Decisions, Decisions By Andrew Tottenham, Managing Director, Tottenham & Co March 20, 2019 at 2:45 am If there is one thing we humans are not very good at, it’s making decisions. Ask people if they are good decision-makers, and most will tell you that they are. In fact, though, they are not, generally speaking. Stock and commodities traders and investment managers, for instance, are notorious for thinking they are better than they actually are. It is not enough to know what information we need to make a decision; we should also understand how we make them, if we want to make good ones. One of my heroes, behavioural psychologist Daniel Kahneman, believes we have two internal systems that work to help us make these judgment calls. There is the ponderous logical system that analyses the data and comes to a rational conclusion – what he calls the slow system – and the intuitive gut system that relies on an emotional response, the fast system. Each has their place. If you are walking along a road and a car is heading straight at you, analysing the speed and trajectory to discern whether the car is going to hit you or not is probably not the best thought process. In this example, you are better using the fast system to help you out. The slow system uses a great deal of brain bandwidth and consumes a lot of energy, and, as humans tend to be lazy animals, they do not engage it often enough. It is hard to carry out the most basic tasks – ones we take for granted without thinking, such as walking – when we are trying to solve a complex problem. Regardless, nearly all management decisions should engage the slow system. We need information and logical thought in such situations to help us decide what is the best way forward. In the main, there are three types of decision makers: those that rely heavily on intuition; those that think they use information to make decisions but, in fact, cherry-pick the data to support the decision they’ve already made; and those that do seek out information to assist them in making their decision. One of the critical factors, I think, that separates a good manager from a mediocre one is not only that a good manager uses logical processes to make decisions, but that they understand what information they need to come to the right conclusion. Business today is very data-heavy; sifting through the data generated to get at what is important is critical. But the process can overwhelm an individual trying to make sense of multiple gigabytes of data. Let´s look at slot machines. How do slot managers determine what machines to remove and what to replace them with? Some slot managers will tell you that deciding what machines to remove is easy: just look at those that are underperforming. If, for example, the win per day (WPD) is less than the average for the whole slot floor, those machines with the lowest WPD can be earmarked for removal. Some look at win per day combined with occupancy rates and try to remove those that fall into both low win and low usage. But are these the correct data points? What might happen to win and occupancy rates if you moved the machines to different places on the floor, or made some other floor change? Could it be that they are underperforming in both instances not because they are not the correct machines but because the location is wrong? A higher denomination machine, or a machine with different volatility or cabinet design, or some other characteristic, might perform better in that location. It could be the influence of the other machines in the vicinity making it perform poorly. Unfortunately, much of the information that manufacturers provide about the characteristics of their machines are either inaccurate or incomplete and do not allow managers to easily segment the machines by type properly. If you do not know what the real attributes of that machine are, how can you be expected to correctly determine what types of machines should be removed or purchased? When it comes to replacing machines, many slot managers are sold new machines; they do not buy them. In other words, generally, the salesperson will tell the manager what machines are hot in the market and why the manager needs to have them, and the manager duly takes note and buys them, expecting them to perform as well, or even better, than at least some of the machines already on-site. Maybe the new machines do well, but the overall performance of the floor does not improve. The manager has then spent money just to move money around the slot floor. It is too simplistic to think that replacing a low performing machine with a machine that performs well elsewhere, even a high performer on the same floor, is going to guarantee the positive impact the manager expects. On a single machine basis, perhaps, but not always. On an overall slot floor basis, rarely. Slot floors are dynamic: they are populated by people, after all. Changing one aspect in one place will necessarily have an impact elsewhere on the floor. There are multiple factors that influence customer spending patterns: game type, volatility, minimum bet, maximum bet, maximum prize, location, lighting, cabinet style, other slot machines, etc. It is practically impossible for a single person to understand how all of these factors affect customer behaviour on a particular slot floor without some kind of external help. Today’s slot managers are lucky. We now have computer software that can simultaneously analyse large amounts of machine and customer data, examine how these factors interact with each other, and measure the impact of previous changes. Using artificial intelligence, we can predict how a slot floor will perform if no changes are made and how revenue can be maximised by relocating machines on the floor. Based on your budget, these tools can recommend what and how many machines to change and what to replace them with, predicting the consequent change in overall revenue and expected return on investment. Nothing is fool proof, and no one can predict the future. But tools like these mean that, in many ways, there has never been a better time to be a slot manager.