NASHVILLE – “Even a little prediction goes a long way,” according to one expert here, who showed how the smallest pieces of information can dramatically improve the results of a financial institution’s marketing campaigns.
Dr. Eric Siegel, founder of the Predictive Analytics World, told TMG’s Executive Summit here, showed that by deploying a targeted marketing campaign using even the simplest of predictive models, financial institutions can expect profits to skyrocket -- all without a new product or new marketing creative.
A large amount of data is also not necessary for valuable, predictive insights, according to Siegel. He pointed to a one-time study by Chase, for instance, that showed mortgage holders with interest rates higher than 7.94% were exponentially more likely to refinance with another lender than those with lower rates. “Already, Chase has value with just that one factor alone.”
Predictive modeling doesn’t always have to be about earning additional revenue from new customers, Siegel explained. Financial institutions can use some of the same methods for predicting those borrowers who are most likely to miss credit card payments, for example.
Meanwhile, there are two of what Siegel called “killer apps” that will be particularly important for financial institutions to master in the coming months and years: target modeling for acquisition and churn modeling for retention. With regard to the latter, Siegel said it’s important to consider the “let sleeping dogs lie” method. This describes being careful not to deploy marketing, discounts or offers that may backfire, triggering a negative behavior, such as the canceling of an account.
The Bottleneck
Half the battle is getting raw data into what Siegel referred to as training data, which is essentially lines of data describing the attributes and behaviors of individual consumers over time. “This is often the bottleneck to leveraging predictive analytics technology,” he said.
Because of predictive modeling, which Siegel calls the next phase of evolution in the Information Age, each of a consumer’s daily experiences is changing. Millions of decisions based on decision trees and other analytical methods are made by companies, law enforcement, government agencies and others every day. These decisions inform how consumers are treated, serviced and incentivized. And often they are based on both straightforward and unusual factors, such as the size of dog collar purchased at a pet store or whether or not a consumer prefers curly fries to straight.
What’s important in this context, said Siegel, is that correlation doesn’t always equal causation. In other words, just because your model indicates straight-fry fans default more often than those who prefer curly fries does not mean eating straight fries directly impacts one’s ability to pay bills.
All of that information can create problems for some leaders. “You will have people in your organization who do not want to believe the data,” Siegel told the meeting. “When you see a connection between two things, be satisfied that it will help predict. Look at the link as a building block, not necessarily as an explanation of why something happened.”