Eric Stratman is Senior Director of Analytics & Insights at ValidiFI, where he builds machine learning systems that help companies verify bank accounts and catch fraud. His team analyzes data across 171 million payment records, 83 million bank accounts, and 1.4 billion inquiries to spot risk patterns that traditional verification methods miss.
ValidiFI’s AI can push account validation coverage from 85% to 96% by learning the structural patterns of routing and account numbers across thousands of financial institutions. The models also flag suspicious connections, like when multiple Social Security numbers show up tied to the same bank account within a short window.
Here, Stratman talks through how ValidiFI trains its models while staying compliant with financial regulations, why quality data matters more than complex algorithms, and what it takes to keep pace with fraudsters who are getting smarter.
ValidiFI positions itself as a leader in “predictive bank account and payment intelligence.” Can you explain how AI and machine learning form the foundation of this predictive capability, and what makes your approach different from traditional account validation methods?
At ValidiFI, the foundation of our predictive capability is our data. Our data teams are very diligent when it comes to the inclusion of data into our solutions. This diligence allows us to create bank account and payment intelligence centric solutions where we can provide analytical insights into the status, ownership, behavior, and performance of a bank account. We believe that when you have high quality data, you can best utilize various AI and machine learning methodologies to provide that predictive capability to our customers. Whether that is through our account validation pattern matching to further extend coverage, or by analyzing account velocity and behavioral data to uncover fraudulent actors that appear legitimate on the surface.
You’ve mentioned that your AI techniques can extend account coverage from 85% to 96% by analyzing bank account and routing number patterns. Can you walk us through how your machine learning models identify these patterns and what specific signals they’re detecting that human analysts might miss?
Our AI and machine learning pattern matching can analyze the structure of account numbers based on their routing numbers. Human analysts are pretty adept at identifying patterns naturally, so it’s not necessarily something they might miss. As an example of one trend our machine learning process can tell you, almost all ACH capable account numbers from bank A are 8 digits long and begin with a 3. If a human analyst was given a list of successful transactions from bank A, they would likely come to the same conclusion. Where AI and machine learning excel is that, at scale. There are thousands of routing numbers and each one has different structural behaviors. With AI and machine learning, we can process this on a large scale and identify even more granular details within the account structure. Add in the fact that real time data is coming in that can change what we thought to be previously true. AI and machine learning also allows us to adapt as more data and information is provided. On top of structural analysis, our pattern matching will also incorporate historical data to ensure we are providing strong predictive indicators of account validity to our customers when direct validation is not available. This allows us to increase validation from 85% to 96% coverage when utilizing pattern matching.
Your fraud detection system analyzes connections between bank accounts, consumers, and payment behavior. What types of AI algorithms are you using to detect these relationships, and how do you handle the challenge of identifying sophisticated fraud that might appear legitimate on the surface?
At ValidiFI, our fraud detection capabilities are built on the ability to analyze billions of data elements across our consortium network—connecting bank accounts, consumer identity attributes, and consumer velocity behavior in real time. We utilize a variety of AI algorithms to identify fraudulent behaviors and high-risk indicators. We will often leverage supervised learning methods to identify the most effective indicators at identifying high-risk fraud patterns. For trying to identify general patterns we are seeing, we may look at other unsupervised learning methods that can uncover relationships and anomalies. For example, we can identify when a bank account is linked to multiple SSNs while also being commonly associated with a VOIP phone. These signals may not raise red flags individually, but when analyzed in a larger context, they can reveal coordinated or emerging fraud tactics.
With over 171 million payment records and 83 million bank accounts in your database, how are you using AI to turn this massive dataset into actionable insights? What’s your approach to training models on financial data while maintaining privacy and regulatory compliance?
ValidiFI’s data network continually grows month-over-month and with it, our ability to deliver deeper, more predictive insights. Along with 171 million payment records and 83 million bank accounts, our database also has coverage on over 1.4 billion inquiries, and over 61 million unique consumers—giving us one of the most comprehensive financial views in the industry.
We use AI as a strategic layer on top of this massive dataset to transform raw data into real-time, actionable intelligence. Our machine learning models analyze patterns across bank account status, ownership, payment behavior, and identity attributes to predict trustworthiness and performance.
We also take privacy and compliance seriously. Our models are trained using strict data governance protocols, including anonymization, encryption, and access controls. We ensure that all data usage aligns with regulatory frameworks like GLBA and FCRA, so our partners can trust that insights are both powerful and compliant.
Your velocity metrics can detect risk patterns like “3+ phone numbers in 30 days” leading to 70.5% increased fraud risk. How does your AI determine which behavioral signals are truly predictive versus coincidental, and how do you avoid false positives that could harm legitimate customers?
To avoid false positives, our models and insights are built on large, diverse samples. We also don’t rely on single data points; we look at combinations of velocity metrics, bank account status, and identity inconsistencies. For example, a phone number change alone may not be risky, but when paired with a surge in inquiries, mismatched identity data, and unusual account velocity, the risk increases significantly.
We also continuously retrain our models and evaluate patterns using real-world outcomes—such as confirmed fraud cases, return rates, and resolution data—to refine accuracy and reduce bias. This feedback loop ensures our models evolve with changing fraud tactics while maintaining fairness and precision.
Traditional credit scoring often lags behind a consumer’s current financial situation. How are you using AI to analyze real-time bank account data to provide a more current picture of creditworthiness, and what advantages does this give lenders over traditional FICO scores?
We leverage AI as a layer in analyzing real-time bank account data—including inquiries, payment history, tradeline performance, and bank behavior patterns to build a dynamic and current view of a consumer’s financial health. Unlike traditional credit scores, which are static and often outdated, our models continuously assess financial behavior to reflect true creditworthiness in the moment. This allows lenders to make timely decisions with precision and the ability to expand consumer reach to those underserved.
You work with companies like Velera and Amount across different industries. How do you customize your AI models for different verticals – for example, does fraud detection for automotive lending require different algorithms than for convenience retail loyalty programs?
When it comes to developing solutions for different verticals, we like to take into account what data we are feeding into the training dataset of a model. As you rightly pointed out, the challenges and patterns found in retail, particularly in convenience, can differ significantly from those in automotive lending. While the underlying AI algorithms may remain consistent across verticals, at the end of the day, it is the data that is fed into the development of the model that shapes how those algorithms perform.
The financial services industry is heavily regulated. How do you ensure your AI models remain compliant with regulations like NACHA rules and FCRA requirements, while still providing the predictive insights your clients need?
As a SOC 2 compliant company, we take data privacy, security, and compliance seriously. We maintain robust data security, privacy, and risk management controls, supported by continuous monitoring and evidence-based verification of controls. Our models are trained using strict data governance protocols, including encryption, and access controls. We ensure all data usage aligns with regulatory frameworks like GLBA and FCRA, so our partners can trust that insights are both powerful and compliant.
As fraudsters become more sophisticated and potentially use AI themselves, how is ValidiFI evolving its machine learning approaches to stay ahead? What’s your strategy for this technological arms race?
We recognize that fraud is no longer static—it’s adaptive, fast-moving, and increasingly powered by AI. That’s why our strategy is built on quality data, first and foremost. And utilizing continuous learning, real-time pattern recognition, and consortium-scale intelligence. To stay ahead of AI-powered fraud, we also invest in model agility. Our infrastructure supports rapid iteration and deployment of updated models, allowing us to respond to new fraud tactics in near real time. Ultimately, our goal is to ensure that our clients are not just reacting to fraud—they’re anticipating it. By combining deep data coverage with AI, ValidiFI helps institutions stay ahead and protect their customers with confidence.
Looking ahead, where do you see AI making the biggest impact in financial services beyond what you’re doing today? Are there emerging AI technologies or techniques that you’re excited to potentially incorporate into your platform?
We are always looking at differing machine learning methodologies that can evaluate consumer and banking behaviors, which can identify trends that are strong predictors or risk and fraud. Looking ahead, we’re particularly excited about emerging AI and machine learning techniques that can help us identify and optimize customer strategy development. By tailoring our solutions to fit the unique needs of each client, we aim to deliver even more precise, predictive account, payment, and credit risk intelligence.
