example, STR data is extremely useful as a benchmarking tool. It is also useful for analysis of different performance drivers, for example, for use in A/B testing, for pricing, and other decisions. STR data (and reporting) represents an industry standard, trusted and used all over the world. The team at STR is top-notch, and I know several of them personally. The value they provide to the industry, and to academia, is monumental. In addition to benchmarking, STR data is useful for controlled experiments, or as a part of a performance analysis. However, this information is often not useful as a standalone pricing diagnostic, and yet, it is frequently used precisely that way. For example, if my hotel is gaining the ADR index, but losing the occupancy index and losing RPI, one could conclude that my hotel is overpriced, because it is growing rate versus the competitive set (often referred to as the compset), but more than offsetting this with a loss of occupancy versus the compset. Of course, it is possible that this hotel is overpriced. But it is also possible it is underpriced... perhaps it has a strong occupancy premium without much room to grow, and should be driving average rate even more. There are any number of explanations for the observed results, only one of which is overpricing. Other explanations include renovations, segment/mix changes, location of demand drivers, and distribution changes. Furthermore, some segments may be overpriced, while others are underpriced; the same holds for different days or timeframes, and these nuances matter a great deal. To quote Albert Einstein, “everything should be made as simple as possible, but no simpler.” Said another way, for the nerds reading this, the scientific principle Occam’s Razor applies broadly but not in this case. Competitive Response What about pricing based, in part, on expected competitive reaction/ response? I’m more inclined to raise prices if I have reason to believe my competition will do so too. This is a tricky topic, as there are some potential anti-trust concerns. That said, here goes... virtually every aspect of business (and of life) involves decisions. We each make decisions all day, on a wide variety of issues, both personal and professional. These decisions have one thing in common: they are made in the context of other people making decisions about the same issues, and these decisions are most certainly not independent. This is at the heart of a branch of economics/mathematics called the game theory. As it relates to pricing, it is quite reasonable to assume that decisions by one competitor may influence decisions by others. In fact, this is not hypothetical at all; there is an abundance of research that says exactly this. For reference, you can Google price war prisoner’s dilemma or something similar. The field of the game theory has plenty to say about price wars, and their severity and duration, based on a number of attributes. A detailed review of the game theory is beyond the scope of this book, but I strongly encourage the reader to become familiar with this field. To use foreign language analogy: if you are in a discipline that involves tactical and strategic competitive behavior, revenue management of course being a prime example, you do not need to be fluent in the game theory, but you do need to be conversational. As an interesting aside: it is fairly well understood that undercutting your competitors can trigger a downward spiral, and yet this is fairly common in a downturn. Why? Are we really collectively so stupid? I’ll address this specifically in Chapter 11. In addition to providing approaches to mitigate price wars, game theory can also be used to drive rates up. I’m going to refrain from putting many specifics on that point in this book; the obvious reason for that is anti-trust. In many parts of the world, we are subject to anti-trust constraints, and appropriately so. It is well beyond the scope of this book, not to mention beyond my own credentials, to offer advice related to anti-trust law. An abundance of caution, along with legal oversight and counsel, is strongly encouraged. Signaling future rates, with the intent to collude on prices, is almost always illegal. But the act of imitating competitor pricing, which is known in academic circles as conscious parallelism, is not. Conscious parallelism is generally legal in the United States, and the European Union, for example, provided that there is no explicit agreement among the competitors to do so. It is important to note whether conscious parallelism plays any role in decision making for your competitors (or perhaps yourself), and if it does, how you might use that to your advantage. Revenue managers with a strategic mindset will note that pricing decisions can be based in part on the degree of conscious parallelism in the market, and that this can be explicitly measured. To put a fine point on it: I’m more likely to raise my rates if I have reason to believe my competitors will follow, and I can assess that likelihood by evaluating their tendencies from the past. Important reminder: obtaining legal counsel is advisable in order to navigate the potential gray area between this and actual price collusion. Demand-Based Pricing One more point before we end this chapter: We should always strive for demand-based pricing, right? Of course yes, but... this may be one of the most misunderstood aspects of pricing, particularly retail pricing. The idea is simple: pricing should be a function of demand. That is certainly true. The confusion comes in when people interpret this to mean pricing should be solely a function of demand levels. No! Let’s consider a common example: Sunday nights at most hotels around the world. Let’s assume my hotel is charging $200 on Sunday, but occupancy is typically 50 percent. I believe that lower demand warrants lower pricing, so I lower my price to $100. What happens? My hotel’s occupancy grows, for example, from 50 percent to 55 percent, and I’ve just taken the lowest RevPAR night, and made it much worse! Retail pricing needs to be based on price response, and displacement. While it is true that Sunday nights have much lower displacement than other nights, this does not mean we should lower the price. Low season at a hotel does tend to have lower rates, but this is because of price response, driven by lower pricing in the market and in competing markets. The point here is that pricing should be based on demand (relative to supply of course), but this means the elasticity of that demand, in addition to merely the level. I hope that one takeaway from this chapter is that the evolution to revenue strategy is dependent on advancing our pricing capabilities. And I hope that a second takeaway is that there is enormous opportunity to do so.  CHAPTER 6 Discounted Rates The Most Misunderstood Aspect of Revenue Management In this chapter, I’ll cover some themes, and misconceptions, related to discounted rates. Specifically, we’ll review the definition of a successful discount rate, as well as criteria for that success. Throughout the chapter, we’ll use a hypothetical example of a college professor rate and discuss why that may or may not be a good idea, using the concept of incremental versus tradedown (dilution), introduced in Chapter 2, as a guide. We’ll look at price discrimination and apply that concept to advance-purchase rates and to last-minute rates. I’ll introduce a price discrimination approach that has yet to be implemented and reasons why it should be. And we’ll end the chapter with measurements of success, including some specific approaches to evaluating a discounted rate. As noted in Chapter 2, I believe this topic of discounted rates to be the most misunderstood aspect of revenue management. As the discipline evolves to revenue strategy, and is involved in, and driving, all revenue generation decisions, we need a holistic and unbiased perspective on discounted rates. The COVID-19 pandemic certainly put a spotlight on discounted rates; in Chapter 11, we’ll look at the competitive nature of discounted rates and how to avoid a discount war in a downturn. Let’s begin with a note on terminology: in this book, I’m using the term discounted rate to mean a rate that is below the prevailing retail rate. Some people use discounted rate to refer to a lowered retail rate, which I will not do in this book. A sound pricing structure for standard rooms will have a retail rate (for a given hotel/arrival date) and some discounted rates (see the two charts in the dilution and displacement section in Chapter 2 for why this is true). Of course, the retail rate for a given day will fluctuate for a variety of reasons, and there will be different retail rates for different room types. Let’s work through an illustrative example. Consider transient demand for a standard room for a given future date. Let’s assume the retail rate is $150. We will capture som