Behavioral analysis as an approach for more accurate fraud risk detection
Season 6 Episode 10
Transcript
Juan José Ríos (host):
To understand fraud, we need to look beyond the simple details of a transaction. Ultimately, an isolated transaction is just a snapshot, a moment in a much larger story. Today, financial institutions must understand the context, habits, and patterns of their customers in order to protect them without affecting their digital experience.
In recent years, this understanding of user behavior has become a key factor in anticipating fraud.
Welcome to Mundo Financiero Seguro, the Plus TI podcast.
I'm Juan José Ríos, and today we'll be talking about how behavioral analysis is transforming fraud prevention in Chile and throughout the region.
To this end, I am joined by two distinguished guests:
Diego López, Assistant Manager of Fraud Prevention at SBP Chile, with more than nine years of experience in payment fraud management; and José Ruiz, Product Manager of Transactional Security and Digital Fraud at Plus TI. Welcome to both of you, it's a pleasure to have you here.
Diego López:
Thank you, Juan José. It's a pleasure to be here with you all.
In my experience, there has been a significant evolution in the way financial institutions understand and measure their customers' behavior to prevent fraud. Previously, the approach was much more transactional and simplistic, based on static rules that generated reactive, delayed detections with high levels of false positives.
Today, analysis is much more dynamic and predictive. Advanced behavioral models are integrated that consider the entire context of customer interactions. We no longer look solely at the transaction, but also at the type of device, geographic location, frequency of use, how the customer interacts with digital channels, and many other factors.
This allows anomalies to be identified more accurately and in real time. In addition, advanced segmentation has been incorporated, which not only analyzes the customer individually, but also according to characteristics shared with other similar groups or segments, considering transactional, demographic, economic, and social criteria.
This broadens the perspective and makes it possible to determine what is normal for both an individual and a specific group, even when that behavior deviates from the general average.
Juan José Ríos:
In that context, have you experienced situations in which a change in customer behavior has required adjusting the detection criteria?
Diego López:
Yes, especially during promotional periods or sales events with large discounts for limited time only. In these scenarios, we observe significant variations in customers' usual behavior.
While the offers explain part of the increase in consumption, we cannot rely solely on that argument. It is essential to have a proactive analysis that considers the history of similar events in previous years and allows us to assess whether the behavioral variation is explained exclusively by the event or whether there are additional risk signals.
To do this, we analyze historical trends, compare behaviors at the individual level and across similar segments, and monitor everything in real time to confirm or rule out hypotheses and take the appropriate actions.
This analysis must be dynamic. It is not enough to define it once and keep it fixed over time. New patterns of fraud are constantly emerging, and the challenge is to correctly classify the behavior of legitimate customers while continuing to detect changes in fraudsters' tactics in a timely manner. It is a constant race to adapt.
Juan José Ríos:
José, from your technological perspective, what is the biggest challenge today for institutions to take full advantage of behavioral analysis?
José Ruiz:
One of the main challenges is legacy systems. Many institutions operate with technologies designed for processes that are very different from today's, which also do not always communicate with each other. This leads to fragmentation of information, with multiple isolated databases and dependencies on external providers.
When data is scattered, it is very difficult to obtain a unified view of user behavior, which is essential for effective analysis. In addition, this fragmentation affects the quality, consistency, and updating of data, especially when real-time information is not available.
Customer information is often scattered across systems such as core banking, CRM, and other platforms, reducing predictive capabilities and limiting the effectiveness of behavioral analysis.
Juan José Ríos:
Do you think institutions are ready to move toward more coordinated and collaborative prevention models?
José Ruiz:
There is a clear trend toward this goal, driven by digitalization. The challenge is to eliminate technological and organizational silos in order to integrate multiple data sources and achieve a unified view of the user.
However, beyond the budget, there are barriers related to organizational agility and alignment between areas. Each institution has different realities: a fintech company is not the same as a traditional financial institution, where decision-making processes tend to be more complex.
When areas do not share a common goal, responsiveness is lost and it becomes difficult to adopt a truly proactive stance.
Juan José Ríos:
Diego, one of the big challenges is defining what is normal in customer behavior. How do you manage that?
Diego López:
Defining what is normal is one of the biggest challenges. To do this, we start from a completely analytical basis, supported by large volumes of historical information, both transactional and digital interactions.
We combine this history with appropriate customer segmentation, which gives us a broader view of behavior and allows us to determine whether a deviation is truly anomalous. Artificial intelligence and machine learning models play a key role here, enabling us to process large volumes of data and identify complex patterns in real time.
Even so, the human factor remains indispensable. The judgment and experience of analysts allow them to cover scenarios that automated models do not always detect. Effective prevention is a balanced combination of technology and human talent.
Juan José Ríos:
José, in Chile, Law 20.009 requires traceability and accountability in fraud management. How does behavioral analysis contribute to regulatory compliance?
José Ruiz:
Current regulations, not only in Chile but globally, tend to shift responsibility to financial institutions, assuming that the user is innocent until proven guilty.
In this context, behavioral analysis is key, as it provides detailed and contextualized traceability beyond an isolated transaction. This makes it easier to explain and support decisions made, both for audits and regulatory defense.
Furthermore, in an instant payment environment, response times are drastically reduced. Understanding normal customer behavior allows controls to be applied selectively and proportionately to risk, reducing unnecessary friction in the user experience.
Juan José Ríos:
Diego, how do you maintain your ability to anticipate without losing accuracy or confidence?
Diego López:
The key is constant proactivity. Technologies, regulations, and behaviors are constantly changing, and we cannot afford to react too late. Payment methods are evolving rapidly: QR codes, digital wallets, payment gateways, fintechs, and new transactional models.
This requires continuous investment in technology, team training, and a mindset of constant evolution. Customers change, but fraudsters change even faster. Staying up to date is essential to maintaining detection accuracy.
Juan José Ríos:
This conversation makes it clear that preventing fraud is not just about having the best technology, but about adopting the right approach and having teams capable of interpreting, contextualizing, and adapting models to each client's reality.
Human behavior is constantly changing, both among legitimate users and fraudsters. And with every technological advance, new risks emerge, but so do new opportunities to prevent them.
Thank you, Diego and José, for sharing your experience and vision.
Thank you for joining us on Mundo Financiero Seguro, the Plus TI podcast.
I'm Juan José Ríos.
Until next time.
