There isn’t a week that goes by it seems where I’m not receiving at least one mailer for a pre-approved credit card or personal loan. Maybe you’re like me and you’re just annoyed at the waste these cost and don’t really consider other thoughts before you promptly recycle them. However, have you thought about all the layers that place you as a target for these advertisements?
There are obvious ones, if you’ve opened a new credit card recently, you’ve ever used a personal loan, you are responsible in paying your current credit cards on time, etc. But what is the history of this idea of credit that we rely so heavily on in this country?
Now I don’t have time to go fully into credit and credit cards but if you’re interested, here’s a video.
What’s fascinating to me is how credit approval was drastically transformed in the 1980s when Richard Fairbanks and Nigel Morris decided to utilize information technology to create a predictive model for profitable customers.
Before this, banks actually had uniform pricing for credit cards because:
- companies didn’t have enough of information systems to be able to handle differentiating pricing
- banks didn’t think that customers would be ok with price discrimination
This point struck me because there were definitely other ways of discrimination when we look at the history of credit approval, redlining being the most obvious. But that’s another blog post in and of itself.
So to give the abbreviated version of the story of how credit approval being differentiated across the board, we look at the work of Fairbanks and Morris. These two wanted to move towards a predictive model to be able to offer different terms with credit cards. The big banks at the time were seeing success, after all it was the 70s when credit card spending continued to grow. Finally, they were able to convince a small regional bank in Virginia: Signet Bank to allow them to model on profitability. This was different than how banks previously looked at customers of whether or not they would default, but rather if they would make the bank money.
Looking at the customers, Morris and Fairbanks knew the proportion of profitable customers was very small (because most customers are either break-even or money-losing). By being able to predict the customers that were profitable, they could make better offers to such customers.
The problem with implementing this was there was no current measurement for the profitability of customers. So they had to acquire the data by conducting experiments. How did they do this? They randomized offereing different terms to different customers. This was a very risky way to obtain data. However, through this experimenting, they were able to gather the necessary data and build the predictive models they were looking for to be able to accertain profitable customers.
This risk is fascinating, as such a small bank could have easily lost on this gamble, but instead, they became a leader in credit card operations.
While some of you may already be aware, while Signet Bank is one you haven’t heard of, I’m sure you’ve gotten a mailer or two from its spin-off: Capital One.
This is a fascinating example of how being willing to accept the costs of data acquisition lead to a strategic advantage in a business.
Source: Data Science for Business