Delving Into Consumer Spending Habits Helps Tailor Marketing
Understanding consumer credit card spending habits has been difficult historically. That's because spend has not been an item found on the consumer credit report. So while you might have known how your members used the credit cards you issued them, you had no real insight into how they spent on other cards in their wallets, or how prospective members spent on their cards.
Recent enhancements to the consumer credit file have allowed TransUnion to estimate spend across a member's revolving credit products. These enhancements include the incorporation of actual payment amounts at the tradeline level, and the presence now of a history of key metrics (as opposed to just at a single point in time), such as balance and actual payment. TransUnion's ability to calculate estimated spend with improved accuracy using credit file data alone opens tremendous opportunities to better understand member behavior and preferences. This article discusses a recent study aimed at gaining those important insights.
Defining a Spend Estimator
Calculating an estimator of monthly spend begins with an understanding of how balances change. Consider the following equation:
We can rearrange to solve for spend.
Interest and fee data are not available on the credit file. However, even at 18% APR the monthly interest charge will only be 1.5% of the balance. As for fees, they only account for about 15% of revenue in most prime card portfolios — the interest counts for the vast majority otherwise. Thus it is fair to estimate spend using the following equation show at the top of this page.
This is the formula we used throughout our study.
Although there are many different ways to look at spend, we decided to build our analysis across three different dimensions of spend behavior:
Level — the amount one spends at any given time, or in a given time period.
Concentration — Also known as wallet share, how consumers distribute their spend across the various cards in their wallets.
Seasonality — The timing of consumer spend, e.g. summer vacations, the beginning of the school year and of course the year-end holidays.
Insights in each of these dimensions have different practical applications in the management of a credit card portfolio, ranging from member acquisition, promotions and seasonal campaigns to line management, pricing, retention, and much more.
We began by analyzing the level of consumer spend, comparing historical spend to current balance as a predictor of future spend levels. Remember, in the absence of historical spend data, most lenders used balance as a proxy. We considered approximately 2 million randomly selected consumers and evaluated their spend from April 2012 to March 2013, looking at it both in terms of rank-ordered balances as of March 2012 and also in terms of rank-ordered aggregate spend from April 2011 to March 2012. The results in the top chart show that past spend level is a far more accurate predictor of future spend across the board than is balance, in particular accurately identifying the highest spenders more than twice as well as did balance.
The average U.S. consumer has 2.6 active general purpose cards in his wallet. This number doesn't include 2.4 private label cards per consumer, for those who carry private label cards. Clearly, consumers have a choice of payment vehicle, and credit card issuers fight for primacy in the consumer wallet. Our study measured what percent of total spend was put on the top card in the consumer wallet over one year, and used that measure to determine the correlation to the wallet share of the top card in the following year (see second chart).
The results demonstrate that a consumer's spend dispersion tends to be persistent year over year and may yield a proxy for member loyalty. In other words, the majority of consumers who concentrated their spend on one card continued doing so in the following year, while those who dispersed their spending over many different cards exhibited the same behavior thereafter.
Consumer spend follows clear seasonal patterns. For example, our study revealed that median spend in December is 35%-40% higher than in any other month, and in fact 20% of consumers had twice as much spend reported on their cards in December as they did on average throughout the rest of the year (see third chart). We also found that seasonal spending patterns are consistent — consumers who were rank ordered into 5% bands based on their 2010 holiday season spending fell into the same 5% bands in 2011. And even for high seasonal spenders (top 20% overall), the level of holiday card spend in 2011 by 5% bands perfectly rank-ordered spend observed in 2012.
Finally, we compared the forecasting accuracy of traditional spend models — those based on legacy credit file attributes — to models based on our enhanced credit file attributes, and also to credit risk scores.
As expected (see bottom chart), credit risk scores are poor predictors of spend. This is not surprising; it is a task for which they were never intended. Also not surprising, our enhanced data model was significantly better at separating between high- and low-spending consumer segments than the model based on traditional credit data.
Our findings indicate significant correlations between past and future spend levels, stable dispersion of spend among cards in the member wallet, and consistent seasonal spend patterns from year to year. This is just the tip of the iceberg. Many cogent insights await now that more insightful data is available. Credit unions can use these findings to better market to members, allowing members to receive more appropriate interest rates and credit lines, timely promotions and customized rewards — in short, a better fit of products to their needs.
Vladimir Prupes is director of research and consulting and Ezra Becker is vice president of research and consulting in TransUnion's financial services unit.