Using Data to Predict and Manage Delinquency
Banks collect vast amounts of customer behavioural and transactional data which is normally used to better understand them through customer value scoring, segmentation, marketing offers, and credit and behavioural credit scoring. This provides the basis for preventing and managing delinquency.
Prevention is better than Collection
We have discussed the process whereby the bank reminds customers that repayment of their loan or credit card bill is due through a gentle ‘nudge’ a few days before it is due, reinforcing good behaviour that ultimately benefits customers through higher credit limits or better interest rates.
Banks can use behavioural credit scoring to undertake pre-collection debt restructuring before the customer’s financial situation deteriorates to a point of no return.
Behavioural credit scoring is a predictive model that calculates the likelihood of any customer within the bank’s portfolio defaulting within a period, normally six months.
Every month, it tracks key indicators of default, such as falling credit turnover against steady or rising debit turnover, changes in cash withdrawal behaviour and credit limit utilisation using 12 months of data or more to calculate a ‘likelihood of default’ score. Previous months’ scores are compared against the most recent one to highlight positive or negative trends.
By analysing negative changes, banks can take pre-emptive action by contacting the customer, offering help to restructure their finances. This might be done by extending the repayment period to match customers’ ability to repay, to avoid having to start a collection process that is costly for the bank and the customer. If the customer’s financial circumstances have dramastically changed due to, say, being furloughed or losing their job because of Covid-19, the bank may help by offering debt repayment holidays for a few months. However, by taking a repayment holiday a customer’s real risk can be masked, so it is important that the information is held as machine-readable data that can be used in behavioural scoring and collections algorithms.