Casino Enterprise Management

This is the second article of the “Where’s the Money?” series that focuses on Jackpot Wharf, a mini casino area within Silverton Casino. This article will look into player displacement and clustering analysis of the player behavior. It will also introduce a new concept, preference filters, which are designed to highlight areas of preference for gaming product on the gaming floor. Preference filters are a critical tool for discovering what drives play from players who have a preference while filtering out players who spread their play widely across different gaming products.

In our last article, we covered how “war room” collaboration drove an ambitious effort to transform Jackpot Wharf from one of the weakest areas of Silverton Casino to one of the strongest. The overall goal was to have the Jackpot Wharf area offer a taste of the Las Vegas Strip experience, enticing the destination market customers who visit the property for its Bass Pro Shops store to stay and play. The changes to the area drove the revenue up by more than 20 percent, and now Jackpot Wharf is the strongest performing area of the gaming floor. Furthermore, the additional revenue has been accompanied by an increase in year-over-year revenue on the balance of the gaming floor.

As we dig deeper into the results of our analysis, our core goals are to enable learning from the successes we find. There is now an expectation that, given the success of Jackpot Wharf and the earlier success of Penny Alley, another mini casino at Silverton (see the December – March 2011 issues of CEM for the complete analysis), that further work with mini casinos will drive incremental revenue. However, learning from one’s successes is sometimes harder than learning from one’s failures.

According to Robert Sutton on his Harvard Business Review blog (June 4, 2007), “After people succeed at something, it is especially important to have them focus on what things went wrong. They learn more than if they just focus on success (so, don’t just gloat and congratulate yourself about what you did right; focus on what could go even better next time).”

Sutton is saying that if we can find the cause of any missteps on the road to our success, as well as identify the key drivers of the success, we will be able to replicate that success next time, and maybe even improve it.

Because we’re dealing with a mini casino, understanding player preference is an extremely important part of the analysis of our success.

Preference Restrictions
A preference restriction is a means of grouping players based on the manner in which they select their gaming product. Here are some examples of preference restrictions:

1. Select all players who played on Machine No. 123 in December.
2. Select all players who played on Machine No. 123 in December and who also had at least 50 percent of their total play appear on Machine No. 123 during that month.
3. Select all players who have at least one machine that captured 50 percent of their total play in December.

We can also request additional data from these restrictions in multiple ways. For example, for Preference Restriction 1 above, we can choose to look at those players’ overall play data in two very different ways:

- For all players who played on Machine No. 123 in December, select their play across all other machines during that month.
- For all players who played on Machine No. 123 in December, select their play across all other machines during that month, but only on days that they also played Machine No. 123.

In the above examples, we selected groups of players who passed our preference restriction during December, but we could have easily selected a specific week in December or even a single day. However, this gets much more complicated when we consider the fact that players behave differently on different gaming visits. As a simple example, what do we do with a player who only plays Machine No. 123 on Dec. 1, but who then returns on Dec. 15 and ignores Machine No. 123 entirely in favor of Machine No. 456? Do we conclude that this player is 50-50 about these two games, or was the player just tired of No. 123 and tried No. 456 on a lark? This is not an easy question to answer, but we need flexibility in our data to allow for the possibility to research these types of questions.

To do this, we need to look what we call the “player visit.” Instead of looking at the play of each player over the month of December, we go down a level deeper and look at the play of each player on each visit to the casino, and consider these separately. So our player’s visit on Dec. 1 is considered completely separately from his visit on Dec. 15. In our example above, when looking at Machine No. 123 and our player who visited on the 1st and 15th, we would only consider his player visit that occurred on Dec. 1, since that is the visit when the player played Machine No. 123. The play from his player visit on Dec. 15 would be completely ignored in the analysis. In practice, it is the experience of at least one of the authors that, when using preference restrictions, the information that is derived when analyzing player visits is far more impactful than that derived when just analyzing players.

Preference Filters
Before we dig deeper into discovery of the gaming floor revenue, let’s look an interesting algorithm called preference filters. To understand preference filters, think about a casino with two players and 10 slot machines. These players are categorized as Play Everywhere (PE) and Play Two (P2). PEs play on every game, and P2s play on only two games. Figure 1 shows a graph of the revenue per day for these two players. A look at the gaming floor shows that Location 4 is the strongest, followed by Locations 3, 5 and 6. The underlying question is, are there location preferences amongst these locations?

To answer this question of preference, let’s filter out all players where less than 16 percent of their play is at that location (see Figure 2). (Note: This 16 percent of play means that the ratings for players will be filtered to, at most, the top six locations and, in practice, only the top two or three.)

Now the results look quite different. We can see that Locations 3, 4, 5, 6 and 7 are driven by play from players who have a strong preference for these locations. This example illustrates the concept of preference filters, and in this case, the numbers are easily understandable. When applied to the larger context of real gaming floor numbers, it still serves to highlight locations that players prefer to play at—a powerful and simple concept. Preference filters become even more interesting when they are calculated against different aspects of the data.

The most important characteristic of a preference filter is that the calculation of the filter is executed at the session or ratings level. This low-level filter means that each view of the data is a new calculation of the results. The challenge with these preference calculations is that they truly exploit detailed data, and that data is often very large. This means that the methods of building aggregates or data cubes are not useful for preference filter calculations.
Now let’s examine some more sophisticated preference calculations.

Question 1
What is the First Choice in Game Theme?
The masses of players moving around all have very different game preferences, but in our experience, these groups of players often fall into distinct groups of players. (See “Market Basket Analysis,” CEM, December 2008). Some players are prone to play on many different themes, while others concentrate their play on only specific games. This combination of different play patterns creates a kind of noise, where many groups of players play across a wide range of gaming product.

By filtering the play to only players who have a preference for a theme, the concentration of play on the gaming floor now shows the magnet games or product that players will almost always play. For example, if we apply a theme-based preference filter to only show players who spend 50 percent of their time on one specific theme (regardless of where that theme is), then the resulting analysis is going to be highly concentrated on games that are magnet or first choice games themes. This preference filter can be further refined using subsets of players. For example, “Show me only local players versus destination players.”

In the case of Jackpot Wharf, the game theme preference filters showed a strong tendency for local players to not play in the Jackpot Wharf area. Furthermore, the preference filters showed that players who played in the non-Jackpot Wharf areas of the gaming floor did not move their play to the Jackpot Wharf area.

The theme preference filter was very powerful in understanding this pattern, as we were able to see that the local players remained with the same preferences and that they did not change to play on any of the new product located in the Jackpot Wharf mini casino.

Question 2
What is the Displacement of Players with Preference?
The alteration of games on the casino floor and the subsequent displacement of play patterns is one of the critical questions in gaming. (Note: We have intentionally obfuscated this example to protect Silverton’s confidential data.) Using the preference filter, we can calculate how many players prefer to play on a bank of slots that we are considering changing.

From the preference calculation comes a powerful selection process. Looking at Figure 3, we can select the players who preferred to play on Bank 04-007. This bank is an illustrative example of a bank that was changed and how the preference calculator can be applied to understand the displacement of game play. Clearly the players on Bank 04-007 are concentrated on these banks and are not playing on some of the surrounding games.

This displacement of gaming revenue is one of the core questions that allows for ongoing optimization of the gaming floor. In addition, the selection of players with a preference enables the marketing team to determine who stopped visiting the property. Combined with a change in gaming product, this change in visitation may well have resulted from the alteration of a product that those players preferred.

Figure 4 shows the same group of players who preferred the old games at Bank 04-007 and how they altered their play patterns after the game had changed. Notice that this group of players did play on the new 04-007 games, but, at least in this example, the players who preferred the old 04-007 game were displaced to other banks, including 04-064 and 04-061. This examples illustrates how a preference filter can be applied to select a group of players to further examine the displacement of their gaming revenue following a change in the gaming product.

Question 3
How Do Math Models Affect Preference?
Math models are one of the core features of a gaming machine, and many themes share the same paytable. The preference calculation based on the preference for a gaming model enables the analyst to address questions such as, “Are there gaming models that are driving magnet products?” Using the selection process, the players with a preference for a paytable can be selected and further analyzed to see if there are other models that they will play on or what kind of volatility of gaming model that they enjoy.

More Results from Jackpot Wharf
Jackpot Wharf truly achieved its goal at Silverton Casino. During November, the marketing department executed an aggressive hotel campaign that resulted in a 150 percent increase in gaming business from hotel guests. Furthermore, as desired, the Jackpot Wharf mini casino responded with a more than 250 percent increase in gaming revenue from this marketing program in the new targeted area.

The increase in non-rated play by over 33 percent in Jackpot Wharf has been a further reinforcement of the targeting of this area to the non-locals market (local players tend to be carded).

Given the volume of the changes, it is the belief of the authors that the redesign and refocus of the Jackpot Wharf mini casino was the central causal factor in these increases.

The bottom line is that Jackpot Wharf continues to have very little impact on locals play and has driven the play from non-locals and non-rated players.

Conclusion­
The calculation of game preference is one of the most powerful techniques in deciphering game play patterns. The case study at Silverton shows that the combination of the mini casino strategy and the preference calculations can, if executed correctly, drive significant revenue. Numbers like a 250 percent increase in revenue from a targeted group are hard to argue with and stand as further validation of the importance of insight-driven business initiatives. Operators now have a choice whether to adopt these powerful techniques that enable them to filter out the noise in the data and see true player preference and displacement or to continue to optimize the gaming floor based solely on revenue metrics.

Andrew Cardno has more than 16 years of experience in business analytics, ranging from modeling health care drive times to casino gaming floor analytics. He often presents on the future of analytics across the world and has spent the last seven years living in the United States and working with corporations around the world. He can be reached at andrewcardno[at]yahoo.com.

Dr. Ralph Thomas is Vice President of Strategic Analytics and Database Marketing for Seminole Gaming. During his years in the casino industry, Thomas has focused on maximizing profitability by applying statistical analysis to the company database. Previously, Thomas spent 15 years in academia, as both a student and a lecturer of mathematics. He can be reached at ralph.thomas[at]stofgaming.com.

Jada Evans is the founder of MindSight Analytics and works with gaming companies to provide analytics and help develop strategies for a variety of topics including game performance, marketing, food & beverage, hotel & labor. She can be reached at jevans[at]mindsightanalytics.com.

Salinda Conklin is a seasoned slot operations and performance professional who has worked in properties in Iowa, Illinois, California and Nevada. Conklin brings a wealth of knowledge and over 20 years of experience to Silverton Casino Hotel as the Vice President of Slots. She can be reached at salinda.conklin[at]silvertoncasino.com.