In today’s gaming world, there is vastly increased game flexibility and supply of gaming products in many, or possibly most, markets. These factors are changing the rules of yield on the gaming floor, necessitating a new framework for optimizing the yield of these games in this new environment.
The term “yield management” is “used in many service industries to describe techniques to allocate limited resources, such as airplane seats or hotel rooms, among a variety of customers, such as business or leisure travelers. By adjusting this allocation, a firm can optimize the total revenue or ”yield” on the investment in capacity. Since these techniques are used by firms with extremely perishable goods, or by firms with services that cannot be stored at all, these concepts and tools are often called perishable asset revenue management.”1
In the gaming machine space, for this article we define yield management as techniques to allocate gaming resources among a variety of patrons. For example, offering different times on a slot machine to different tiers of patrons. One measurement of this yield is revenue per square foot. Let’s explore revenue per square foot, then set a framework for optimizing the yield of the gaming floor and maximizing this revenue. (In later articles in the series we will dig deeper into examples of these techniques.)
In the modern casino, there are dozens of revenue streams. Casinos make money from hotel rooms, entertainment, retail shops and, of course, gambling. The question of how to optimize these multiple revenue streams traditionally in many instances has come down to the revenue-per-square-foot model.
The model is quite simple. Just figure out the daily revenue per square foot (RSF) of each revenue stream and make sure you allocate the space to ensure the highest yield.
At this point, you might declare, “If that were the case, then every casino would simply allocate all their space to the revenue stream with the highest RSF!” And you would be correct … if RSF were a constant. But it’s not.
Consider the slot machine. For the sake of mathematical simplicity, let’s assume that each slot machine takes up 1 square foot and has the ability to collect $100 per hour of play. Let’s also assume that there is no competition for the casino we are examining, and that there are plenty of local gamblers itching to play.
If the casino were to open with exactly one slot machine, it’s safe to assume that the machine would be played 100 percent of the time. The RSF for that machine would therefore be $2,400 ($100 per hour x 24 hours / 1 square foot). (See Table 1.) Two machines would also be played 100 percent of the time, and thus the RSF for two machines is still $2,400 ($2,400 x 2 machines / 2 square feet). In fact, this pattern will continue until we finally place enough slot machines that demand subsides to provide something less than 100 percent utilization. At this point, the RSF would then slowly decrease as we continue adding more games. For this example, let’s assume this occurs at 100 machines.
Now suppose we’ve continued to add games until we’ve reached the point of saturation, that is, the point at which adding games ceases to add any total revenue from all slot machines combined. For the sake of this example, let’s assume saturation occurs at 10,000 machines and that our RSF is $120 per day, meaning we are making $1.2 million per day from all machines combined ($100 per square foot x 1 square foot per machine x 10,000 machines). If we were to add another 10,000 machines, our overall win per day would not change due to already being at the point of saturation, but now our RSF plummets to $60 ($1,200,000 per day / 20,000 machines).
Table 1 – Revenue Per Square Foot CalculationsSo, instead of RSF being a constant for this particular revenue stream, it is in fact a complicated, piece-wise step function—it starts off being constant, then quickly begins a slow descent when demand is met until we reach saturation, at which point it finally plummets toward zero. While rarely calculated precisely, the fact that casinos are not simply 100 percent devoted to the top revenue stream shows that operators have (at least intuitively) known for years that RSF is a highly variable metric. To unlock understanding of this problem, we need to look deeper into yield management and defining gaming patrons.
Patrons vs. Customers
Patrons are dramatically different from customers. A customer is defined as “a person who buys,”3 with the assumption that there is a seller and a price. Patrons are different for at least three reasons:
1. Patrons choose their own price for their gambling experience. The gambler chooses how much to wager and so controls the price of their experience; they choose to spend pennies on a game or tens of thousands of dollars on that same game.
2. While the house advantage underpins the whole industry, it is always possible for patrons to win, resulting in a negative purchase price. In a retail situation, the customer always pays.
3. Some patrons are actually skilled enough that they gamble with the odds in their favor.
When looking at yield management in the hotel industry, we look at business customers vs. leisure customers. When looking at gaming patrons, we need to consider other factors, as there are very few business customers and typically we try to stop them from playing at all. When looking at gaming, we need to consider the constraints that drive a patron’s gaming experience.
These constraints include:
1. Time – Some patrons are limited in the total amount of time that they can spend gaming.
2. Wallet – Some patrons are limited in the total amount they can lose during their gaming experience.
3. Time of day – Some patrons can only play at specific times of the day.
4. Day of week – Some patrons can only visit on specific days of the week.
In yield management, there are a number of metrics that can and possibly should alter the way that we execute our plan. However, how these metrics are normalized can provide quite different views of the patron. Normalization is defined as the denominator in the calculation of the average or rate. Following are examples of different time-based normalizations:
1. Per Hour of Play (PH)
2. Per Day (PD)
3. Per Month (PM)
4. Per Trip (PT)
To illustrate the use of normalization, consider two patrons who have the same theoretical win per trip but dramatically different theoretical win per hour of play. While their trips have basically the same value to the business, their customer profiles are vastly different, as is their suitability for yield management.
Combining the patron constraints (as discussed above) with a ranking of the time normalized metrics generates a powerful breakdown of customers suitable for yield management. Table 2 shows a breakout of players by their yield management types. For illustrative purposes, it may help to imagine that all of these players have the same theoretical win per trip and the same theoretical win per month. They do, however, present very different yield management challenges. Consider the aggressive time constrained players. It is likely important that members of this segment find the gaming experience they are looking for in small amounts of time.
Table 2Yield Management Actions
• Closing whole sections of the gaming floor
• Changing secondary gaming effects
• Time-of-day marketing events
• Promotional offers, such as meals that are available only at a specific time of day
• Changing the density of gaming machines on the floor
• Changing the mix of gaming of machines
• Changing the available patron-selected options, such as theme or denomination
Beyond Gaming Machines
We’ve seen that the gaming yield management model is in practice much more complicated than it seems at first. But don’t worry, it’s actually much worse!
The various revenue streams do not exist in isolation. Instead, certain streams serve to support others. In heavily competitive environments, companies have even resorted to loss leaders—for example, a buffet that loses money but is a great marketing tool for driving incremental gaming revenues. At this point, the gaming yield optimization model falls apart completely, but do not be discouraged, analysis is still possible.
When confronted with a low- or negative-yielding revenue streams, one must answer the following questions:
1. Does this revenue stream prop up other revenue streams, like the buffet example above?
2. If so, how much incremental revenue does it provide?
3. Does the ROI on the low-performing revenue stream provide justification for the low- or negative-performing square footage?
Question 1 is really for operators, and usually the answer is a simple “yes” or “no.”
Question 2 requires some heavy analytics and is perhaps best left for a future article to explore. However, simple moving averages can often provide a gauge as to the effectiveness of one revenue stream on another.
Question 3 is quite interesting. To properly calculate the ROI when we allow square footage to be used for underperforming revenues, we cannot assume that if it’s not a loss-leader, there is no cost (or if it is a loss-leader, that the cost is simply whatever amount we lose per day). Instead, we need to measure the opportunity cost. If there is a better use for that space, then that has to be factored into our ROI calculation.
An example might make this clearer. Suppose that we have a 10,000-square-foot buffet that loses $1 per square foot per day. However, we’ve been able to measure its impact to the gaming floor and have discovered that it adds $15,000 per day in incremental gaming revenue. So the 10,000-square-foot buffet loses $10,000 per day but adds $15,000 per day in gaming revenue for a net gain of 50 cents per square foot per day. Sounds good, right? But what if we had another use for that space that generated $1 per square foot in revenue? Or if we could spend $10,000 more per day on marketing and thereby generate $20,000 per day in incremental gaming revenue? In either case, we may want to remove the buffet after all.
In this article, we have explored the optimization of various products on a casino’s footprint. We have looked at both non-gaming and gaming revenues, and have explored different analytical techniques for attacking this optimization problem. Never before has the gaming floor been so dynamic and never has the customer had so many choices from both the gaming industry and its non-traditional competitors.4 Moving to a yield managed gaming business gives a fascinating twist on the operation of the gaming experience, and like the results of yield management in other industries, may result in what seems to be counterintuitive actions. However, these counterintuitive actions could become the next competitive advantage.
2 Based on the assumption that each slot machine takes up 1 square foot.
4 P. Laube, M. de Berg, M. van Kreveld (2008). Headway in Spatial Data Handling (Eds. Anne Ruas, Christopher Gold), Lecture Notes in Geoinformation and Cartography Series, pp. 575–593.