毕博上海银行咨询Final Deliverables Retailfinal.ppt

Retail Scorecard Implementation:Recommended Approach,Contents,Retail Scorecard Implementation,cutoff 2 Score band:cutoff score band():Quick hit:,loan review,Establishing initial score cutoff:ConsiderationsIn theory,the scorecard cutoff should be set at that point where the marginal contribution of new accounts obtained at the cutoff score is equal to the marginal cost of the bad accounts predicted at that probability level.There are two reasons why this theory,while valuable,must be tempered by other factors in current use:first,most banks are unable to precisely measure these values;second,it is also important to evaluate the rate at which the ratio of goods-to-bads is changing when establishing score cutoffs.By reviewing the odds ratio for accounts near the cutoff,the bank can determine where the odds ratio is dropping below the population odds for the entire scorecard.This will suggest where the returns from any additional analysis are unlikely.Finally,once the initial score cutoff is established,Hanvit will need to incorporate the improvement of the second score in its underwriting practice.,Retail Scorecard Implementation,cutoff:cutoff cutoff:1)2)cutoff cutoff odds ratio,odds ratio odds ratio.cutoff,2.,Establishing initial score cutoff:MethodologyIdentify the initial cumulative probability of bads that is the desired risk management target(for example,no more than 3%likelihood of bads 60 days or more past due)and the associated score valueReview the rate of change of the ratio of goods-to-bads around that score value,and evaluate the impact on total volume.Compare the impact on total volume with the impact on overall portfolio delinquency.,Retail Scorecard Implementation,cutoff:(:60 3%),Adjusting for the second score strategyUse of the second score enables us to harvest an additional number of goods at an acceptable levels of bads due to the additional probability information provided by the second score.When using a second score strategy,we need to consider the effects of the sequence in which the steps are applied:In a strictly sequential approach,modelers generally advise to adjust the threshold of the first test downward in order to maximize the discrimination of the second test at the cutoff margin(that is,expand the grey zone).It is possible to develop a matrixed approach in which the second score and first score are explicitly considered in the establishment of the initial cutoff margin.In effect,however,by using additional underwriting intervention on the grey zone accounts,these interactions should be able to be taken into account without complicating the automated strategy.,Retail Scorecard Implementation,2 2 2,cutoff 2 1.(,)cutoff 1 2.,.,Identifying the score bandsEstimation of the score band ranges can be accomplished by using two complementary techniques to group score ranges with similar probability of default into the desired number of score bandsFirst,the points at which the cumulative probability changes by a given increment(e.g.,.5%)are identifiedSecond,the odds ratio(probability that an account will default within that score range;the ratio of goods to bads within that score range)is computed and compared across the score bands.Finally,the change in odds ratio across the score bands is reviewed and adjustments are made to ensure reasonably similar probabilities are grouped into bands.Two contiguous score bands should not have very similar odds ratios;this would suggest the bands should be re-analyzed to achieve a more uniform decline in probabilities between the bands.,Retail Scorecard Implementation,Score band score band 2,score band(:5%)odds ratio(;),score band score band odds ratio,band.score band odds ratio;band band,Establishment of loan limitsThe primary methodology for establishing loan limits in the United States is based upon the concept of the debt-to-income ratio.These ratios were first developed in association with mortgage loans,and their use has expanded to include maximum limits based on combined mortgage and installment debt-to-income ratios.Mortgage debt-to-income ratios have come to include a ceiling on housing cost debt to income,plus total installment debt.Typically,housing cost debt-to-income should not exceed 30%of a borrowers gross income,while installment debt should raise the total to a maximum of no more than 40%.It should be noted that these figures vary by institution,and have developed in an environment reflective of US housing costs,tax burden,and overall living costs.,Retail Scorecard Implementation,.mortgage loan,mortgage.Mortgage.30%,40%.,Establishment of loan limitsThese overall limits can be applied to develop general factors as follows:Debt-to-income ratio=(Interest rate*Limit)+(Amortization factor*Limit)IncomeDebt-to-income ratio=(IR+AF)Limit Income DIR*Income=(IR+AF)Limit DIR*Income=Limit(before deducting other bank debt(IR+AF)and guarantees),Retail Scorecard Implementation,=(*)+(*)/(DIR)=(IR+AF)*/DIR*=(IR+AF)*(DIR*Income)/(IR+AF)=(),Establishment of loan limitsApplying this ratio to a mortgage loan with a maximum debt to income ratio of.30,an 8%rate,and 25-year amortization,for a consumer with an annual income of$45,000:.30*45,000=$112,500(.08+.04)Note that this is equal to approximately three times annual income,although our amortization factor is higher than usualSecured loans typically also have a secondary limit established by the value of the collateral,generally 80%loan-to-value for a mortgage loan.To qualify for the above loan,the collateral would need to be worth:$112,500/.8=$140,625,Retail Scorecard Implementation,0.30,8%,25,$45,000 mortgage loan:(0.30*45,000)/(0.08+0.04)=$102,000,3 2.(mortgage loan 80%),$112,500/0.8=$140,625.,Establishment of loan limitsApplying this ratio to an installment loan the same consumer,with an interest rate of 12%,an amortization factor of five years,and a maximum DIR of.10:.10*45,000=$14,000(.12+.20)Note that this amount can vary significantly depending on the amortization factor and interest rate assumption,Retail Scorecard Implementation,12%,5,DIR 0.10:(0.10*45,000)/(0.12+0.20)=$14,000,Application to the Hanvit score card:score cutoff and score bands-Unsecured/guaranteed scorecardCutoff is established based on combination of:Cumulative probability of desired risk management targetAdjustment for second score strategyOdds ratio bottom scoreband:8Improvement of second score:2:1Resultant odds ratio:16:1(population odds),Retail Scorecard Implementation,Score cutoff:350,Auto decline,Auto approval,Grey zone,Scorecard:scorecard cutoff score band-/scorecardCutoff 2 scoreband odd ratio:82:2:1 odds ratio:16:1(odds),Application to the Hanvit score card:Limits-Unsecured/guaranteed scorecardLimits are established based on maximum debt-to-income ratio adjusted for required amortization as scores decreaseLimits for guaranteed loans will be based on the income and risk grade of the borrower,not the guarantorMaximum limit:(DIR*Income)-Other bank debt-GuaranteesMaximum DIR=12%,Prime rate=10%Limits for new customers(no transaction history)should be set at no more than 75%of the maximum and increased based on satisfactory performance,Retail Scorecard Implementation,Scorecard:-/scorecard:(DIR*)-DIR=12%,Prime rate=10%()75%,.,Application to the Hanvit score card:Pricing-Unsecured/guaranteed scorecardPricing is based on a base rate,expense factor,risk factor,and product adjustment Note:at this time,base rate,expense factor and adjustment factor are incorporated into base ratePricing for guaranteed loans will be based on the risk grade of the borrower,not the guarantor,Retail Scorecard Implementation,*Assumes a 10%prime rate,Scorecard:-/scorecard,(,.),Grey-zone underwriting criteriaCharacterWhat is the purpose of the loan,and is it consistent with bank policy?What is the customers previous history of handling loan obligations?Number,amount,payment historyCapacityWhat is the expected source of repayment for the loan?What is the customers income history?Is it stable,and is the trend favorable?What is the customers total debt,and what is the trend?What is the customers current debt-to-income ratio,and what is the pro-forma ratio including the payments on the proposed loan?What is the likelihood the customer will be able to repay the loan in accordance with the proposed terms and conditions?,Retail Scorecard Implementation,?(,)?,?,?,?,Grey-zone underwriting criteriaCollateralAutomobile/major purchaseIs the loan advance consistent with bank policy?Will the bank acquire a security interest in the purchase of the item?Real estateIf the purpose of the loan is not for purchase or refinance,what is the proposed use of the proceeds?Marketable securitiesIs the purpose of the loan consistent with bank policy?Can the customer support repayment if the securities decline in value?Will the bank take custody of the pledged securities?,Retail Scorecard Implementation,/?,?,Override policy and strategyOverrides can be distinguished by classifying them into three different types:Informational overrides,where information not taken into account by the scorecard is obtained that results in a change in the credit decision;Policy overrides,where certain loans that would otherwise be qualified(or disqualified)are declined or approved for management reasons;Judgmental overrides,where the decision is made to approve or reject a loan based on judgment outside the scorecard,Retail Scorecard Implementation,3:():,Override policy and strategyHanvits override policy should:Identify the specific situations and factors where an override may be permissibleIdentify the required authority to approve the overrideEstablish monitoring systems to ensure that override policies are adhered toMonitor the overall level of overrides for indications that some credit characteristics may not be adequately reflected in the scorecard,Retail Scorecard Implementation,Override policy and strategy:underwritingSpecific policy recommendations should include the following:As a general rule,the low-side override ratio(incidence of overrides below the cutoff divided by total applications below the cutoff)should not exceed 10%.The number of types of overrides should be kept to a manageable number,usually ten or less.Override factors should not include characteristics included in the scorecard.The meaning and use of override codes should be clearly defined;a priority of codes should be specified to determine which code should be used where multiple codes may apply.The use of override codes should be closely monitored;trends may identify factors that should be included in future improvements in the scorecard itself.,Retail Scorecard Implementation,:(cutoff)10%,10;,Scorecard management and enhancementThe plan for management of the scorecard should include the following objectives:Front-end report trackingPopulation stabilityFinal score reportOverride tracking and analysisSecond score evaluationCorrelation of second score/underwriting resultsCorrelation of second score/performanceLow-score performance tracking,Retail Scorecard Implementation,Front-end report 2 2/2/,Scorecard management and enhancementThe plan for improvement of the scorecard should include the following objectives:Data collection:Year 1Accumulation of loss data by score to support the improvement of pricing bands and establishment of the cutoff pointAccumulation of performance data by score band to measure the effecitveness of the modelAccumulation of improved training/validation data to support the development of improved future scorecardsCharacteristic analysis and refinement:Year 1+Segmentation analysis:Year 1+Process enhancement:Delinquency/loss forecastingRoll rate analysis,Retail Scorecard Implementation,:1 cutoff training/validation:1:1/Roll rate analysis,Customer Relationship Management(Marketing):Marketing has traditionally encompassed a number of separated activities:Product development involves the introduction of new products and enhancement of existing products.It also includes such methodologies and market segmentation to identify what products appear to specific markets,as well as market research to study customer needs and what makes consumer choose one product over another.Channel management involves the selection of the proper mix of sales,fulfillment and service methods.Institutions are increasingly developing mixed channel strategies that involve multiple service and fulfillment systems and both push and pull sales strategies.Sales management involves the training and motivation of the direct sales force.It includes the management of direct staff(branch and calling officers)and centralized strategies,including call center marketing,Retail Scorecard Implementation,():.:,.push pull:.call center,call center,Customer Relationship Management(Marketing):Customer Relationship Management has emerged as a new discipline that brings these desperate efforts together in one activity:knowledge-based marketingKnowledge-based marketing both improves the execution of the Banks existing strategy,but provides continuous feedback information for improvement.Channel strategy is improved through up-to-the minute feedback on channel utilization and effectivenessProduct development is enhanced through the understanding of multi-product use patterns and the development of effective cross-selling programs.Sales management is enhanced through tools that put meaningful customer information in the hands to sales staff to improve their sales performance.,Retail Scorecard Implementation,RM()RM,cross-selling programs,Marketing Initiatives(quick hit)Improve Hanvits customer market information retrievalQuery toolMarketing information data added to DBLeverage Hanvits superior market presence(leverage existing channels)Branch calling programs(company-oriented)Customer re-acquisitionBuild on broad-market convenience marketing efforts(channel enhancement)Supermarket branch pilotDevelop integrated approach to key product segmentsReal estateHome buyer seminarsBroker calling programsPre-qualification programs,Retail Scorecard Implementation,:Quick Hit-Query toolDB-()calling programs-()-,Marketing Initiatives(quick hit)Emphasize developing multi-product relationships(sales management)Personal banker program(mid-to-high net-worth)Sales incentive programs.,Retail Scorecard Implementation,(Quick hit)-()PB(,)Sales incentive programs,Marketing Initiatives(quick hit)Branch calling program(company-oriented)Focus:people like to bank where they workEach branch should identify target firms close to the branch with likely target customersProfessional services firmsCorporate headquartersDevelop a package of special services for presentation to the firmDiscounts on feesSpecial loan packagesFinancial counselingVIP delivery serviceIdentify ways to make it easy for customer to change banksMonitor program success,Retail Scorecard Implementation,(Quick hit)calling program():)VIP,Marketing Initiatives(quick hit)Customer re-acquisitionDevelop target lists of attractive custom