“Big Data” Revenue Management

Back in October, Kelly rather succinctly stated that “Revenue management has always been a “big data” problem.” This is very true. This week, as we continue our exploration of big data in hospitality, I’m going to delve deeper into the data needs for revenue management. I’ll be exploring the important sources of information that should be driving revenue management decisions – what they are, and why they are important. My primary focus here will be data used to support automated revenue management decisions – but I’ll also touch on the type of data that revenue managers should be looking at regardless of whether they use an automated system or not.

The Basics

For any revenue manager looking to make data-driven decisions, the starting point is to make sure that you have the following types of information at your disposal:

• Stay history – data regarding the final number of rooms sold by date, and the revenue generated in the past
• Inventory history – data regarding the number of rooms available for sale by date in the past
• Future reservations – data regarding on-books reservations (this data should be captured and stored regularly – daily or weekly) for future arrival dates
• Future Inventory – data regarding the number of rooms available for sale for future arrival dates
• Future Rates – information regarding the rates that are available for sale in the future

If you are a revenue manager trying to do this job by hand, you probably have a spreadsheet that contains the stay history and future reservations data. You might also have some of the other data listed above in that spreadsheet – but my experience is that most revenue managers don’t. Why? Because that information is in their head! After all, every competent revenue manager knows the capacity of their property, as well as the rates that they are selling. Many revenue managers are surprised to learn that an RM system needs these other pieces of information – after all, they don’t have them stored – why would a system need them? I’ll cover some of this as I describe the uses of each of these pieces of information, starting with the data most commonly kept by revenue managers.

Stay History

This information is absolutely critical to understanding the overall demand patterns for a hotel. What are our high seasons and low seasons? What is our day of week pattern? Is the property full on weekdays or weekends? Does our average rate fluctuate according to this pattern, as well? The most important and basic information that we need to manage revenue is based off of this data.

Future Reservations

Future reservations information is critical to understanding how many rooms are left to sell. It’s pretty difficult to manage revenue on a date that’s already sold out – so this information tells us how much space is left to manage. In addition, as we keep track of this information, we’ll be able to discern a booking pattern for our property – how far in advance of arrival do guests book reservations – and determine if this booking pattern fluctuates by arrival day of week or by season?

Future Inventory and Inventory History

Revenue management is, at its heart, the act of balancing demand and capacity through management of pricing. As such, future inventory information is essential to every revenue management decision imaginable. In fact, capacity is essential to the understanding of information provided by both stay history and future reservations – without matching capacity information, those other pieces of data are of limited value. In particular, the process of unconstraining (i.e. estimating sales lost when rates and rooms were unavailable for sale) depends upon the availability of accurate historical capacity information over an extended period of time.

This is why I am so surprised at how frequently hoteliers fail to track this data: the assumption that capacity is stable over time in a hotel is frequently an incorrect one. Wings are upgraded, and rooms out of service for a long period. Everyday maintenance issues take individual rooms out of service for short periods of time. Room type configurations change – and so on. If you are a revenue manager that is currently managing revenue using spreadsheets, but thinking that you might look at a revenue management system in the future, do your future vendor a favor, and start collecting this data today. They’ll thank you later, I promise.

Future Rate Information

This, of course, is where it all comes together. After all, if managing rates and availability is the problem, then knowing what the rates are and how they are sold is pretty essential information, isn’t it? And most revenue managers do have a strong understanding of the rates that they sell, including important factors such as qualifications for rates, typical contract elements (for contracted rates at the property), distribution costs (for distributed rates), and so on. It is this broad understanding of rates that allows the typical revenue manager to handle what is obviously a very complex job.

Capturing this information outside of a Property Management System (PMS) or Central Reservation System (CRS) is often quite difficult, though – rate data is both tremendously complex and highly dynamic. In fact, revenue management systems frequently don’t really even try to keep up with all of the complexity and change. They do this by simply capturing historical average rate information and forecasting the rate value patterns forward. In most cases, and done properly, this compromise works fine.

But that’s not all…

Next week I’ll continue our investigation into “big data” and revenue management, with an exploration of the data that is critical to revenue management analytics, including the “next generation” needs of true rate optimization solutions.

“Big Data” Revenue Management (part 2) Published May 31, 2014 

Last week I opened this topic of “big data” revenue management by introducing the typical information sources that revenue managers rely upon when managing revenues manually. This week, I’ll be exploring revenue management data further, with a focus on data required to support revenue management analytics. I’ll finish off with a look at the type of data required to drive “next generation” rate optimization.

Taking the Next Step with Revenue Management Data

Last week I wrote about the very basic data sources that drive every revenue management function: stay history, inventory history, future reservations, future inventory, and future rates information. This data is essential – and generally, the more of it that you keep, the more valuable the data becomes. For example, by storing future reservations data over time, we can assess the booking profile of our guests – a critical piece of information for revenue management. By keeping the information longer, we can see if these profiles are seasonal, if they change over time, and so on. So – keep your data. All of the other information I’ll be talking about here are useful – but of limited value if the basic information from last week isn’t stored over time and readily available.

With that said, let’s move on to the information that will help take you to the next level in managing revenue for your properties:

• Holiday and event information – information regarding events and holidays that have substantial impacts on demand levels or patterns.
• Segmentation & length of stay – breakdowns of reservations and stay information indicating the customer group or rate type, and the length of the guests’ stay.
• Market / share information – information regarding reservations, stay, and rate levels for competitive properties in your market.
• Competitor price information – information regarding the rates being charged for future arrival dates by competitive properties in your market.
• Walks, upgrades, and cost of walk – data regarding guests that have been “walked” from your property (due to oversales or out of service rooms), and the costs associated with re-accommodating them, and data regarding guests upgraded to a more luxurious room type.
• Ancillary revenues – revenues associated with non-rooms sources, such as food and beverage, retail sales, meeting space and associated rentals, and so on.

Again, many revenue managers working without an analytic revenue management solution try hard to keep much of the data I’ve listed here – with varying degrees of success. The sheer amount of information and complexity of the data really can become overwhelming – and spreadsheets begin to reach their limits – when dealing with this many different types of data.

Holiday and Event Information

Holiday and events data is critical to revenue managers and analytics. Without this information, seasonal demand and mix patterns can be affected – causing systems to over- or under-estimate demand or inappropriately predict the mix of guests. In addition, many holidays or events affect day of week demand patterns, as well. For these reason, most revenue managers are very aware of the need to keep abreast of upcoming events.

Analytics have different needs with regards to events. Analytics require event history in order to predict the impact of upcoming events. This means that repeating events need to be recognized in the past, as well as the future. In addition, non-repeating events need some categorization, so that historical event information can be used to predict the impacts of future non-repeating events.

While event information is very important, one of the most frequent issues I have witnessed in hotel revenue management is over-specification of special events. Most forecasting analytics do not react well when a significant portion of the historical data is identified as impacted by events. For this reason, it is very important that revenue managers limit their labeling of special events to periods that have truly significant differences in behavior patterns, or use analytics that are capable of making this assessment automatically.

Segmentation and Length of Stay

Segmentation and length of stay represent levels of detail regarding stay and reservations information – critical information, when the guest base varies significantly in value by customer segment or multi-night stays are common. Most revenue managers recognize the importance of breaking down customers by at least the roughest segments: transient and group. Analytics, on the other hand, often requires much more detailed segmentation – especially in breaking transient segments into various qualified segments (such as corporate accounts), and non-qualified segments (such as promotional, or “best available”), and even breaking out booking source, when significant direct sales costs are implied. For group-heavy properties, similar segmentation of group reservations is necessary.

Length of stay information can contribute significantly to the return from revenue management analytics, often by improving the revenues on shoulder days. While most revenue managers have the ability to weigh the value of different customer segments, length of stay effects are much more complex to assess. Sometimes a short length of stay is preferable, while at other times, longer length of stays is preferable. Determining which strategy to apply to any given period can be difficult, as there’s a wide variety of factors that can affect that decision, including demand levels, unsold room capacity, expected mix of demand (both LOS and segment mix), expected cancellations and group wash, and variability of demand.

Market / Share Information

Many hotels and hotel brands make use of competitive market information. Traditionally, market information has come in the form of aggregated and averaged historical performance – occupancy, ADR, and RevPAR. More recently, forward-looking competitive reservations information has become available. This aggregated market information is extremely useful in gauging market trends and assessing competitive strategies, but can be of limited value in assessing specific revenue management decisions.

Competitor Price Information

There is a variety of ways to obtain competitive rate information, including through service offerings. I wrote at length in this post regarding the value of competitor rate information, so I won’t go into detail on that subject again here. There is a difference in the way that a typical revenue manager, or even traditional revenue management analytics uses this data – I’ll cover that difference later on in this post.

Walks, Upgrades, and Cost of Walk

Many hotel revenue managers have mixed feelings regarding overbooking, and this is understandable. Improper overbooking can be costly – not only in terms of guest service, but also to the bottom line. Hotel situations vary – so I don’t want to get into a defense of the practice of overbooking here. What I will say is this: if you are overbooking your hotel, and not paying attention to these figures, you are taking risks that you probably shouldn’t.

Ancillary Revenues

For some hotels, ancillary products and services represent a significant revenue source. As such, it is important to keep track of them and, where possible, manage them. Unfortunately, revenue managers often have difficulty obtaining sufficient data regarding ancillaries, because of the difficulty in accessing and consolidating the data in the systems used by those operations.

Price Optimization Analytics – What Is Different?

Back in August of last year, I wrote a 2-piece entry explaining the differences between traditional revenue management analytic approaches, and more sophisticated price optimization analytic methods. Price optimization methods essentially have the same data requirements as classic revenue management methods – and more. In particular, price optimization methods require these two additional sets of data:

• Historical rate and availability information
• Historical competitor price information

The primary distinction between classic revenue management approaches and price optimization analytics is the requirement in the price optimization approach to relate price to demand directly. Historical rate and availability information across the booking period (since rates for a given arrival date typically change, and averages like ADR introduce errors) is critical to calibrating this sort of demand model accurately.

As we have discussed at length here at the Analytic Hospitality Executive, the hospitality industry today is highly competitive – and our customers have an ever-increasing number ways to see competitive rates. So, competitive rate information is critical to understanding how customers will react to rate changes. I did, however, already mention competitive rate information in the section above relating to classic revenue management analytics – so why revisit this topic? The reason is that classic revenue management analytic approaches typically only use future competitive rate information –as a guide for decisions. Price optimization analytic approaches, as noted above, need to relate price to demand – and, given the highly competitive nature of the industry, this assessment is best made in the context of what the competition is charging. Therefore, price optimization analytics for hotels need historical competitive rate information in addition to future competitive rate information. Like the own-rate data discussed above, this information will be used to calibrate the sensitivity of customers to rate changes.

Revenue Management Really is Big Data

In this post and my earlier one, I have covered the different types of data that are useful to assist in making comprehensive and revenue-maximizing decisions. Combining the different types of data, the frequency of collection, the historical and prospective timeframe, and the changes that occur regularly and you’ve got the recipe for “big data.” How big? Well, typical revenue management input data includes:

• Customer or market type segments optimal for analytics (described in the segmentation and length of stay section): 60
• Different accommodation types: 12
• Historical dates (2 years’ of history): 730
• Future dates (1 year): 365
• Length of stay types: 8
• Snapshots stored for each occupancy date: 40

The combination of all of this input data for just one property is 252 million observations. Note that this only includes only a subset of the information that I’ve covered in these two posts. That said, if you then generate decisions based on this data and store those decisions, you will need to store approximately 10-20 gigabytes per property. For a hotel chain with 2,000-4,000 properties, that would equate to 20-80 terabytes of data.

I hope that you’ve found this discussion on “big data” in revenue management useful and informative. Are using information to manage revenue that wasn’t discussed here? Please share – we’d love to hear from you.