One problem with training bank staff is the proverbial “Man with a hammer syndrome.” If you only have a hammer, then every problem appears to be a nail. In other words, if your bank has only solved problems one way, then it is a fair bet to say that they will keep on solving problems the same way. The problem is that bank staff needs to be trained with new mental (or physical) models so that they have more tools than just a hammer.
Let’s take customer profitability for example. Many bank staff are not clear on what constitutes a profitable customer. They may assume it is the contractor doing construction, the frequent branch visitor or borrower that is willing to pay a high margin. However, this problem is easily and quantitatively solved as by looking a set of existing profitable customers we can analyze the data and find out what is correlated to profitability.
Below, is just one of our many models at CenterState Bank that shows that our ultimate customer is a commercial business, with experienced management, in a growing industry, with limited bank relationships that produces 15% top line revenue growth per year and grows its employee base by 10%. If bankers stay on the upper outward bounds of the decision tree, they are likely to end up with the optimal customer. To the extent they move into the interior, the customer is less likely to be profitable.
In one fell swoop, we can now quickly train all staff, but particularly business development, on how to prospect for profitable customers. In addition, we have given them an ongoing tool to use after training. In essence, what happens is that signal detection is simplified and more accurate so bankers can pick a profitable customer from the noise of the general community population.
The Impact of models are exponential. Like an Erector Set, the more models we have, the more we can build. The more we can use multiple models together, our bank’s advancement can occur in an exponential manner. These models give bank staff a checklist and a set of tools to solve problems in a variety of disciplines from loan origination to BSA compliance.
Keep in mind that each node of the decision tree algorithm can be created and then iterated so that over time models can be refined and knowledge can be passed on. Boxes can be added or subtracted and the process restructured. In addition, each variable can be set to be maximized (such as revenue growth) or minimized (such as credit blemishes).
Creating an algorithmic decision tree helps bankers visualize interactions and start to quantitatively define what is important. A good algorithm allows your bank to better institutionalize knowledge and speed up choices while improving accuracy.
Submitted by Chris Nichols on March 31, 2015