Extracted from 'Principles and Practice of Consumer Credit Risk Management' 3rd Edition 2008 by Helen McNab and co-author Peter Taylor ISBN 978-1-84516-750-9 enhanced with definitions from ScorePlus seminars.
Please note terminology and institutional references are UK based.
A
Wide-ranging operational function dealing with customers during the active life of their account(s), i.e. while their account(s) are in order.
Astatistically-derived behavioural scorecard(s) developed for predicting the future risk on accounts that are currently in order; generally used for setting credit, shadow and target limits, and for authorizations decisions.
An organization that recruits a retailer who then accepts credit card transactions. Usually a bank.
Action Fraud is the UK’s national reporting centre for fraud and cyber crime. actionfraud.police.uk/what-is-national-fraud-intelligence-bureau
The proportion of the outstanding balance of a product that the lender expects will not be repaid. Used in provision estimation.
The information available about an applicant that is taken from the application form.
Operational function dedicated to the timely processing of new applications in line with company lending policies.
A computer system or software, which automates all stages of the credit review process. It can include validation checks, policy rules and on-line credit referencing. It may be run on a mainframe or on personal computers (either singly or networked).
The score calculated at the time of application. Some organizations use this term for the score based on the application data alone (i.e. excluding scorecard characteristics that use credit bureau data).
A statistically-derived model that predicts the credit risk for applications if they are accepted. It comprises a scorecard and a set of associated statistics for interpreting score in terms of risk. The traditional term for this was credit scoring.
A table derived from a scorecard development, which shows, for any score or scoreband, the cumulative percentage of applicants expected to achieve that score or lower.
An item of information about an applicant or an account holder. A range or set of these items is interrelated to form a characteristic. For example, ‘tenant’ is one attribute of the characteristic ‘residential status’.
The measure of an individual attribute’s ability to discriminate between any two outcomes (usually good and bad).
A form of behaviour scoring in which the predicted outcome is the likelihood of an account becoming dormant or closing in the near future. Used by marketing departments to retain good customers.
The process whereby the credit card issuer permits, or declines to permit, a new cardholder transaction.
B
British Bankers Association. The trade body for the banking and financial services industry. From 2017 merged into UK Finance. ukfinance.org.uk
The percentage of accounts that perform in an unsatisfactory manner as defined by the good/bad definition that was used in the scorecard development.
Accounts with unacceptable repayment behaviour. The credit grantor is unlikely to transact any future business with these account holders.
A statistically derived model for predicting a specific dimension of future behaviour on an existing account. Models can be designed to predict, for example, risk, attrition, churn, propensity, response and revenue. The score is recalculated regularly (typically monthly) and can be used to make decisions on marketing, authorizations, limit setting and debt collection/recoveries activities.
Industry group that seeks to set benchmarks of best practice/performance for a given industry/sector.
The ratio of the number of good accounts needed to balance the losses generated by one bad account.
C
A UK proprietary payment history system used for sharing financial account information. The acronym stands for ‘Credit Account Information Sharing’.
County Court Judgments – a court order to a debtor in serious default that he or she must repay the debt. An English term whose equivalent elsewhere is Judgments, or in Scotland is Decree.
A non-competitive, industry-wide information exchange system to help members identify fraudulent applications.
Civil Procedures Rules. The definitive book on court procedures. justice.gov.uk/courts/procedure-rules
The means by which the presenter of the card may be identified as genuine, for example a signature or PIN.
Debtors who have defaulted on payments because of their financial inability rather than any intent not to pay.
Organizations that manage and control the operation and clearing of plastic card transactions. Issuers must be member s of the appropriate schemes to issue cards and acquire merchants. Examples are Visa, MasterCard/Europay and American Express.
The last three or four digits of a number printed on the signature panel on the reverse of payment cards.
An experimental set of two or more competing strategies designed to determine the extent to which a new strategy (challenger) can improve an outcome or performance, when it is tested against the existing (champion) strategy. The experimental strategy is applied in a single decision area, e.g. collections, and can be repeated any number of times with new challengers.
Any data item that could appear in a scorecard. These can be derived from the application form, internal files or credit references. For example, applicant’s age or type of vehicle being purchased.
A regular statistical report that compares the current attribute distribution of a particular characteristic with its corresponding distribution in the development sample. It shows changes in applicant profiles over time. It can also be used for non-scored characteristics.
A measure of a characteristic’s ability to discriminate between two outcomes (usually good and bad). This is the maximum contribution that the characteristic can make to the scorecard.
A credit card with an embedded microchip which sores secure data for the purposes of identity validation and transactions authorization.
The propensity for, and the rate at which, account holders leave one lender to take up a better offer with a competitor.
The overlap in the predictive content of two or more data items. For example, time in job, time at address and time with bank all predict risk in the same way - the longer the time, the better the risk. Individually each data item may be predictive. Used in combination the predictive power of the three together will be less than the sum of the predictive power of the three data items when analysed as separate data items. Once a data item is used in a model, other data items may contribute marginal additional information. Model building reflects these overlaps of information. Also known as interaction or correlation.
A group of items sharing a common quality, e.g. all accounts opened in the same month. Also known as a tranche.
Operational function dedicated to the process of managing accounts in the early stages of arrears/non-payment, or in excess.
A statistically derived behavioural scorecard for predicting future risk for accounts in the early stages of collections.
A loan offered as a means of amalgamating a number of other, smaller loans (usually from multiple lenders), particularly where the customer may be in arrears. The loan principal is generally paid out to the other lenders.
A commercial organization that collects and supplies credit-related information about individuals and companies. Also known as Credit Reference Agency.
The total number of points gained when characteristics only available from a credit reference are scored. This can be generated in-house as part of the credit scoring process or it can be a generic score that has been generated by a bureau based on all the information it holds on a customer.
The information resulting from an enquiry on a credit bureau’s files. In the UK this includes judgments, voters roll, previous search and shared financial/non-financial account information.
Strategic function responsible for optimizing credit losses versus business development. Sometimes known as credit strategy or risk management.
The sum of the points allocated by the scorecard to the attributes given on an application form plus the answers retrieved from a credit bureau search.
A term for a statistically derived model for predicting credit risk. It comprises a scorecard and a set of associated statistics for interpreting score in terms of risk. Or, the process of using a scorecard for assessing credit risk.
Developing and applying scorecards at the customer level instead of at an individual product/account level. It takes into account all the products held by the customer, and is used to set an overall credit limit that can be apportioned between different products.
The score that represents the boundary between accepted and rejected applications, i.e. the lowest acceptable score. This figure is determined by the credit grantor and is adjustable.
A plan that determines the cut-off score. This is usually decided on the basis of volume of accepted accounts required, or acceptable bad rate, or a combination of these. It varies for different products, different scorecards and over time.
A four-weekly or monthly date at which accounts are updated with financial transactions – payments, fees, interest, purchases, cash advances. For credit cards, statements are produced at this point.
D
The raw ingredients (usually facts or figures) that, when processed, produce information or scorecards or management information reports.
The process of collecting data from application forms or other paper-based sources, and transferring it to computer media. This may involve keying or scanning.
Operationally, a term generally used in Recoveries management. Under Basel, default is defined as the obligor being past due more than 90 days on any material credit obligation, with the delinquent days counter based on "days out of order", or the lender has made an account-specific provision.
A notice that must be sent to customers in arrears before litigation or other recoveries actions can be taken to recover the debt. Once the notice has been served the lender has the right to demand repayment of the full balance outstanding rather than only the arrears outstanding.
A table derived from a scorecard development, which shows, for any score or scoreband, the cumulative percentage of applicants expected to achieve that score or higher.
The ability of a scorecard to differentiate between two polarized outcomes (or two potential outcomes) – usually goods and bads.
The pattern of observed values in a dataset. For example, score distribution is the number of occurrences of each score in a sample.
The extent to which a scorecard can discriminate between potentially good and bad accounts based on average scores of goods and bads, taking into account the variance around the average scores.
A regular report that monitors delinquency levels by score for a set of accounts which have been open for approximately the same length of time (i.e. they have the same exposure period). The reports are typically produced monthly (sometimes quarterly). It provides the source data for the dynamic performance cohort matrix.
A report that is regularly updated to monitor the growth in the delinquency levels for a set of accounts which have been open for approximately the same length of time (i.e. they have the same exposure period). The reports are typically updated monthly (sometimes quarterly).
E
Electronic purse is a prepaid card that contains an electronic value which can be exchanged for goods and services.
A computerized fraud prevention technique in which data transmitted over a network is scrambled before transmission and unscrambled on receipt.
Credit card accounts where the current balance exceeds the declared limit. Also called over-limit.
Applications/accounts that are not included in a development sample because either the applicant or the account behaviour patterns are atypical, or because there is no intention to score such applications in the future. Examples include staff, VIP, deceased, stolen, or fraudulent applications/accounts.
Expected loss (EL) is the estimated loss from credit commitments. It is used for provision (estimated loss from current commitments) and as a component part of the regulatory capital calculation (estimated loss arising over the next 12 months). Formula: PD Probability of default X EAD Exposure at default X LGD Loss given default. Also known as Expected Credit Loss (ECL).
When the balance on a credit card is only partly repaid each month, the cardholder has taken extended credit.
F
A credit score that includes application data and all other, internal or external, data.
A statistical report that relates the final score distribution for a particular set of applications to the accept/reject decision.
A financial ceiling, allocated by the merchant acquirer, given to retailers who accept credit cards; all transactions above this limit must be authorized. It is a credit and fraud control measure.
Uses a scorecard(s) to identify potentially fraudulent credit card transactions and/or applications.
Credit cardholders who pay their balance off in full each month, thereby incurring no interest.
G
Gone Away Information Network. A non-competitive, industry-wide shared information scheme for identifying and tracing absconders who are in arrears.
A new characteristic created from two or more characteristics. For instance, ‘loan-to-value’ is a mortgage scorecard characteristic generated from the property value and the amount of the loan requested.
These are generalized risk models, developed and sold by outside vendors, based on information held by the vendor. The information used in such models is generally bureau-based and includes public records, shared-account information, bureau searches and geo-demographic indicators.
Classification of personal data based on various demographic criteria, e.g. census, voters roll, financial or neighbourhood data.
Precise commercial definitions of good and bad accounts. This includes definitions for indeterminates and exclusions.
Accounts with acceptable repayment behaviour patterns. The credit grantor would wish to transact future business with these account-holders.
Attributes of similar strength are amalgamated into larger groups. This reduces the total number of attributes and ensures that the sample count for each one is large enough to be statistically sound.
H
I
3 stage framework for calculating provision. Stage 1 accounts have a 12 month assessment. Generally, non delinquent debt. Stage 2, where risk has increased, lifetime assessment calculated, interest bearing. Generally, early stage delinquent debt. Stage 3 lifetime assessment, non-interest bearing. Generally, post-notice of default debt. Successor to IAS 39.
Insolvency Practitioner - Person authorised to act as the office holder in an insolvency proceeding.
Internal ratings based approach – within the Basel II Accord framework, the use of own company estimates of customer behaviour as inputs to the calculation of Expected Loss.
An Individual Voluntary Arrangement - a formal agreement between a debtor and their creditors; generally enables debtors to make reduced payments towards the total amount of debt, to pay off a percentage of what is owed. Generally after 5 years any remaining debt is cancelled.
Those accounts that fall between the definitions of good and bad. They may have a mildly delinquent history or not have sufficient maturity to be clearly classified. Ideally these should be no more than 5% of the development sample.
Industry Hot Card File. Computerised list of lost or stolen cards, maintained by the banking indistry, used by merchants in fraud prevention.
A table of scoring results that shows, for each score, the percentage of applicants which can be expected to achieve any particular score, together with the bad rate and marginal bad rate.
An organization that recruits cardholders and issues the plastic card. Usually but not always a financial institution.
J
A legally binding decision handed down by a court of law requiring a debtor to pay the creditor and the means by which this is to be achieved. Abbreviated as CCJ.
K
L
LGD stands for Loss given Default. Under Basel Accord, LGD is defined as % exposure at default which will be lost. Estimated LGD should be conservative and “adapted to economic downturn”. There are minimum floor values which vary by product (and country) e.g. 25% to 50% for unsecured credit; 0% to 15% for secured credit.
Parameters for LGD have changed over time. These definitions are correct for the UK as at February 2022.
Known as LGD. Under Basel Accord, LGD is defined as % exposure at default which will be lost. Estimated LGD should be conservative and “adapted to economic downturn”. There are minimum floor values which vary by product (and country) e.g. 25% to 50% for unsecured credit; 0% to 15% for secured credit.
Parameters for LGD have changed over time. These definitions are correct for the UK as at February 2022.
M
A statistical method that can be used for estimating provision levels based on generalized roll rates.
Maximum unexpected loss is the worst level of “conceivable” losses. Under the Basel II framework, a factor is applied to the Expected Loss. This varies by product type. The key idea is for regulated lenders to hold enough capital to absorb the worst level of “conceivable” losses “TTC” (though the cycle).
Representation of the future based on past experience, generally derived through statistical analysis.
The illegal process of passing illicit funds through an account (or several accounts) to disguise the original source of the funds.
The business function of producing regular management reports in order to control the scorecard(s) and the lending functions. Industry-standard reports are population stability, characteristic analysis, override analysis, final score report and dynamic delinquency reports. Also known as tracking.
N
National Fraud Intelligence Bureau (NFIB). A rapid response unit based at APACS that shares information between banks and the police for combating fraud. actionfraud.police.uk/what-is-national-fraud-intelligence-bureau.
Modelling technique based on artificial intelligence or the network of nerve cell connections in the brain.
O
Optical Character Recognition – software that scanners use to convert images into characters or digital files.
The probability that a specific event will occur expressed as a ratio of one event to another, e.g. good/bad odds.
In a scorecard development, the time period between the date the sample data is extracted and the date the sample is classified into the principal sets.
Credit card accounts where the current balance exceeds the declared limit. Also called excess.
A regular statistical report that relates, by score-bands, the number of overrides occurring.
A decision that contradicts the scored recommendation, e.g. an accept below the cut-off or a reject above the cut-off.
P
PD stands for Probability of Default. Under Basel Accord, PD is defined as the probability of an account being 90-days arrears over 1 year, or specific provision raised. Delinquent days counter based on consecutive "days out of order".
Estimated PD should be "represent a conservative view of long-run average PD" and "be grounded in historical experience and empirical evidence". "Long-run" is not defined but is assumed to be "economic cycle".
There is a floor minimum of 0.03%.
Parameters for PD have changed over time. These definitions are correct for the UK as at February 2022.
Fraudster method for eliciting information from genuine account holders, by re-directing internet traffic.
Fraudster method for eliciting information from genuine account holders, by pretending to be the lender, generally by email.
Any rule that is applied in addition to the scoring process in order to arrive at an accept/reject decision, e.g. decline all applicants aged under 18.
All the applicants who have applied for a particular credit product, irrespective of whether they are ultimately accepted or rejected.
A statistical report that reflects changes in a population by score-band.
Telecommunications equipment that stores future outbound telephoning requirements, automatically dials telephone numbers and presents connected calls to an operator. Also known as a predictive dialler.
The process of cleaning a mailing list by removing any entries that do not meet a predefined set of conditions, e.g. the existence of CCJs.
A scorecard used to identify accounts that are likely to be most suitable for a mailshot.
The two groups used in a two-outcome statistical modelling process, e.g. good/bad or active/inactive.
Known as PD. Under Basel Accord, PD is defined as the probability of an account being 90-days arrears over 1 year, or specific provision raised. Delinquent days counter based on consecutive "days out of order".
Estimated PD should be "represent a conservative view of long-run average PD" and "be grounded in historical experience and empirical evidence". "Long-run" is not defined but is assumed to be "economic cycle".
There is a floor minimum of 0.03%.
Parameters for PD have changed over time. These definitions are correct for the UK as at February 2022.
Offers made by customers whose accounts are in arrears to repay some or all of the debt by an agreed date and in agreed instalments.
Marketing scorecard that predicts the likelihood (or propensity) of an existing customer taking more credit/another loan, etc.
An accounting measure for the amount of outstanding balances that a lender expects will not be repaid.
R
RWAs are the lender's assets (loans) weighted by the underlying risk. The RWA value determines how much capital a regulated lender must hold.
Operations function dedicated to the collection of debts post-notice of default, where the full outstanding balance has become payable.
The illegal practice of declining an applicant for credit solely on the basis of his or her address.
In the Basel II Accord framework, the amount of capital to be held, to cover Maximum Unexpected Losses 99.9% of the time.
A credit card that has been routinely produced to replace an expired card. See also replacement plastic.
The process of deducing how a rejected applicant would have behaved had he or she been granted the credit. The resultant data is then included in the scorecard modelling process.
A credit card that has been issued as a one-off event to cover the loss or theft of an existing card. See also reissued plastic.
The process of using a scorecard to predict the likelihood of targets responding to a mailshot. If the likelihood is too low, they are removed from the list before mailing.
Card account-holders who do not pay off the full balance each month. See also Transacters.
A form of consumer credit typified by credit, debit and charge cards, in which the outstanding debit balance may revolve from one payment period to the next.
A strategy in which the interest rate or other terms are set in line with the perceived risk.
The percentage of accounts at any level of arrears that progresses, or rolls on, to the next (worse) level of arrears.
S
A one-dimensional summary of a particular aspect of customer behaviour, based on the information known about that customer at a specific point in time.
A set of questions, called characteristics, with a set of answers, called attributes. Each attribute has a different score value. For any given customer, the attribute scores are summed to give a total score that is interpreted using the scorecard statistics.
An undeclared credit limit, which is usually greater than the declared limit, to which the lender will extend credit on a particular product.
An agreed amount, which is less than the full amount due, that the debtor will pay, and the creditor will accept, in full repayment of the debt.
A form of fraud where a card’s magnetic stripe details are electronically copied and use to create a counterfeit card.
Within the Basel II Accord framework, the broad grouping of debt categories for which a ‘standard’ level of capital is required.
Proprietary systems for implementing account management strategy design and testing for application or behavioural scoring.
Any sub-group within the main population that is sufficiently different from the rest of the population to justify being identified separately.
Those applications that would have been accepted under the previous decision process, but would be rejected by the implementation of a new scorecard, and vice-versa.
T
The maximum amount that a lender is prepared to extend to a customer across all his or her product holdings. There is no tolerance in this limit.
All the applicants for any particular credit or financial product irrespective of the final accept/reject decision. Sometimes abbreviated to TTD population.
A means of illustrating where different activities occur in relation to each other over time.
Operational function dedicated to finding debtors who have absconded without leaving a forwarding address or contact point.
Gives the likelihood of an account in any one state at one point in time moving into any other state at a second point in time.
U
The process of assessing the risk of any request for credit and reaching an accept or decline decision based on that assessment.
Unexpected loss is the difference between the actual and expected loss. This can be positive or negative. See also Expected Loss and Maximum Unexpected Loss.
In the Basel II Accord framework, this is the ‘worst case’ estimate of losses – it determines the amount of regulatory capital to be held.
V
The process of testing the final scorecard before delivery. The supplier undertakes this before delivery, and the user should repeat the process before the scorecard is implemented.
A small percentage of a scorecard development sample (typically 20%) which is excluded from the scorecard modelling process so that it can be used to test the final scorecard.
A government list of all the adults who have registered their eligibility to vote in national or regional elections. Also known as electoral roll.
