Rick: We are speaking with Thomas Grady today regarding utilizing Artificial Intelligence and Predicative Analytics in the casino industry, Welcome Tom, we should start with your background and experience in such a unique combination of skill sets.
Tom: Thanks Ricks, it was a long, long time ago when I started my career in the IT industry, back in the early 90s. I began my studies with certificates and diplomas in Database, Programming, Networking and Hardware Design, and I found the most enjoyment in developing Database solutions combining the structure with programming in different languages to present the data. In 2001 we moved to Las Vegas where I fell in love with working in the Casino Industry. First, Information Technology then a transition to Marketing where I was able to combine Marketing, Database and Programming focusing on guest behavior and predictions.
Rick: Speaking of predictions, isn’t Artificial Intelligence and Predictive Analytics the same thing?
Tom: That’s like saying plain and peanut M&Ms are the same thing. Yes, they are both M&Ms but very different. Say we want a car to drive through a box, picture a box with an opening at one corner (start) and another opening at the opposite corner (finish), easy draw a straight line from one end to the other, done. Now put in obstacles between the two openings and cover the box to hide those obstacles. Now you need help. So, you ask GPS for directions. The GPS system collects data from numerous scenarios and will tell you it's best guess based on what the data says. You, being human, decide if you want to follow its directions or try something different. That’s predicative analytics. Now take a self-driving car and tell it to find the quickest way from start to finish. You set up rules like turns must be 90 degrees right or left, or it can go straight then you let it go. The self-driving car makes the decisions and tries over and over till it makes it through. Then it will keep trying to find quicker routes until it finds the quickest route and sets that as th
e best possible route for a quick arrival. That is Artificial Intelligence or AI.
Rick: Interesting, but how can that be used in the Casino Industry? Specifically in Marketing?
Tom: It can get very complicated, very quickly, but for this discussion let's simplify things. Let’s take new members only. We can simply say our promotion is new members get $10 Free play. This is just to try and build the database. That’s a form of blind marketing. The promotion fills the database but what’s the value? Now let’s say a promotion of give us your email address, we will send you $10 the following week or $10 match play to produce a bounce back. Now we have 2 trips to base future offers on. Now bring in predicative analytics, if your first trip Theo is $X you get $X within the next week based on previous first-time players, this simple calculation you would still need a few full-time employees to monitor the program. Lastly, to expand the program to Artificial Intelligence, like the self-driving car, we give the program our goal. We want to develop players from the new member pool into high frequency and high value players and as casino operators that is what we want, right?
Rick: Absolutely, but can Artificial Intelligence find high frequency and value players?
Tom: Yes, just like the self-driving car you give the program li
mitation, the turns of 90 degree right, left, or straight. You give the program instructions such as sending an email in X number of days after signing up such as 3,4 or 5 days after signup to reengage, then the next turn is sending the player free play/match play, Dining offer or Retail offer depending on the results. Let’s say we sent free play. The next instruction becomes base free play offer on first day of play, or 2 X’s offer based on first day of play or an earn and get offer. The list goes on and on. You can see how it can get complicated very quickly but if you want to maximize your ROI, you will have to give the program room to test and determine the best possible offer for the player.
Rick: Wow! I can see how the choices can get complicated
Tom: Yes, it can. Imagine all the possibilities to where you quickly get outside the possibility of human analysis, consider segmenting based on Distance from property, time of Signup (early AM vs. Midday vs. Late PM signups) these are different player types and have different behavior patterns. What does the
player respond to? Is it special events (Birthday, Anniversary), or offer by age grouping, or natural neighbor?
Rick: Wait, what is a natural Neighbor?
Tom: Aww, it is very difficult to determine a married couple in the database, yes it can be flagged at the players club, if they know and if they remember, but would you want to know married
couples, unmarried couples, siblings, groups of players like frequent bus riders or common residence building. If we set the system to find people that join the club within half hour, and have the same last name, or email address or physical address. We can also look at players that play next to each other or in the same area consistently and the start and end time are similar, we can consider those natural neighbors. We can further the programming and expand offers. Knowing we're getting in 2, 3, 4 even sometimes more natural neighbors, increases the ROI as a group and not by individual, but we're getting off topic.
Rick: Yes, but it is very interesting and another great topic to expand on another day, so to conclude the discussion of artificial intelligence vs predicative analytics it really comes down to giving the program your final goal, setting limitations, and trusting the program even when it is wrong?
Tom: Exactly! I think everyone reading this interview would agree you can’t send a player one hundred times his worth to return the next week, they are going to play it and leave. The AI program would try it a couple times before determining it’s the wrong thing to do. That’s why we put the limitations in place. But if you have ever thought…what would happen if we did this or that? Let the AI program do it and see what happens. You might lose once or twice but what if you don’t and that decision turns into the most profitable promotion or offer and creates more loyal players?
Rick: Terrific way to end this topic, it was a pleasure speaking with you Tom
Tom: Pleasure was all mine, thank you Rick. I love talking about this stuff, and keep in mind we just discussed new members, imagine the possibilities of core mailers, decliner offers and disengaged campaigns. The adjustments that can be made and allowing AI to do the work. There are so many ways we, as casino executives, could expand and find profit that is currently being left on the table or lose wallet share to competitors
Rick: I agree, it sounds like artificial intelligence has its place and will be utilized in the near future, thanks again Tom and I look forward to our next discussion.