Analytics can be an extremely profitable investment – assuming efficiency and completion. Financial institutions, whether they handle it in house or turn things over to a trusted provider, need to know if they’re victims of leveraging or being served silo analytics to dictate their credit decisioning.
For those who’ve never heard the term, silo analytics can refer to using a single piece of (or incomplete) information to represent a holistic view of the consumer. It can also refer to a decentralized or fragmented process for analytics. For the latter, each department within an organization (such as marketing, credit risk, acquisitions and collections) often rebuild their analytic infrastructures from data gathering to the creation of analytic attributes rather than partnering across divisions to improve the speed and ultimately implementation.
For the first part of this two-part series, we’re going to focus on how to avoid using a single piece of information to represent a holistic view of the consumer.
Analytics is not just a report or score
A single report or score doesn’t truly give you the insight you need to run your business. A traditional credit score provides lenders with very point-in-time information on the consumer’s risk profile. What is often left out is information regarding where that specific consumer is going. Supplementing the traditional credit score with an application score can indicate if a high-risk consumer is applying for a low-risk loan. These consumers are trying to rebuild credit the right way instead of biting off more than they can chew. Low-risk loans show a low payment-to-income ratio and have a low overall change in the consumer’s debt-to-income profile. For auto loans, a low-risk loan will have a lower loan-to-value ratio (LTV), indicating some immediate equity in the vehicle.
If you are a lender with multi-client relationships, compare strategies that review your institution’s consumer risk profile to the risk provided from the full credit report. Relationships obviously carry a lot of weight. Consumers with low internal risk based on historical account information should warrant a review before declining new business opportunities. Simply limit the new credit line and price the loan based on the consumer risk to maintain profitability.
Use what-if analysis and inference for policy changes
Don’t make policy changes based on isolated results or single decision characteristics and policy rules. A what-if analysis and inference methodology can ensure you have all the facts. Using a what-If simulator tool will save both time and money compared to continual champion-challenger testing of new strategies. These tools will simulate the impact of one or more policy changes on auto decision rates, approval rates and delinquency rates. Only strategy and policy changes expected to yield the greatest benefit should then be thoroughly tested using a champion-challenger method.
Rebuilding custom scorecards from within your lender data can lead to having a custom score that is volatile and less predictive. Using inference methodologies on the rejected applications can remove the over-sampling biases from the data prior to optimizing a custom scorecard.
Moving forward
Getting the most out of your analytics solutions is critical to maximizing profits while minimizing exposure to unwanted risk. Unless your institution can afford to staff its own analytics department, this means you have to find a proven and trusted provider to guide you along the way.
MeridianLink’s MLX Consulting team has the expertise to help institutions of any size avoid silo analytics. But that’s not all. Indirect lending is a segment that presents a fundamental need for effective analytics. Our experts are hosting a complimentary webinar at 2 p.m. ET on Wednesday, March 20 for a deeper understanding of how data analytics should be leveraged to boost value generation. Please register by clicking the button below.
(For the second part of this two-part series, please click here)
Photo Credit: Jan Tik
