Complexity of credit card data

This entry is just a little bit of what I have experienced after more than one year working at a bank. By no means am I trying to imply that I am an expert in credit card data. Not even close. This is simply to shed some light or pull the curtain a little bit into the world of banking. It’s also a way to put into writing what I have learned or think I have learned so far.

Lost and stolen

One reality that makes compiling and analyzing credit card data complicated and tricky is that people lose their cards or have them stolen all the time. When that happens, banks issue new cards to customers with totally new account numbers. On the front end, customers don’t see any change. New cards just work like old cards. However, on the back end, it’s a totally different ball game. If an analyst like myself wants to analyze performance of cards, especially accounts that have a complicated history, such as spend, balance and profit history, a reconciliation of all the account numbers associated with the original account number can be daunting and easily lead to mistake.

Tricky balance between risk and profit

If you are like me who uses cards regularly and pays off balance every month to avoid financial charges, you are not a profitable customer to credit card issuers. To make money off customers, issuers usually charge interests on late payments and fees such as annual fees, foreign transaction fees, over limit fees. No late payments mean no financial charges and no profit. On the other hand, customers who are usually late on payments and accrue balance are usually the riskiest bunch. Such customers, as we call revolvers, are more likely to charge off and cost issuers money. As a result, issuers must find a sweet spot between risk and profitability. Having the majority of customers under 680 FICO can lead to higher profitability, but also higher risk at the same time. Acquiring great credit customers poses a lower risk, but also a lower chance of profitability.

One popular practice in the credit card world is balance consolidation. In a layman’s term, it is an offer allowing customers with a balance at a higher interest rate to transfer the balance to another account at a substantially lower rate which can be 0% for several months. This offer appeals to credit trapped customers who are likely to be late on payments, but want to avoid interests. By taking on more accounts and balance, issuers take on more risks, but balance consolidation may not always lead to higher revenue and profit.

Ambiguous definition

Sometimes, metrics in the credit card world may not be as straightforward as they sound. For example, if a customer calls to close a credit card, but chooses to pay balance in installments instead of one big payment, is that credit card considered closed or still open with an outstanding balance? When discussing “active” accounts, are we talking about “balance active” (having balance > 0) or “debit active” (having net purchase which is comprised of purchases minus refund > 0)? When discussing spend per account, are we talking about spend per account or spend per “debit active” account only? Those are just a few examples on top of my mind. There are plenty more cases where a question, straightforward as it may seem, is really not.

Apple to apple comparison

A performance analysis needs a benchmark to compare against. To analyze a cohort of accounts, it’s important to find another cohort that is as similar characteristically as possible to the one at hand. The challenge is that there are multiple elements to consider such as acquisition channels, vintage (when accounts were opened), cashback or rewards cards, FICO, credit bureau segments (more on this later), brands that cards are associated with (FCA, Delta Airlines, Walmart…) and so on. If the task is to perform an analysis on the same set of accounts over time to test the effect of an offer, seasonality needs to be considered. Obviously, consumers spend more during holidays. So, comparing spend in Dec to spend in Feb isn’t really a fair one.

External data

Issuers work with agencies such as Acxiom to acquire demographic data. To my best knowledge, issuers send account information such as addresses and names to Acxiom to match up with their demographic database such as household income, net worth, how likely they have a mortgage, how likely they have children, how they are etc…Such information is sent back to issuers for future marketing purposes. One of the challenges arising from this practice is that an account can have more than one user. Joint accounts can have two users whose demographic profile can be different. In that case, a decision has to be made as to which profile should be used and data needs to be handled with care. Otherwise, it would lead to double accounting.

Issuers can work with credit bureau agencies for credit score and other information. An individual banks with multiple corporations and has multiple cards with different issuers. It’s helpful to know how big of a customer’s wallet share your issuer makes up. For instance, it’s helpful to know how much a customer allegedly tends to spend in a year and the percentage of that spend goes to your issuer. The same goes for all credit card balance and revolving balance. Agencies such as Experian receive data from different issuers and return their estimated data back to issuers for marketing purposes.

Another challenge is that between the time when data is sent to external data partners and the time when data is sent back to issuers, accounts can be lost or stolen. Reconciling that change can be problematic.

I have worked intensely at my current company for the past year. It’s really satisfying to learn how data works, what goes under the hood, what the nuts and bolts are and what can be done to help the business. There are a lot to learn and do such as

  • How to optimize the acquisition process? How to get the best response rates for Direct Mail, an important practice in this industry
  • How to encourage spend?
  • How to balance risk and profitability?
  • Who are the “best” customers?
  • How to use data to improve customer experience
  • How to improve interchange?
  • How to predict charge off?
  • How digitally engaged are customers?

I hope this entry has been helpful somewhat. If I miss anything or any point is incorrect, please leave a comment. I hope I’ll learn more and be able to share with you in the near future.