How Natural Language Processing Improves Consumer Credit Loss Recovery

Banks extend credit to consumers with the expectation that they’ll be repaid on time and in full.  However, economic conditions can change suddenly, especially in the labor market, and borrowers may fail to adhere to their monthly loan repayment schedule.  

A loan is delinquent when a payment is 30 days or more late and the delinquency becomes “serious” when the payment is 90 days late. An unsecured loan, such as the outstanding balance on a credit card, that’s 180 days or more delinquent is considered a loss and charged off against the bank’s loan loss reserve. 

For a loan secured by an asset, such as an automobile or real estate, a loan default precipitates a legal process where the bank seeks to recover the collateral pledged as security for the loan. Recovery of the collateral is an expensive and time-consuming process, and the resale value of the collateral deteriorates materially during this process. 

A key metric for bank credit quality is the “roll rate” of loans from one state of delinquency to another. As loans move from 30-to-90 days and from 90-to-180 days, each month of delinquency increases the likelihood of the loan going into default.  

Consumers are aware that banks report the repayment status of outstanding loans to centralized credit bureaus and that delinquent or defaulted loans negatively impact on an individual’s credit score. While adhering to the legal and regulatory constraints on the methods used to recover on bad debts, the optimal course of action for the bank is to make every possible effort to get the borrower back on track to repay the loan. This is typically the best outcome for both the bank and the borrower.  

To address consumer loan delinquencies, banks communicate with borrowers to remind them of their obligations and remediate obstacles to repayment. The ability of the bank to communicate directly with the person responsible for the outstanding debt, or “right-party contact”, is critical in debt collection. With the decline in telephone land lines, it is increasingly challenging for banks to reach people to engage in a discussion about delinquency. Debtors sometimes wish to avoid being contacted or are experiencing a precarious employment situation which may cause them to change residence repeatedly. Creditors must sometimes pursue “skip tracing”, or the process to track down people who are missing, unresponsive or hard to find.  

For delinquent loan customers who also have deposit accounts at the bank, information gathered from these deposit accounts can be used by the bank to impute the borrower’s current income or observe whether payments are being made to other creditors. This allows the bank to assess the borrower’s capacity to repay the loan.   

A key activity for the bank is to collect and maintain a customer contact history, including all the physical and email addresses provided by the borrower. It’s also important to retain the phone numbers associated with the borrower, both those provided directly and, given the bank’s compliance policy, those collected when the borrower contacts the bank.  

For inbound calls, the bank can use Automated Speech Recognition (ASR) to processes a digital recording of the inbound call to generate a transcript of the call. The transcript is then evaluated by Natural Language Processing (NLP) via application of a Language Model (LM) and can be used to extract caller sentiment, and to identify call topics via Entity Recognition. Teradata ClearScape Analytics™ provides text analysis tools as well as exposure to open-source ASR and LM models via Bring Your Own Model. 

Personalization of the outbound customer contact strategies related to collections will generate the greatest likelihood that the borrower will “cure” the delinquency and restart repayment of the loan. Personalization involves identifying the outreach model with the highest likelihood of right-party-contact, based on the borrower’s stated preferences and previous behaviors. Successful collections personalization requires a multi-dimensional view of the borrower as a customer –including exhaustive information on customer interactions and modes of contact.  

With the vast amounts of data collection and analysis, it’s imperative for banks to leverage the most sophisticated analytics solutions. See how Sicredi uses Teradata VantageCloud on AWS and ClearScape to manage their credit risk portfolio and drive increased profitability.