Machine Learning for Real-World Credit Risk

by Alan Ikemura 3 min read September 12, 2018

Machine learning (ML), the newest buzzword, has swept into the lexicon and captured the interest of us all. Its recent, widespread popularity has stemmed mainly from the consumer perspective. Whether it’s virtual assistants, self-driving cars or romantic matchmaking, ML has rapidly positioned itself into the mainstream.

Though ML may appear to be a new technology, its use in commercial applications has been around for some time. In fact, many of the data scientists and statisticians at Experian are considered pioneers in the field of ML, going back decades. Our team has developed numerous products and processes leveraging ML, from our world-class consumer fraud and ID protection to producing credit data products like our Trended 3DTM attributes. In fact, we were just highlighted in the Wall Street Journal for how we’re using machine learning to improve our internal IT performance.

ML’s ability to consume vast amounts of data to uncover patterns and deliver results that are not humanly possible otherwise is what makes it unique and applicable to so many fields. This predictive power has now sparked interest in the credit risk industry. Unlike fraud detection, where ML is well-established and used extensively, credit risk modeling has until recently taken a cautionary approach to adopting newer ML algorithms. Because of regulatory scrutiny and perceived lack of transparency, ML hasn’t experienced the broad acceptance as some of credit risk modeling’s more utilized applications.

When it comes to credit risk models, delivering the most predictive score is not the only consideration for a model’s viability. Modelers must be able to explain and detail the model’s logic, or its “thought process,” for calculating the final score. This means taking steps to ensure the model’s compliance with the Equal Credit Opportunity Act, which forbids discriminatory lending practices. Federal laws also require adverse action responses to be sent by the lender if a consumer’s credit application has been declined. This requires the model must be able to highlight the top reasons for a less than optimal score.

And so, while ML may be able to deliver the best predictive accuracy, its ability to explain how the results are generated has always been a concern. ML has been stigmatized as a “black box,” where data mysteriously gets transformed into the final predictions without a clear explanation of how. However, this is changing.

Depending on the ML algorithm applied to credit risk modeling, we’ve found risk models can offer the same transparency as more traditional methods such as logistic regression. For example, gradient boosting machines (GBMs) are designed as a predictive model built from a sequence of several decision tree submodels. The very nature of GBMs’ decision tree design allows statisticians to explain the logic behind the model’s predictive behavior. We believe model governance teams and regulators in the United States may become comfortable with this approach more quickly than with deep learning or neural network algorithms. Since GBMs are represented as sets of decision trees that can be explained, while neural networks are represented as long sets of cryptic numbers that are much harder to document, manage and understand.

In future blog posts, we’ll discuss the GBM algorithm in more detail and how we’re using its predictability and transparency to maximize credit risk decisioning for our clients.

Related Posts

How GenAI Fraud Bots Fly Under the Radar (and How to Ground Them Before It’s Too Late)

Today’s AI-powered fraud bots are exceptionally good at mimicking human behavior. Learn why a modern defense strategy is key.

November 18, 2026 by Laura Burrows
What Is Incident Response? Key Steps, Planning and Best Practices

Learn what incident response is, the key steps in the process and how to build and test an incident response plan for a data breach.

November 17, 2026 by Laura Burrows
Rethinking Charged-Off Portfolio Prioritization

Experian’s new Recovery Scores model turns portfolio intelligence into more focused recovery strategies, helping organizations improve returns, direct effort and accelerate speed to cash. The challenge isn’t simply working more accounts. Whether a debt buyer is evaluating an acquisition, an agency is deciding where to assign collector time or a lender is managing its own post-charge-off accounts, the question is similar: How can limited capacity be directed toward accounts with the greatest potential for recovery? Prioritizing accounts mainly by age, balance or when they were last worked may not show which ones are most likely to generate a recovery. With the new Recovery Scores model, organizations can rank charged-off accounts by expected collectability, providing a consistent signal to inform placement strategy and treatment decisions across the recovery lifecycle. Make collector capacity a strategic resource Working every account with the same level of attention isn’t always practical. The more useful question is: Where should teams focus first to protect speed to cash? Our model performance analysis found: 80% of recovered dollars come from the top 40% of accounts worked. The value isn’t simply a better ranking. It’s how organizations can use it for collections optimization. Score bands can inform different strategies for placement, outsourcing, auction and legal channel treatment rather than applying the same approach across an entire portfolio. For debt buyers: The scores can inform acquisition pricing and how a newly purchased portfolio is divided across internal collections, legal channels and resale. Expected collectability can shape debt portfolio analysis, including how a portfolio is valued and segmented before collection work begins. For agencies: Score bands can help teams focus collector resources on accounts with stronger recovery potential. For first-party lenders: Managing recovery internally, the scores provide another signal to help structure decisions about which accounts to work and in what order. Look beyond collectability alone Prioritization rarely comes down to a single variable. Consider two accounts that both rank highly for expected collectability, but one has a substantially larger balance. Their recovery likelihood may be similar, but the potential dollars recovered are not. The Recovery Scores model uses debt collections analytics by combining Experian’s bureau-based score with current balance information. This helps organizations consider both the likelihood of recovery and the potential dollars at stake, while sensitive balance information remains within the organization’s working environment. Monitor accounts as they change The first decision about an account is rarely the last. Teams continually make choices about treatment, placement and resource allocation as accounts move through the recovery process. Recovery Scores is also available as a trigger. When underlying information changes, teams can receive an update and reconsider an account’s placement without waiting for the next portfolio-wide re-score. An account that shows lower recovery potential today may have a different recovery outlook later. Turn recovery potential into a portfolio strategy When portfolios contain more accounts than teams can work at once, the question isn’t simply, “How many accounts can we work?” It is, “Where can our resources have the greatest impact?” The latest Recovery Scores model helps organizations sharpen placement and treatment decisions. It gives debt buyers a way to price and segment portfolios with greater confidence and gives lenders another tool to direct limited collector time across their own post-charge-off portfolios. Learn more about Experian Recovery Scores

November 10, 2026 by Zohreen Ismail

Subscribe to our Newsletter

Enter your name and email for the latest updates.

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.

Subscribe to our Newsletter

Don't miss out on the latest industry trends and insights!
Subscribe