Analytical Engine

If you thrive on uncertainty, now is a great time to be an analytics practitioner in the credit card industry. Consider the following scenarios.

Today - Mr. Smith opens a new credit card account with an attractive APR and minimal annual fee. After the first four billing cycles, Mr. Smith has already defaulted once on the minimum payment due, thus becoming a riskier proposition for the issuer. While banks in general generate fee and interest revenue on such accounts, the risk of default (60-90 DPD/write-offs) starts increasing. At minimum, issuing the card at that particular “teaser” APR seems like a wrong decision. Damage control begins with the card issuer increasing the APR, which slowly helps the issuer cover its losses.

February, 2010 – Ms. Jones follows Mr. Smith’s bad example and quickly becomes another high risk customer. However, since CARD Act regulations are now in place, the issuer cannot increase the APR in the first six months of issue and without 45-days notice. The probability of default and hence, the loss-given-default amount starts increasing in the wake of a few decisions gone bad, and because the bank now lacks the flexibility to re-price.

These scenarios highlight implications of the CARD Act and UDAP regulations, legislation that aims “…to establish fair and transparent practices relating to the extension of credit under an open end consumer credit plan, and for other purposes.”

Among other things, UDAP and CARD:

• Restrict all interest rate increases during the first year:
• Restrict interest rate increases on existing balances;
• Allocate payments in excess of the minimum payment first to the balance with the lowest APR:
• Treat a payment as late unless the issuer provides a consumer with a reasonable amount of time to make payment:
• Place limits on fees and penalty interest:
• Increase notice for rate increases on future purchases.

The CARD Act takes effect in February 2010, and has significant cost and strategic implications for credit card issuers, not just in terms of their acquisition strategy, or risk profile of their current/future base, but also a fundamental shift in their strategic outlook towards the credit card business.

We think the implications for the use of analytics within the card industry are profound.

• One of the first and foremost thoughts is: “Will the card industry shrink?” Issuers must carefully consider their exposure vis-à-vis the risks associated with it, given the new limits imposed on them. How should issuers evaluate their risk exposures differently, given the new regulatory constraints?

• Do we see “revenue-at-risk” models (a term more commonly used in telecom) becoming the latest analytics investment for credit card issuers, to better understand the pockets of revenue that are most at risk for issuers? How are the issuers going to deal with the challenges of revenue replacement? How will the issuers treat the transactors vis-à-vis revolvers?
• Will Lifetime Value (LTV) or profitability modeling supplement or supercede risk/price modeling as related to the overall underwriting function, both for acquisition and base management?
The big strategic question is this:How should issuers look at their portfolio and their products now to determine the customers they want, and the products and pricing they offer in order to retain profitability and position for competitive growth in the new era?

For analytics thinkers, here is an example where numerous forms of credit card analytics (pricing, underwriting, risk management, customer acquisition, customer management, and customer attrition, ) are suddenly in play, and all at the same time, together. Everyone wishes for a quick solution to this development, realizing fully well that there is none. The risk models and pricing strategies used by issuers have been built and stabilized over decades with great attention to detail. Even so, they require careful handling every few cycles as the underlying economic and social structure of the population changes. Given all this, is there a good near term fix that can be rolled out for testing?

The road ahead will be strewn with a series of analyses and tests that issuers will experiment with to assess the impact of the changing regulatory environment on the credit card industry.
In a subsequent post, we will talk about the different analytics that we think will be of immense value in these changing times. But for now, let us all think about the change and its impact.

Steve Shu had quoted his wife (Suzanne Shu, a Professor of Marketing at Southern Methodist University) on the issue of indefinitely delaying reward redemptions.

Shu’s comment is interesting from a behavioral model for rewards redemption point of view. A credit card holder will be interested in those redemptions that suit his choice and utility function. True- a rewards program in itself manages to alter the choice function available for the credit card holder.

The linked article and another US News article highlight some interesting attributes of rewards redemption by consumers

1. Incentives –

1. Leading to purchase acceleration
2. Guilt/Excitement of opting for something that they would not otherwise

1. Indefinitely delaying an aspirational redemption for a special occasion

1. Cost of Card
2. Benefits from Reward
3. Cost of payments/credit card usage

The closest we can come to creating a consumer choice/utility function in the database driven world is by collecting and integrating 5 different sources of information –

1. Demographic Data – Age, Gender, Ethnic Background, Education level, etc. Available through census, application forms filled by customers, etc.
2. Economic Background – Income levels, saving and investment data. Available through application forms, relationship level view of a customer, etc.
3. Behavioral Data – attitude, interests, etc. Can be collected through surveys and lifestyle data bureaus.
4. Bureau Data – e.g. risk bureaus such as Experian, FICO score, or lifestyle data from Claritas.
5. Transaction Data – prior credit card usage data, payment data, etc.

Demographic and behavioral data together would be ideal to define a person’s wants, while to build a choice model, it would be a good idea to include Economic and Bureau data as well. The transaction data can be used for validation and refinement of the choice model.

Back to the articles, a choice model can help us explain and manage all three aspects–

1. Incentives – through a greater understanding of choice limitations and utility functions, we can customize offerings that maximize the utility for a customer at the lowest possible cost
2. Disincentives – any alternative outside the “want” area and significantly beyond the choice area suffers from possibility of being delayed, as pointed out by Suzanne Shu.
3. Cost-Benefit Analysis – an economic model of choice can bake in the tangible costs of a reward program to optimize profitability.

Some of our other posts on credit card analytics are here

Ron’s post made me think about the implications of this suggestion on the global banking system. Getting a dynamic price optimization algorithm at a relationship level in place requires an integrated view of the customer.

Almost every bank that I’ve worked with or have talked to, suffers from the problem of disintegrated views, LOBs competing amongst each other, and the customer being valued at the product level and not the relationship net worth level. What’s surprising is that every bank is aware of this problem, is trying to address it in its own way, and still far from being successful.

Ron talks about the futility of this pursuit and the importance of Loyalty Programs in achieving it. Let me look at another side of the story – The strategic focus required for solving this problem. Banks are either sales focused, or profitability focused.

Sales focused banks play the volume game with hundreds and thousands of cards getting shipped every day to unsuspecting customers. The same customer may have a silver, gold, platinum card from the same bank. The idea is to carve out a bigger and bigger share of the wallet without getting bogged down by the profitability levels, as long as even marginal amounts are being made. The same customers are being approached for mortgages, auto-loans, refinancing schemes, personal loans, etc. The pricing of each of these products is done individually, and I mean “individually” at product level – each product is priced separately, The debate is around moving to each individual getting priced separately for her/his bouquet of products.

Profitability focused banks operate on the P&Ls and margin levels for each product. They keep the back-end and front-end of each of their products separated. There is positive money and higher margin, but not a lot of it (in some cases, there might be a lot of money in niche operating areas). However, at some point all such players realize that scale comes from the masses. A seemingly premium American Express card gets offered free for a lifetime in some of the developing countries.

Now, having said that there are two kinds of banks, both of them eventually will try to move in one direction – the integrated view of the customer followed by a relationship management approach. Loyalty programs may help get more data and act as an incentive for customers to proactively get their unified integrated presence in the bank databases. However, loyalty programs are still a piece-meal solution. The effort required is higher, turnaround times longer, and vigilance is incremental. Changes need to be introduced gradually and their effect/success needs to be monitored.

Compare that to the “break the bank apart” module before reorganizing it. The upfront investment is higher, the complexities overwhelming and the implementation time can be long initially. What the banks need to realize is that the investment required to upgrade to an integrated view of the customers is a strategic investment to be amortized over several years. Any bank ready to take the pains and go through the organizational and technical barriers will find itself way ahead of its competition. The proof of this lies in the gradual advent of Direct Banking, which forces an integrated view on the customers for all the products offered since there aren’t multiple human-to-human touch-points.

At Diamond, one of our key strengths has been the ability to think about, create plans for, and help our client implement both. Helping client redefine their data systems, architecture, program managing it to successful implementation, followed by the ability to create a relationship view and management approach. Our analytics approach has been geared at looking not just at the problem at hand (increasing sales, improving profitability) but also at increasing the impact of the business for overall success.

A discussion I had with a group of analysts suggested, based on rigorous data analytics, that rewards points and redemptions as value added services do help credit card companies retain a few more customers [This image here does prove the retention hypothesis], but its a huge cost on the account level P&L⁽¹⁾. Probably, one of the worst rewards offer is Air travel related redemptions. So, if you are a credit card company and you want to advertise your rewards catalogue, be cautious about advertising Airlines Redemption, is what the group suggeted.

The premise of rewards program offered by credit card companies, traditionally, has been low redemption rates and a lot of people not even redeeming their points ever. Add to it the hassles faced by customers in redeeming, you have defeated the purpose of the program. Why offer a value added service you don’t want people to use (and create negative customer experience)? And, just in case someone ends up using it, why compound the problem with an unfavorable cost structure for such a service?
Assuming that rewards programs are aimed at retaining customers, are the redeemers those customers whom I want to retain through this offering? In other words, am I retaining the right customer?

Lets focus on the second question - Who is the right target for these rewards? Someone who stretches the margin the most (through frequent high cost redemptions, hitting P&L through rewards costs as well as high customer servicing costs), or someone who is generating good revenues through revolving credit, relatively infrequent bulkier redemptions and is sticking on to the credit card because of a certain variety of reward?
Offering rewards as a retention ploy would imply ceilings and floorings to redemptions (by type) for managing customer profitability. For instance, credit card companies impose hypothetical floorings by allowing air ticket redemptions only if someone is utilizing at least 20,000 points or more.

• Another strategy for rewards programs could be aimed at identifying customers with a lower propensity to redeem while maintaining a higher retention probability given rewards utilization.
• However, the best strategy in such cases would be to build multi-stage rewards and retention model that predict point balance utilization, preferred category of redemption, retention driven by rewards usage and most importantly, P&L sensitivity models for rewards usage. This requires deep analytics expertise and a thorough understanding of credit cards’ rewards program.
• Additionally, this report coming through Celent highlights how rewards disposition is likely to change going forward and move towards blended rewards and cash-backs and the economics are likely to become more favorable by leveraging low cost internet, mobile advertisement channels.

⁽¹⁾For P&L, however – lets look at this example⁽²⁾

Include the various maintenance cost items, program costs, and servicing costs on top of the air ticket cost and you probably end up with a non-existent or a very “thin” margin!
⁽²⁾ Under the assumption of zero revolving credit and other sources of revenue that credit cards’ companies generate. While revolving credit is the biggest source of money for credit card companies, the attempt here is to show that at the base level, the credit product is becoming unprofitable because of a faulty rewards program
⁽³⁾ Optimistic, my friends tell me!