Authors' Note: In the ninth of a 12 article series themed on where the money is now for “smart” casinos, VizExplorer executives discuss the fallacy of one-to-one marketing, and how a one-to-one communication strategy, combined with third party data, can prove more valuable for the land-based gaming operation. Please note these articles are meant to stimulate thought and that we are using some deliberately provocative metaphors which should be taken with a grain of salt.

We have been writing about one-to-one marketing for years, and it remains our conviction that the theory is complete fiction when it comes to the gaming industry. One-to-one communication, however, holds a lot of marketing and other potential for the casino enterprise.

What follows is an examination about what is wrong with one-to-one marketing and what is right about one-to-one communications when it comes to the land-based casino business.

A one-to-one marketing strategy focuses on each individual customer receiving a personalized marketing offer. The strategy of one-to-one marketing is tailoring a campaign to a customer’s specific needs, and has been utilized for decades, especially outside of gaming. In the 1996 book The One to One Future, Don Peppers and Martha Rogers described a future of marketing and selling to customers over their lifetimes, as opposed to selling products to masses of customers at one time. The one-to-one marketing strategy starts with customized collateral that is personalized for each customer and incorporates messaging, offers and artwork.

Today, the amassing of customer collateral is allowing us to directly target customer groups with offers that will appeal to them, and it seems intuitive that this should extend to one-to-one marketing. An offer could include a specific message such as, “Mary, I know you enjoyed our Sundance Grill last week, so we are writing to give you the option to reserve a table and extend a free dinner offer to a friend.” The message would also include a picture of the Sundance Grill. 

This is clearly a personalized marketing offer, and could be further broadened to include a named acquaintance of Mary (if this data is available) that we know also likes to frequent the casino and restaurant. This kind of highly-personalized, mass-marketing program is now possible as we are adding deeper and deeper knowledge of the customers by drawing together data streams from across the enterprise. One-to-one marketing opportunities are countless when we know your customers from their demographic information, favorite activities and even gaming preferences. This information allows us to tailor customer messages, include pictures of favorite restaurants, table games and slot machines.


In our book The Math That Gaming Made, we described how “Casinos can... cluster their customers with loyalty cards and use this information to better understand their customers and increase their market share. Searching a large group of customer transaction data to find natural groupings or clusters of customers is an exploratory multivariate statistical method, which has become an important data mining tool and is referred to as ‘unsupervised learning.’”

The theory around one-to-one communication needs to be considered along with the fundamental requirement of gaming database marketing. This fundamental requirement is that we can group customers together into cohorts, and we can execute test and control on each cohort to measure the response to each of our marketing initiatives. In the world of gaming where one of our main marketing tools is free play and free play is essentially cash, it is essential that we can measure the results of invested marketing dollars.

The process of grouping cohorts is called clustering or segmentation. Clustering creates groups that contain items that are statistically similar. For example, a customer may have numerous attributes, a few of these being age, gender and visitation frequency; it is extremely likely that will share the same combination of attributes with many other customers. These similar attributes are used to create the groups.

In some cases, attributes that differentiate between customers groups are counterintuitive. To illustrate how attributes can become counterintuitive, consider a simple clustering based on only four attributes: age, gender, frequency and location. If there are 10 age bands, two genders, 10 frequency segments and a flag for local or out of state, then there are a total of 400 (10x2x10x2) combinations of customers. The out-of-state attribute is confounded with frequency, as high frequency (daily) customers are unlikely to be out of state. This confounding factor means that even though we have 400 groups, there may only be 250 groups that are active. The nature of higher dimensional space is that the more attributes that are added, the more confounded they become. In the example of recency (the number of days since a customer’s last visit from a fixed date), clearly high frequency customers are likely to have visited recently. Thus, the total number of groups have grown, along with the number of inactive groups.

A great majority of casinos segment players using some form of “worth” calculation which can range from average daily theo (ADT), total points earned for a period and so on. After the worth calculation is decided, bands of worth are created, usually between five and 20. Some casinos stop there, while others continue to add bands such as local, out of state or even frequency. This can mean a casino may have 80 total segments (e.g., two for local or out of state, two for frequency and 20 for worth). Even with tens or hundreds of thousands of customers, some of these segments may not have many customers while others have a great many customers. Introducing the one-to-one marketing strategy her, would increase the number of fixed segments rapidly. Adding additional attributes such as gender would result in 160 segments, 10 age bands results in 1,600 segments, distance to casino 5,800 (4,000 for the five distance bands among the local population plus 800 for the out-of-state population), and so on.

These fixed segments do have limitations however. Many of the segments have very few customers—or possibly no customers at all. Many of the segments offer no real separation of customers… is there really a big difference between $20 to $30 low frequency locals and the $30 to $40 low-frequency locals? Also, how different is the $29 customer from the $31 customer?

It is here that statistical clustering techniques show their value:  Clustering is a statistical modeling technique that basically says, “Let’s force the data into buckets!” If we force the data into, say, 20 buckets where the groups of customers are as different as possible within each bucket, we will have created buckets of customers that are both meaningful and distinctive. Marketing in clusters versus a fixed segmentation strategy allows us to know that we have created segments in which the customers included truly belong—and are truly grouped with like players.

Now that we have shown that we can, in fact, group customers into similar segments, the beautiful thing is that statistical clustering segments have customers that we have mathematically determined to be similar. This similarity allows us to run experiments on different sub-segments of the customers while having a strong control, and control is essential to the experimentation process. It is our position that the single most important thing casino marketers can do is experiment on their customer bases via this process, fine-tuning offers to maximize profits as often as possible.


Statistical clustering gives us segments that contain groups of customers with similar attributes. These segments can be the basis for different communication methods, with messages that are relevant to each segment. In practice, clustering can be applied, but if that is not available, a well-considered segmentation strategy that creates meaningful and distinct segments can suffice. 

One-to-one marketing means that we quickly lose what we desire in a segmentation strategy.  Each time a new attribute is added (to get closer to one-to-one marketing), the number of segments increase, they are confounded with attributes that may typically be correlated and the number of customers within that segment is declining. One-to-one marketing, by design, creates small customer segments—ideally with only one customer. Small segments mean we no longer have the ability to apply test and control strategies. 

In this area of discussion, gaming data is more volatile than in other industries, and thus it is vital to maintain a certain level of customer count in any test and control segment. So, we sadly wake up from our dream of personalized marketing for our customers and face mathematical reality of the need for test and control against cohorts of similar customers. This, however, does not mean we have given up on one-to-one communication.


As described above, to maximize the profitability of direct marketing offers requires manageable and testable segmentation strategies. However, we can redefine the dream of one-to-one marketing to be focused on customer-centric individualized messaging. With this approach, the desire is no longer to maximize profit (that is what offer testing is for), rather it is to maximize response—we are trying to communicate with our customers, the most important outcome is they respond to those communications. 

With this new threshold for success, the customer’s play behavior matters far less and behavior with other data sources becomes more important. For example, the decision to place a picture of a steak on an e-mail communication has nothing to do with the ADT of a customer; what matters is if they have been to the steakhouse before or not.

Here is where third-party data streams enter the picture to give us a complete profile customers. It is this complete picture that enables intelligent, automated, one-to-one communication. Here is a list of data that is available today (either directly or indirectly via data matching) for the casino—these data sources are ripe with valuable information about customers that can help ensure that messaging is timely and relevant (see Chart 1).

Social marketing—by its nature a kind of one-to-one marketing though a dialogue—should also be integrated into communication strategies. With a seemingly endless stream of personal data in this multichannel environment, we have to think carefully about how we engage with our customers in the active real-time social environment.

To drive the social dialogue, the third-party sources are wide and varied, including: Facebook, Twitter, surveys and third-party booking sites such as The casino database collates all its customers’ online comments through data streaming software. This software enables marketers to build rules to react to their customer’s life events (engagements, marriages and vacations) and time-based events (such as weekend plans).

An example of automated social dialogue: customer Karen is going out with friends on Friday. The one-to-one communication system sends an offer for her and four friends, valid Friday at her favorite in-casino restaurant. The marketing system knows Karen—she’s 42 years old, lives in an affluent suburb and receives magazine publications about sports and cooking. It weaves this information into the creative to ensure the right message is sent. As marketers, we dream of the opportunities this kind of one-to-one interaction would provide, and it’s all due to the amount and types of data available.

In gaming we are faced with the challenge of constantly optimizing our marketing programs. This optimization needs a feedback loop that is able to accurately measure the impact of each optimization step. The great challenge with one-to-one communication is to make it relevant to the customer by using data from across the enterprise and building this communication around a statistically valued database marketing program that can measure the value of our actions.