Articles and Whitepapers on Collection & Recovery

Transcription

1 Collection Scoring This article explores the scoring technologies utilised for defaulting accounts. Best practice collection strategies apply the most appropriate scoring technology, depending on the status and delinquency level of an account. In order to create a cost-effective and robust delinquency management strategy an organisation should: 1. Segment accounts according to their delinquency stage within the credit life cycle, namely: Early Stage typically 1 or 2 instalments in arrears Late Stage typically 3 or more instalments in arrears Write-off Stage account balance has been moved to the organisations profit & loss statement 2. Profile accounts according to their behaviour patterns, using one of the following scoring technologies: Risk Scores predicts the probability of mildly delinquent accounts becoming highly delinquent Recovery Scores predicts the payment ability of highly delinquent accounts 3. Apply the appropriate collection tactic, using the profile to drive the collection action according to: Type - letter, telephone call, SMS, handover, etc Tone - severity, content, etc Timing - immediate, delayed, etc Date: August, 2010 Compiled by Profitera Corp. Page 1 of 9

2 The objective for any organisation is to cure accounts that are in an early stage of delinquency (mildly delinquent). The scoring technology applied is a behaviour risk score, which predicts the probability of an account becoming highly delinquent. Typically early stage delinquency is defined as an account having missed 1 or 2 instalments, however the definition will vary according to the credit product in question (bank card, personal loan, retail card, mortgage bond, etc) and an organisations view on what a mildly delinquent account is. Late stage delinquency (highly delinquent) is typically defined as an account being 3 or more instalments in arrears. Organisations generally treat these accounts as bad and the objective is to recover as much of the outstanding balance in the most cost-effective manner. The scoring technology used is a behaviour recovery score, which predicts the likelihood of an account repaying the outstanding balance. A credit bureau risk score can be used instead of, or in conjunction with, an internal risk score. The latter is achieved by creating a cross-matrix between the internal risk score and the external credit bureau score in order to enhance risk prediction capabilities. The process of applying this technique is discussed later in this article. Premise of Scoring Scoring is based on the premise that the future is like the past. Scorecards are based on the relationship between historical data elements and an accounts subsequent performance. Significant economic turmoil, customer behaviour shifts, etc will result in changes over time, which a scorecard cannot cater for. For this reason, the application of the scorecard and the impact of changes must be monitored on a regular basis. Where significant changes occur and the scorecard is no longer effective, the scorecard will have to be redeveloped. On the whole, however, the future is more or less like the past and scorecards are consequently robust over a number of years. In more changeable economies or industries, scorecards may often require realignment or rebuilding. Predictive Modelling Predictive modelling attempts to establish rules (scorecard) to guess (predict) a future outcome ( bad status) using historical variables (characteristics). A scorecard is therefore risk or recovery assessment tool and assigns a score to an account on a regular basis, typically monthly. The score provides an objective view on a customer s future risk or ability to repay an outstanding balance. A behaviour scorecard is a set of characteristics for which each value (attribute) attracts a weight (points). The sum of the weights attained for each of the characteristics constitutes the overall score, which translates into a particular level of risk or recovery propensity. Benefits of Scoring There are many benefits to using scoring within decision-making. Scoring can be easily automated so that it occurs quickly and consistently. The assessment of risk or recoverability is objective and Date: August, 2010 Compiled by Profitera Corp. Page 2 of 9

3 unbiased. Given the same attribute combination, the same score will be allocated, ensuring that the decision will be the same regardless of which branch the customer visits. Scores allow for a high level of management control, as decisions are made based on quantifiable risk or payment ability. The statistics from a scorecard development indicate the risk or payment ability at each score, and consistent collection strategies can be based on this. Performance Definitions Every scorecard has a good/bad performance definition. This is the definition of what constitutes a good account and what constitutes a bad account. Typically the good definition for a behaviour risk scorecard is no delinquency or only very mild delinquency during the period that the account is being measured in. A bad definition is typically a serious level of delinquency during the performance period. The level of delinquency used to determine a bad customer depends on the local conditions and target market of the specific portfolio. An example of a bad definition for a behaviour scorecard is 3 cycles delinquent (missed 3 required instalments) during the performance or measurement period. An indeterminate is an account that falls between the good and bad definition. These are removed from the scorecard development to ensure that a scorecard can be built to differentiate more strongly between good and bad. Behaviour Risk Score A behaviour risk score predicts how an existing account will perform on a specific portfolio. In terms of the behaviour risk scorecard performance definition, the score will predict the likelihood that an account will reach the bad definition within the next 6 months. This knowledge allows an organisation to use the score to make risk-based decisions on each account, for example how harshly or leniently the account will be treated within the collection department. MAXIMUM DELINQUENCY MONTHS SINCE DELINQUENCY > Date: August, 2010 Compiled by Profitera Corp. Page 3 of 9

4 MONTHS ON BOOKS This is an example of a behaviour risk scorecard. Typically these scorecards contain between 8-10 characteristics. The above characteristics are typical of those found in any risk scorecard. The characteristics can be defined as follows: Maximum Delinquency during a particular period in the past shows the extent of the delinquency on the account during the most recent history. It is intuitive that the greater the delinquency level, the fewer points that should be allocated. Months Since Delinquency > 1 means the number of months since the delinquency level on the account was greater than 1. Intuitively the longer the period since delinquency has been >1, the better the account is and the more points should be allocated. Months on Books is how long the account has been on the specific portfolio. It is an established fact that the bad rate of accounts increases until a certain point at which it levels off. This is because within the immature (new) accounts, the accounts that are going to go bad will tend to go bad fairly quickly. More points are therefore allocated to older, more mature accounts. Behaviour Recovery Score A behaviour recovery score predicts how likely a highly delinquent account will repay in the future. Typically a recovery score relates to the percentage of outstanding debt that is likely to be repaid over a specified period of time. Here is an example of a behaviour recovery scorecard. MONTHS SINCE LAST PAYMENT MONTHS SINCE UP-TO-DATE Date: August, 2010 Compiled by Profitera Corp. Page 4 of 9

5 CURRENT DELINQUENCY Months Since Last Payment indicates the likelihood of repayment in the future. The longer an account continues without repayment, the less likely it is to repay in the future. More points are therefore allocated to accounts that have recently made payments. Months Since Up-to-date is another typical recovery score characteristic. The longer the period since an account was up-to-date, the less likely it will repay in the future. Therefore fewer points are allocated to accounts that were up-to-date a long time ago. Current Delinquency is another common recovery score characteristic. Although recovery scores only apply to accounts that are already significantly delinquent, the higher the level of delinquency, the less likely an account will repay. Fewer points are therefore allocated to those accounts that are highly delinquent. Risk and Recovery Odds Odds is an accepted method of measuring risk or recoverability. It is an objective measure of risk or a way in which one can quantify risk or payment ability. It is an expression of the ratio of good accounts to bad accounts for a particular group of accounts In this simplistic illustration there are 11 apples, one of which is bad. Another way of expressing this relationship is by saying that 10 apples are good and one is bad, therefore the good/bad odds of this group of apples is 10:1. Odds are therefore calculated as the number of good accounts divided by the number of bad accounts. The higher the odds are for a group of accounts, the better the risk quality or the better the recovery propensity. Creation of Profiles Scoring is a tool used to predict risk. A particular score relates to a specific odds or level of risk. The norm in any scorecard development is that the scorecard is scaled in such a way that the higher the score, the better the quality or lower the risk is. For example, a score of 675 reflects a lower risk than a score of 600. This is commonly referred to as a score to odds relationship and assists with the development of robust and understandable collection strategies. In order to use scores effectively, it is best practice to group scores into profiles that are easy to understand and apply. Profiles are used so that different actions can be taken on groups of accounts that display significantly different levels of risk. If scores were available between 550 and 700, for example, one would not wish to take different actions at each individual score point. This would be impractical and too fine a distinction would have Date: August, 2010 Compiled by Profitera Corp. Page 5 of 9

6 to be made. Instead, one is more likely to take a set of actions on groups of accounts in a particular score range, and another set of actions on another score range. Once score ranges have been grouped according to similar behaviour patterns, each profile can receive different treatment within the collection area. It is also important that treatment is consistent across the organisation. For example, it is important to not send mixed messages, such as granting a credit limit increase on the one hand while simultaneously applying harsh collection calls on the other. Here we have an example behaviour risk scorecard which has been scaled so that a score of 630 represents odds of 3,75:1 while a score of 645 represents odds of 7,5:1, and so on. Each score range has been grouped into score bands or profiles, and given a name. In this example we have Platinum, Gold, Silver, Bronze, and Ore profiles. The level of risk represented by each profile can now be determined, as the score to odds relationship is known for this scorecard. As an illustration, the Silver profile equates to accounts with odds of 15:1 to just less than 30:1. This is the group of accounts that score between 660 and 674. Grouping scores into profiles simplifies the use of scores as the groups are formed according to known levels of risk and are more understandable than a wide range of scores. Practical use of profiles within strategies is also simpler and easier to understand than the use of scores. Here we have an example behaviour recovery scorecard. As previously described, recovery scores are built to predict the likelihood that the outstanding balance will be repaid over a particular period in the future. Again, as with the behaviour risk score, the scores have been grouped into profiles, and have been named, in order to make them more easily understandable within the collection strategy. In this example the Silver profile, which refers to the score band of , equates to recovery percentages of 20% - just less than 30%. That is odds of recovery of 0.25:1 to 0.43:1 Risk in Delinquency Management Strategies In order to comprehend the usage of scoring technology in the collection arena, we need to understand how scores are used to prioritise delinquent accounts. Date: August, 2010 Compiled by Profitera Corp. Page 6 of 9

7 In this illustration, accounts that are 6 or more months on books have been examined. The accounts have been split into low, medium and high balances so that each group of accounts can be treated differently. Furthermore, the accounts have been split into high, medium and low risk groups (profiles). It is good business practice to treat low-risk accounts differently to high-risk accounts, and to treat highbalance accounts differently to medium-balance accounts. High-risk high-balance accounts represent the greatest potential loss to an organisation and should therefore receive the highest collection priority. High-balance medium-risk accounts represent lower potential loss, and should therefore be allocated a lower priority level. As an illustration the relative priorities for each account segment has been assigned to aid with the development of effective collection strategies. Proxies for Risk and Recovery Prediction Scores will always give more predictive power than the use of proxies, as they are statistically and empirically derived. In a scorecard a combination of characteristics are used, which are correctly weighted to take into account the interrelationships (correlation) between individual characteristics. When scores are unavailable, other predictive information can be used as proxies for scores. Characteristics can be used as alternatives for scores when we know that these pieces of information are highly predictive. Examples of characteristics that can be used as proxies for behaviour risk scores include: Maximum delinquency during the last 6-12 months, as the higher the delinquency the higher the expected future risk; Number of times delinquent in the last 6-12 months, as the greater the number of times delinquent, the higher the potential risk; Months since last delinquency, as the longer the period since the last delinquency, the lower the projected risk; Months since maximum delinquency, as the greater the period since maximum delinquency, the lower the anticipated risk. Date: August, 2010 Compiled by Profitera Corp. Page 7 of 9

8 Examples of characteristics that can be used as alternatives for recovery scores include: Payments as a percentage of balance over the last 6 months, as the higher the % the greater the likelihood of repayment; Months since significant payment made, as the greater the period since significant payment the lower the likelihood of future payments; Number of payments made in the last 6 months, as the greater the number of repayments the greater the probability of further payments. Credit Bureau Scores Credit bureau scores were developed in the U.S. in the 1980s and have become standard industry practice for both application and ongoing account management. A credit bureau score is developed on pooled data, provided by all contributing credit grantors. Characteristics are based on payments, usage and delinquency across all accounts that an individual has. Expressed as a score, this enables an organisation to have an industry-wide, global view of the credit risk and predicted future performance of a customer. The Credit Bureau score is dynamic and changes according to the individual s behaviour on all of their reported accounts. Credit bureau scores not only assist in the initial granting of credit, but also the ongoing management of a customers risk. For example, a possible early warning of future delinquency is when a customer is performing well with the organisation in question but is performing poorly with other lenders, based on their credit bureau score. In the same way that application scores can be cross-matrixed (dual score matrix) with credit bureau scores, splitting out and analysing each behaviour score range by credit bureau score adds to the overall predictive power: Dual scoring can be applied to all areas of account management and brings additional business benefit when used correctly. The table detailed below illustrates how a dual matrix approach would be applied in a delinquent collection strategy. When using dual scores in account management strategies, it is very important to consider the age of the credit bureau score, as the score s predictive power tends to decline over time. A behaviour score is dynamic and is typically calculated on a monthly basis and is therefore always up-to-date. A credit bureau score is typically refreshed on a quarterly basis via an off-line batch procedure with accounts sent to the credit bureau for scoring. In many countries, the yearly FatMAN schedule has become the standard with batch refreshes in February, May, August and November. Promise-to-Pay Scores Many organisations have invested heavily into predictive dialler technology, risk and recovery scoring solutions, improved collection and workflow management systems, and a multitude of other productivity enhancement technologies. These solutions are geared towards maximising collection resources in order to reach a promise-to-pay (PTP) with the defaulting customer. Simply making an arrangement with a customer to pay a specified amount on a specified date is not the end of the collection cycle. Many customers will agree to the PTP with the full intention of not honouring the agreement. Leading- edge credit grantors are implementing new scoring techniques to Date: August, 2010 Compiled by Profitera Corp. Page 8 of 9

9 maximise their collection activities, namely PTP Scores. These PTP Scores predict the likelihood that the customer will honour the arrangement. Summary This concludes the scoring technologies applied by leading-edge credit grantors to ensure that the collection activity is maximised and cost-effective. About Profitera Corporation Sdn. Bhd. Profitera Corporation is a leading technology provider of Revenue Collections, Debt Recovery and Agency Management Software Solutions. With offices in Malaysia, Singapore, the US and Partners in Asia, Middle East and Europe, Profitera provides solutions to help its Customers focus on their Clients and optimize Profits throughout their Revenue and Profitability Lifecycle. As experts in Revenue Collections, Debt Recovery and Agency Management, Profitera continuously brings real-time scalable technology to the doorstep of its Customers to help reduce Bad Debt and improve Profitability. Enterprise Revenue Collections & Debt Recovery Software Systems SMS Notification and 2-way SMS Interaction software platform Data Management and Software Integration Services Data Analysis, OLAP and Multi-dimensional Cubes for Online Interactive Reporting Date: August, 2010 Compiled by Profitera Corp. Page 9 of 9