August 2011ADVANCING INSIGHTS ADVANCING COMMERCE.
TRANSACTION SCORING: WHERE RISK
U.S. card issuers, broadly speaking, are employing risk management techniques
developed during a time of growth. These techniques need significant enhancement,
given issuers’ current lack of liquidity as well as a more conservative and cautious
consumer population. Transaction scoring can increase the predictive accuracy of risk
management by identifying risky behavior in real time, while simultaneously reducing
false positives. This technique enables issuers to intervene in accounts earlier and with
greater precision, providing a wider range of options in terms of next steps. New
technology in form factors, like mobile, make real-time intervention a reality. Using
transaction scoring as a component of credit risk management also allows issuers to
guide account holders into more responsible use of credit.
Transaction scoring in conjunction with existing models can
deliver lifts of 10 to 15 percent in Kolmogorov-Smirnov (KS)
tests over traditional scores alone.115%
U.S. issuers face a daunting challenge: Finding growth opportunities in an industry that is being
buffeted by contraction and constraint. Transaction scoring—the use of real-time transaction data
in scoring for credit risk—can increase the predictive accuracy of risk management by creating
account-level variables that identify patterns of behavior in real time. Using transaction scoring
helps issuers see more deeply and clearly into their customers’ behavior, resulting in more precise,
measured, and responsible treatment of risk. The benefits of transaction scoring extend beyond
risk management to identifying opportunities. By reducing false positives, transaction scoring can
provide greater visibility into positive or neutral use of credit cards, which issuers can leverage in
cross-sell, up-sell, promotions, or product fine-tuning. Transaction scoring:
• Is built on a foundation of industry best practices
• Can assist institutions in more accurately pricing new exposures at origination, given Credit
CARD Act limitations
• Provides a benefit to both financial institutions and consumers by allowing issuers to more
finely segment their portfolios and more accurately design interventions for each kind of account,
risky or not
The last point is especially important in light of current economic and regulatory conditions.
Transaction scoring provides a true social benefit by keeping the banking system and payment
networks open to as many consumers as possible.
BY Westley Koenen and Theodore Iacobuzio
2 ADVANCING INSIGHTS ADVANCING COMMERCE.
By creating account-level
variables that identify patterns of
behavior in real time, transaction
scoring helps issuers see more
deeply and clearly into their
customers’ behaviors, resulting
in more finely grained and
NEW TIMES CALL FOR NEW MEASURES
The challenge facing U.S. credit card issuers today is clear: How to expand a
business that has undergone large-scale contraction. MasterCard believes that
the use of transactional data in managing portfolio risk represents a significant
opportunity—one that can play a major role in meeting this challenge.
Transaction scoring, along with the proprietary techniques necessary to leverage
the data in actionable ways, can increase the predictive accuracy of risk
management. It does so by identifying risky behavior, reducing false positives
and, importantly, making that information available earlier than is possible in
traditional approaches. By creating account-level variables that identify patterns
of behavior in real time, transaction scoring helps issuers see more deeply and
clearly into their customers’ behaviors, resulting in more finely grained and
The growth the industry experienced between 1992 and the crisis of 2008
is a matter of record. The U.S. credit card business grew at a compound
annual growth rate (CAGR) of better than 10 percent from 1993 to 20072
and consistently delivered a 3 percent annual return on assets between
1992 and 2008.3
The contraction is also public information. Total consumer debt shrank $1 trillion,
or 8.5 percent between October 2008 and the second quarter of 2011, as shown
in Figure 1.4
In 2008, there were nearly six general purpose credit cards, excluding
debit and private label, per U.S. household.5
By the end of 2010, each household
had about four.6
Total credit limits (contingent liability) declined 26 percent,
nearly $1 trillion, between June 2008 and the first half of 2011.7
In broad measure, of course, the deleveraging accomplished its goal. Credit
card charge-off rates continue to decline—to above 8 percent in April 2011,
down from a high of more than 13 percent a year prior (see Figure 2).
FIGURE 1: Total U.S. Consumer debt outstanding (balances) have declined
$1 trillion, or 8.5 percent, since October 2008
Total U.S. Consumer (Balances) Outstanding
Source: Equifax, April 2011.
3TRANSACTION SCORING: WHERE RISK MEETS OPPORTUNITY
Transaction scoring represents
a much-needed addition to
the issuer’s compliance toolkit.
Once the predictive value of
transaction scoring is brought
to bear, issuers will likely find
it easier to comply with the
operational and credit risk
requirements of Basel II.
figure 2: Bankcard write-offs are declining, but not to pre-recession
Bankcard Write-Off Rates
3-Month Average Write-Offs
3-Month Average Write-Off Dollar Volume
Source: Equifax, April 2011.
Reducing loss is only one side of the risk management equation. Managing the
changing outlook with regard to revenue and income is the other. That’s what
makes the environment currently confronting U.S. issuers so unfamiliar and, in
many ways, so threatening. U.S. consumers have become more cautious, more
committed to savings, and more determined to pay down credit card debt.
Issuers’ risk management approach, developed during a time of unprecedented
growth, must change to reflect this new reality.
Transaction scoring represents a much-needed addition to the issuer’s
compliance toolkit. Once the predictive value of transaction scoring is brought
to bear, issuers will likely find it easier to comply with the operational and credit
risk requirements of Basel II.
Transaction scoring also has value in enabling issuers to see the flip side of risk,
which is, of course, opportunity. This flip side has two dimensions. First is the
ability to isolate those accounts that might seem risky but, in fact, represent
healthy use of the card and carry the potential for future revenue—i.e., false
positives. For these good accounts, transaction scoring can provide greater
visibility into positive or neutral use of the card that issuers can leverage in up-
sell, cross-sell, promotions, and product fine-tuning.
The second dimension has to do with accounts that represent an actual—
or potential—risk for the issuing institution. For these accounts, the risk
management process can become the basis for constructive conversations with
account holders in addition to more nuanced credit treatment. Both would be
aimed at inducing more judicious and, eventually, more profitable use of the
card—ideally, before an adverse action code is generated.
4 ADVANCING INSIGHTS ADVANCING COMMERCE.
HOW TRANSACTION SCORING WORKS
To put transaction scoring into practice, specialized and proprietary techniques
must be used to leverage the data and integrate it with the issuer’s existing
decisioning mechanism. That said, transaction scoring:
• Is built on a foundation of industry best practices, including historical scoring
• Provides significant and quantifiable lift in identifying delinquent accounts
earlier and reducing the number of false positives
• Can assist institutions in more accurately targeting new exposures, given Credit
CARD Act limitations; credit card transaction data has proven to be highly
predictive for up-selling existing cardholders to premium payment products
and/or cross-selling other products such as home equity loans and mortgages
• Provides a benefit to both issuers and consumers by allowing issuers to more
finely segment their portfolios and more accurately design interventions for
each kind of account, risky or not
The key benefit that transaction scoring provides is visibility into account
activity. For example, by using transaction scoring in tandem with aggregate
consumer transaction behavior, issuers can develop segmentation strategies
based on real industry movement in terms of consumer merchant preference.
They also can use aggregate consumer behavior data to map individual patterns
against industry patterns, regional patterns, and even local patterns. This could
enable issuers to more quickly identify early warning signs of industry sectors
that may be experiencing weakness. These signs may become visible long
before issuers see problems surfacing in their small business or commercial
borrower financial statements.
Transaction scoring is not an alternative to models based on payment history.
Rather, it is a complement to traditional risk models and can provide a
substantial and verifiable lift to traditional portfolio management. Several
examples are outlined in Figure 3.
The key benefit that transaction
scoring provides is visibility into
account activity. For example,
by using transaction scoring in
tandem with aggregate consumer
transaction behavior, issuers can
develop segmentation strategies
based on real industry movement
in terms of consumer merchant
5TRANSACTION SCORING: WHERE RISK MEETS OPPORTUNITY
figure 3: Transaction-Based Applications Enable Issuers to Toughen Risk
Detection and Maximize Revenue
Segment Decision Areas Actions
• Product selection
(e.g., affluent, cash,
• Decision criteria
• Credit limit
• Pricing assignment
• Segment new accounts more finely
(e.g., distinguish potential revolvers/
transactors from customers who
would be well-suited to variations on
a spend versus revolve product)
• Develop pricing and limit-assignment
strategies based on historical
spending and payment behavior,
(e.g., fees vs. APR for transactors and
• Credit-limit strategy
• Collections strategy
• Adjust over-limit pads based on
changes in spending patterns
(particularly helpful for No Preset
Spending Limit, charge, and
• Detect rule and policy violations by
commercial account holders (e.g.,
spending in restricted Merchant
• Institute proactive and reactive
limit changes (e.g., customers with
frequent purchases and high revolve
rates may be targeted for credit limit
• Create micro-segments based on
spending categories to align with
• Implement trigger-based collections-
block strategies so that customers
who continue to attempt purchases
while delinquent will be immediately
Source: MasterCard Advisors.
Transaction scoring provides an added dimension—a view into the behavioral
patterns of individual consumers on a near real-time basis as compared with
the lag time built into traditional credit report data. By creating account-level
variables, transaction scoring greatly increases the accuracy and timeliness of
existing models, as shown in Figure 4.
Transaction scoring provides an
added dimension—a view into the
behavioral patterns of individual
consumers on a near real-time
basis as compared with the lag
time built into traditional credit
report data. By creating account-
level variables, transaction scoring
greatly increases the accuracy and
timeliness of existing models.
6 ADVANCING INSIGHTS ADVANCING COMMERCE.
FIGURE 4: Transaction Elements Are Combined to Create Account-Level
Variables that Capture Purchase Behavior Over Time
combinations of transaction
Capture all transaction
dimensions at industry
and merchant levels.
• Date and Time
• Unique ID
• Merchant Industry
• Transaction Type
• Transaction Flags
• Smoothed Time
• Target Weighted
• Customer Activities
models to create
scores in real time.
Source: MasterCard Advisors.
Transaction scoring completes the risk-management equation by providing an
opportunity to increase revenue while decreasing exposure. When the economy
was growing, simply reducing risk was often enough. But in an environment
characterized by contraction and caution on the part of both issuers and their
customers, more finely tuned risk management is required.
The advantages and quantifiable benefits of transaction scoring, both from cost
reduction and revenue enhancement points of view, are outlined in Figure 5.
Transaction scoring completes
the risk-management equation
by providing an opportunity to
increase revenue while decreasing
exposure. When the economy
was growing, simply reducing
risk was often enough. But in an
environment characterized by
contraction and caution on the
part of both issuers and their
customers, more finely tuned risk
management is required.
7TRANSACTION SCORING: WHERE RISK MEETS OPPORTUNITY
figure 5: Transaction Scoring Can Increase Revenue and Reduce Losses
Behavioral Scoring • More accurately predict risk
(identify more “bads”)
• Identify risk earlier,
affording more time to take
• Increase marketable
universe/ weed out false
• Better predict where
strategies are optimal
• Optimize contact
rate through better
understanding of behaviors
• Focus costly outbound
programs on likely
• Increase marketable
universe/ weed out false
positives (i.e., high-
Fraud Modeling • Optimize real-time
• Optimize transaction
scoring for fraud prevention
• Increase authorization
approvals vs. false positives
• Reduce marketing expense
to recruit new cardholders
• Increase response rate and/
or lifetime profitability
Source: MasterCard Advisors.
Quantifying the Benefits of Transaction Scoring
Leveraging cardholder information from large U.S. issuers, account transaction
data from individual consumers can identify behavior patterns and detect
changes in these patterns over time. Building on the results of the data, the
algorithms become complementary to the issuer’s traditional credit decisioning
practices and portfolio management techniques, significantly enhancing lift as
measured by a variety of metrics.
MasterCard Advisors’ Financial Stress Propensity Model provides a powerful
example. This model is able to identify specific patterns of behavior in accounts
that have never been delinquent, but are likely to progress to delinquency within
12 months. In a recent application, the model’s transaction-based algorithms
were used in conjunction with existing models to deliver lifts of 10 to 15
percent in Kolmogorov-Smirnov (KS) tests over behavior scores not employing
the transaction-based algorithm as an overlay.8
Broadly speaking, the segment
analyzed comprised accounts that not only were never delinquent, but were
high active transactors with mid FICO scores. The results are shown in Figure 6.
information from large U.S.
issuers, account transaction data
from individual consumers can
identify behavior patterns and
detect changes in these patterns
over time. Building on the results
of the data, the algorithms
to the issuer’s traditional
credit decisioning practices
and portfolio management
enhancing lift as measured
by a variety of metrics.
8 ADVANCING INSIGHTS ADVANCING COMMERCE.
figure 6: Financial Stress Model Enables Issuer to Significantly Reduce
Delinquent Balances in a key segment
Applying the model to High active transactors, mid fico range
Percent more “bads” identified 3.32%
Average line increase $800
Average utilization of incremental line granted 69.3%
Total delinquent balances avoided $3-7MM
Lift over internal behavioral score 12%
Source: Proprietary MasterCard Advisors analysis with major U.S.-based issuer.
Advisors’ transaction scoring capabilities have achieved similar results in early
month on book (EMOB) applications. A recent project focused on an issuer’s
accounts that were six months or newer and at risk of charge-off in 12 months.
By using transaction-based algorithms in combination with existing models,
the issuer realized an increase in the range of 13 to 17 percent in KS over the
historical score alone.
By providing a much more robust model, as measured on a KS score, transaction
scoring allows a more consumer-centric treatment of adverse actions. Leveraging
the positive and neutral behaviors isolated by transaction scoring, whether or
not they were potential false positives, issuers can more accurately segment their
portfolios based on what consumers are actually doing at the point of sale.
BUILDING ON BEST PRACTICES
Transaction scoring has been used for years in the U.S. credit card business in a
variety of applications, as reported in the Federal Reserve’s 2010 study of 100
While transaction data has been used since the mid-1990s in fraud risk
applications, bankruptcy scoring is probably the most widespread credit risk
application of transaction scoring at present. Specific cardholder actions (e.g.,
cash advances at casinos) can trigger a bankruptcy overlay of the historical
scoring to which the issuer regularly subjects the account. But an overlay is
based on discrete events rather than identifiable patterns and thus is not a
transactional score as defined in this paper. A robust variable-based transaction
scoring system, which looks at a series of events and the patterns that emerge,
can more finely segment risk into immediate or prospective. By enabling a more
nuanced decision tree to guide actions (rather than a unitary “red light,” as
exemplified by the cash advance), transaction scoring gives the issuer greater
flexibility in responding to the series of events. Only some of these actions
may be linked to an adverse action code. This increases the likelihood that the
action, whatever form it takes, can preserve on a timely basis the viability and
future profitability of many more accounts.
By providing a much more robust
model, as measured on a KS
score, transaction scoring allows a
more consumer-centric treatment
of adverse actions. Leveraging
the positive and neutral
behaviors isolated by transaction
scoring, whether or not they
were potential false positives,
issuers can more accurately
segment their portfolios based
on what consumers are actually
doing at the point of sale.
9TRANSACTION SCORING: WHERE RISK MEETS OPPORTUNITY
Having extended credit to the consumer, an issuer could be just as interested in
guiding that consumer into more judicious paths as in curtailing the cardholder’s
access to credit. By allowing the creation of more narrowly defined segments of
risk, as well as of opportunity, transaction scoring fosters a new, positive model
of bank/consumer relationships, and allows issuers to look at risk management
and marketing as two aspects of a unified account management process.
Benefiting CONSUMERS AS WELL AS ISSUERS
Transaction scoring enables banks to screen changes in an individual customer’s
behavior patterns. This information gives the bank the means to help the
customer by demonstrating that he or she, perhaps unknowingly, is going
down a path beset with hazards. Transaction scoring helps put the element of
time on the side of both the consumer and the issuer. With account scoring
methods based exclusively on historical data, the destructive pattern of behavior
may be detected when it is simply too late to intervene, resulting in a burden of
debt that is likely negative for the institution and the cardholder.
From the consumer’s point of view, this time lag often causes huge problems.
If the consumer has begun to accumulate debt to the extent that his or her
household’s financial position and future—short- or long-term—is affected, it
could take months or even years before the consumer can pay his or her way
out of trouble. Transaction scoring makes it possible for the bank to intervene
before the consumer is too far down the road to delinquency, charge-off,
bankruptcy, or balance levels that preclude further card use.
Transaction scoring also allows the issuer to continue to afford the consumer
access to liquidity at the point of sale over the long term. In an environment in
which some consumers are being forced, through a variety of factors, out of
the banking system altogether and into the arms of payday lenders, rent-to-
own plans, and others, access to liquidity enables banks to provide a much-
needed social benefit that was not possible before.
The U.S. economy’s larger problems, as well as those of the banking business,
result from a lack of liquidity. By definition, a simple rescission of credit will
do nothing to solve this problem and may prolong its effects. But issuers,
working together with consumers, can create a healthier environment in which
the benefits of responsible consumer spending combine with a reasoned
extension of credit. Proper management of the bank/consumer relationship
using transaction scoring can encourage a more financially educated consumer,
whose access to credit money at the point of sale would benefit the consumer
and his or her family, as well as the larger economy—including merchants—
and the banking system as a whole.
the Semantic issue
The concept of adverse action,
or rather the words themselves,
deserve some discussion.
“Adverse action” is a term of
art. But taken as part of a larger
toolset of customer engagement
strategies, issuers can turn
adverse action into a positive
event for consumers, as well as
for the bank.
Obviously, no consumer wants to
entertain the prospect of having
his or her card declined at the
point of sale. But just as in the case
of a compromised card, the decline
is, in fact, in the cardholder’s best
interests, so adverse action, or
contacting the consumer before
adverse action is necessary, can
initiate dialogue on better and
sustainable use of the card.
The new ability to provide mobile
alerts to customers at the point
of sale is one example of how
this dialogue could begin.
Obviously, in certain extreme
cases, the issuer may not have
an alternative to immediate
drastic action—declining the
card or capping at balance. But
if the issuer keeps the lines
of communication open with
the cardholder, the prospect
of “adverse action” may be
an opportunity for consumer
education, credit repair, or even
cross- or up-sell.
Solutions could include different
products, additional products
(debit, prepaid, other lending
vehicles), or help in financial
management and spending
control with scenario planning,
invitations to formal or informal
credit counseling, or credit repair
10 ADVANCING INSIGHTS ADVANCING COMMERCE.
TRANSACTION SCORING, THE NEW BUSINESS MODEL
The U.S. consumer is not the only entity that has changed since 2008. The
transformation in banks’ business models has also been radical. One of the
key tools issuers used to grow their business from 1992 to 2008 was the
securitization of credit card assets. According to the Fed, pools of securitized
assets declined from $661 billion to $131 billion between 2006 and 2010,
a compound annual decline rate of 50 percent.9
Changes in rules from the
Financial Accounting Standards Board (FASB) are part of the reason; risk
aversion on the part of investors is another. But whatever the cause, U.S. credit
card issuers are again being forced, after more than 20 years, to lend from their
By providing a more nuanced and finely grained risk management framework,
transaction scoring can facilitate a greater level of compliance with Basel II
and Basel III in terms of operational and credit-risk management and capital
In for a Dime, in for a Dollar
Banks contemplating the adoption of transaction scoring should understand
that such a decision has large-scale implications for their operations going
forward. The adoption of transaction scoring is, in fact, the tip of the iceberg,
as Figure 7 shows.
FIGURE 7: transaction scoring mandates an enhanced operational
and Channel Management
and Information Provisioning
Core Account Processing
Source: MasterCard Advisors.
Banks contemplating the
adoption of transaction-based
scoring should understand that
such a decision has large-scale
implications for their operations
11TRANSACTION SCORING: WHERE RISK MEETS OPPORTUNITY
Transaction scoring is not a quick fix. Adopting transaction scoring in a way that
realizes the technique’s potential may require from-the-ground-up retooling
of the issuer’s operations, not simply a refurbishing of the risk management
systems. Affected areas may include:
• Core account processing
• Database management
• Customer Relationship Management (CRM)
• Channel management (mail, in- and out-bound telephone, email, mobile)
• Collections and recovery
The issuer must create or retool its existing data management and customer-
touchpoint systems to ensure that the results of transaction scoring are
integrated into the bank’s larger strategy, product design and implementation,
marketing, and electronic payments platform. This becomes especially
important if the bank has multiple relationships with the consumer, including
While by no means transparent, adopting transaction scoring does not have to be
disruptive, or even onerous. It builds, or ought to build, on efforts that progressive
institutions and their processors and partners have been working on for years;
these likely include database management in the interest of superior customer
service, Internet and mobile functionality, and modular core account processing.
Leading institutions are already starting to leverage the power of transaction data
to their advantage. Banks that fail to move in this direction will find themselves
employing out-of-date risk management tools that are incapable of reflecting the
near real-time requirements of decisioning in the new environment.
Widespread adoption of transaction scoring will result not only in a more
competitive playing field among financial institutions, but also a more fruitful
lending environment for consumers.
1 Proprietary MasterCard Advisors analysis with major U.S.-based issuer.
2 Source Media, “Card Industry Directory,” 20th Edition, 2008.
3 Board of Governors of The Federal Reserve System, “Report to the Congress on the
Profitability of Credit Card Operations of Depository Institutions,” June 2010.
4 Equifax, April 2011.
5 Nilson Reports #965 and #942, U.S. Census Bureau.
7 Equifax, April 2011.
8 Proprietary MasterCard Advisors analysis with major U.S.-based issuer.
9 Federal Reserve Statistics Release, “Consumer Credit G.19,” June 2011.
While by no means transparent,
adopting transaction scoring does
not have to be disruptive, or even
onerous. It builds, or ought to
build, on efforts that progressive
institutions and their processors
and partners have been working
on for years; these likely include
database management in the
interest of superior customer
service, Internet and mobile
functionality, and modular core