Four strategies to optimize your decisioning

Today’s consumers need speed and convenience in virtually every experience. They are busier, contend with more distractions, and, as a result, are less willing to wait than in years past.

Their expectation for obtaining credit is no exception. Whether at a brick-and-mortar retail store or pushing a virtual shopping cart through an online checkout, progressive financial institutions and fintechs make credit offers available in just a few minutes. Their success is a direct result of a high percentage of automated approvals.

When approvable applicants are sent for manual review—often waiting one or more days to receive an offer—opportunities are lost. Borrowers may continue to shop and accept a loan offer from another lender willing to approve them instantly. Financial institutions that do not evolve to meet borrowers’ increasing expectations risk being left behind.

As lending becomes more competitive, it is essential to optimize your decisioning for efficiency and effectiveness to better serve borrowers. Begin by considering these vital decisioning strategies:

1. The Strength of Artificial Intelligence (AI)

Traditional risk models may use less than 20 attributes and often only consider information found in a credit report. Underwriting strategies typically consist of 10 to 30 cutoff rules—some set without data to back them. For example, length of residence, length of employment, and debt-to-income (DTI) are frequently used to trigger manual reviews, yet data shows these are less predictive of risk than other factors.

Machine learning algorithms (AI) can evaluate hundreds of attributes from the application and credit file, relationship and performance data from the core processor (host), and data from alternative sources to produce a risk model significantly better at separating high- and low-risk borrowers. When developed against your own performance data, these models improve revenue from increased loan production, introduce cost savings through efficiency gains, and reduce losses.

With more accurate risk assessments, more decisions are automated because fewer applicants fall into the middle tiers sent for review. More borrowers can be instantly approved. Lenders that leverage AI models reduce potential bias and become more efficient while positioning themselves to pick up qualified borrowers not approved by lenders who still use traditional models.

2. Configurability and Testing

Even if your LOS can’t leverage AI, a configurable decision engine with testing capabilities can significantly improve automation. Credit unions should be cautious about purchasing a LOS or decision engine that doesn’t have an administration tool or only supports a finite number of “canned” rules.

A low-code tool that offers rule-builders, decision trees, integrated AI, and the ability to model changes using a “champion challenger” or A/B testing enables you to validate new decision strategies for a positive impact before moving them to production. The more flexible the engine is, the more important the ability to test and model changes to underwriting strategy becomes. A “sandbox” where administrators can safely make changes and run what-if scenarios is essential.

Testing should be available using a single application or in batches of thousands, along with reports that provide insights on the impact of the proposed changes. Transparency into live decisions, versions, and audit history should not be overlooked, as these all enhance the ability to support, analyze, and refine your decision strategy.

However, not all focus should be placed on the approved status. Instant decision ratios will remain low if the credit union is not confident that the correct loan amount, terms, and conditions are being applied. The engine that approves borrowers must also set appropriate approval amounts for borrowers based on their credit experience and capacity to pay.

A modern decision engine can be configured to set smaller loan amounts, a limited repayment term, and conditions—such as a modest down payment and proof of income for first-time and thin-file borrowers.  It can also be configured for those with established credit, with amounts that correspond to credit depth and highest comparable loan amount.

A robust credit line assignment strategy enables more approvals because the lender knows they will not be overextending credit, and the amount of risk being taken is appropriate.

3. Customize Credit Offers and Services

Many lenders obtain “written instructions” from borrowers before or during the application process to provide additional offers of credit and relevant services. When applying for credit, the decision engine should evaluate all available data to offer relevant products to the borrower. The most common are auto loans, credit cards, and personal loans.

Some systems customize offers based on loans in the credit file, such as refinance offers, to reduce interest and monthly payments. They may even be able to identify borrowers who have fallen into the trap of relying on payday loans and offer credit union products designed to assist in breaking free from high-cost financing.

Members are generally receptive to these efforts to provide valuable services that can improve their financial health. They help build better relationships and solidify their credit union’s reputation.

4. Implement Enhanced Fraud Tools

It’s become easier for fraudsters to falsify information. Fake paystub sites are abundant, easy to use, and are common fraud tactics. Income is overstated on roughly 38% of loan applications, leading to a 90% increase in delinquency within the first 60 days, according to Informed.IQ, a developer of AI-based software that verifies, streamlines, and optimizes loan processing.

Not only does this impact the credit union’s ability to assess credit risk and portfolio performance, but it can be costly. In auto loans alone, employment and income fraud cost lenders $4.9 billion in 2023, according to Informed.IQ. This type of fraud can be complex and challenging to detect.

Many credit unions still rely on physical IDs in their Customer Identification Program (CIP). Others have implemented AI processes to leverage data collection from millions of documents in a fraud database to identify fraudulent paystubs and IDs. One lender tested these AI processes on documents collected from a sample of already opened accounts. The results? Fraudulent documents were identified with accounts and loans they had charged off— proving the potential for preventing losses.

Collecting a physical form of ID satisfies Know Your Customer (KYC) requirements, but it is not the most effective way for the digital channel. Modern fraud solutions can confidently validate a person’s identity using the data and devices used to apply for accounts. In their digital channels, large financial institutions have used these tools in lieu of physical IDs, significantly reducing fraud while at the same time removing the friction of requesting photo ID uploads.

The above are just a few readily available fraud prevention options. Today, lenders can verify that a social security number belongs to the applicant along with other personal data, identify red flags associated with an email address or phone number, check device reputation, run behavioral analytics, and more.

Together with AI underwriting and carefully tailored underwriting strategies based on actual performance data, credit unions can approve the most creditworthy applicants in real time. Credit unions have found that by considering these strategies, they have increased revenue, reduced losses, improved efficiencies, developed better credit offers, and satisfied customers.

While technology will continue to advance and improve efficiency, lenders should consider strategies to optimize their lending processes and tools to stay competitive.