TSYS Foresight ScoreSM with Featurespace®: FAQs

What is Foresight Score?

Foresight Score is a fraud- and risk-management decisioning scoring tool that incorporates innovative machine learning capabilities to deliver significant benefits and advantages to an issuer's ability to fight transactional fraud. It was developed with UK-based behavioral analytics specialist, Featurespace, and the solution relies on Featurespace's powerful adaptive behavioral analytics platform. This platform, known as ARIC®, combines statistical algorithms with advanced machine learning to take predictive analytics to an entirely new level in dynamic risk and fraud management.

Why is this technology necessary?

Fraud losses are on the rise. According to The Nilson Report®, in 2015, total fraud losses incurred by financial institutions and merchants on all types of cards — credit, debit and prepaid general-purpose and private-label cards — was $21.84 billion and is projected to reach $31 billion by 2020.

An increase in ecommerce is exacerbating the problem. While still representing a relatively small percentage of total retail sales volume, ecommerce's share continues to grow at a faster rate than brick-and-mortar store sales.

All of this increases the opportunity for 'cardnot-present' (CNP) and digital fraud. According to Aite Group, CNP fraud now represents 45 percent of all card fraud.

In addition:

  • Global card volume was $31.310 trillion, so for every $100 in purchases, approximately 6.97 cents was fraudulent.

  • Fraud in 2015 increased by 21%, which outpaced the 7% increase in card volume.

What can Foresight Score to do address these problems?

The machine learning technology makes it possible to identify new, previously unidentified fraud transaction types in near real time. Predictions and models are built and updated quickly and efficiently based on each customer’s and merchant's individual profile, rather than relying on historical averaged data from a wide group of users.

The product uses 'Bayesian-based' statistical models, making it uniquely and especially adept at spotting subtle variances from predicted human behavior, allowing it to accurately predict both new and unknown fraud types. Foresight Score continually self-learns from every change in the datasets related to individual customers and merchants.

What's the difference between consortium-based models and Foresight Score?

In the simplest overview, the way consortiumbased models work is that they aggregate enormous volumes of data sampled from a group (or 'consortium') of financial institutions. This data essentially represents a snapshot in time — for example, a specific three-month, six-month or one-year period. Consortium-based models have been in use for decades and continue to be highly effective at identify and preventing many types of fraud. However, the efficacy of this fixed data degrades somewhat over time. The fraud environment is evolving faster than ever, so if current data hasn't yet been included or properly weighted in the existing model, these previously unknown types of fraud may be improperly scored or missed — a time-consuming and costly process.

How is Foresight Score different?

Foresight Score is different from consortium-based fraud-management scores. While there are multiple ways that it differs from other systems, it can be boiled down to three key areas.

  • It uses flexible data. Fraud solutions that rely on a fixed data set sampled over specific timeframes are at a distinct disadvantage in identifying new and previously unknown fraud types. Scoring systems must enable customer profiling and anomaly detection that makes use of machine learning and behavioral analytics to process end-to-end information – using both monetary and non-monetary data. Foresight Score does exactly that.

  • It captures this data in real time while continuously improving. Business and fraud schemes are evolving and changing moment to moment, so fraud-detection systems must do so as well. Our solution captures data in real time. This allows Foresight Score to be constantly updated over time based on changing datasets, so they can be used to predict customer behavior and score individual transactions in real time.

  • It has self-learning capabilities. With fraud evolving at an accelerated pace, humans simply can't keep up. While human involvement will likely always be required to review transactions and behavior to detect dangerous trends, suspicious activity and block fraud attacks, leading-edge systems need to self-learn to stay current and adapt based on each customer’s individual behavior. Foresight Score does this and makes it possible to truly understand customer actions and spot anomalies, while not getting overwhelmed by false positives.

What type of fraud does Foresight Score address?

Two specific areas where Foresight Score shines are in card-not-present (CNP) and high-ticket transactions such as commercial purchases.

  • CNP Transactions. The advanced algorithms and rules some consortium models use to auto-detect irregular patterns and suspicious transactions while limiting false positives are at a disadvantage because they rely on a data model that generalizes behavior across a broad, static data set. Foresight Score builds and dynamically updates a model of each individual customer's behavior, giving a clearer picture of which CNP transactions are fraudulent.

  • High-dollar-value transactions. Consortium based scoring models are also at a data disadvantage in helping institutions recognize fraud among commercial transactions or high ticket consumer transactions. That’s because these models are built on an enormous dataset that reflects the typical consumer transactions within a large representative sampling of customers from a consortium of institutions. Regardless of the institutions sampled, there are likely to always be more small- and mid-range transactions in that dataset. So, the model best recognizes suspicious transactions among these dollar-value ranges. Foresight Score instead builds that customized model for each individual customer, giving a much better handle on departures from the norm among these otherwise underrepresented customers within the consortium-based fraud-scoring models.

What are some of the features of Foresight Score?

  • It's focused on the customer level. Foresight Score essentially builds individual customer profiles around what good transactions look like, rather than following the consortium model of profiling what bad transactions look like.

  • Algorithms improve over time. This feature relies on self-learning algorithms that do not degrade over time like models based on fixed-period snapshots. This allows Foresight Score model to constantly evaluate its accuracy against the rate of change in fraud attacks — effectively balancing reacting too quickly to new and unknown behaviors, while understanding which changes are and aren't acceptable.

  • It is efficient at case management. Foresight Score is fully integrated and ready to write rules within the proven TSYS CardGuard℠ ecosystem.

  • It uses real-time machine-learning anomaly authentication. It automates real-time anomaly detection to spot fraud attacks and identify concerning trends as they happen.

What are some of the benefits of Foresight Score?

  • Accurately detects and prevents transactional fraud. Spots anomalies in real time to identify a wide variety of types of transactional fraud at the moment it occurs, without costly and time consuming manual intervention.

  • Enhances precision in detecting new and unknown fraud types. Is faster and more responsive than other models at detecting new and unknown types of fraud that are evolving minute-by minute, as well as keeping pace with subtle changes in existing fraud attacks, because it's continually being 'trained.'

  • Increases revenue and improves the customer buying experience. Enables you to accept a higher percentage of genuine transactions, reducing customer friction by significantly decreasing false positive alerts.

  • Ideal for CNP and high-ticket transactions. By modeling behavioral for each individual customer and not relying on averages across a representative sampling of all customers from a large consortium-based dataset, the scoring tool enhances the handling of CNP and high-ticket transactions.

  • Improves operational efficiency and reduces operating costs. Limits the need for manual intervention in managing false positives and can provide risk-score reason codes for each detected event to enable faster research by fraud analysts, helping to reduce operational costs.

  • Eliminates the need to build new models. Because Foresight Score model is naturally learning and adapting based upon individual customer behavior, the expense of periodically rebuilding models from a completely new set of data can be avoided.

  • Continuous learning. Supports self-learning with nightly batch updates, so the scoring tool will be able to continue to become faster and more responsive as fraud attacks evolve over ever shorter timeframes.

  • Highly reliable. Built on a highly resilient, fault tolerant architecture for the highest possible reliability to avoid leaving you unprotected and inconveniencing customers.

Does my data leave TSYS with this product?

No. While Featurespace is an outside vendor, all data stays within TSYS for a secure transaction.

How do I access and use the Foresight Score?

Once implemented, the Foresight Score can be accessed through the existing TSYS CardGuard ecosystem. You are able to determine how and where you would like to use Foresight Score to predict customer behavior and detect fraud.

What are the three top points I need to know about Foresight Score?

  1. Foresight Score combines machine learning and behavioral profiling to accurately predict what is and isn’t a fraudulent transaction — and it only gets more accurate as it observes more transactions and learns more about each customer.

  2. Foresight Score will save you money by effectively addressing the three primary components of a successful risk- and fraud management strategy: managing acceptable fraud losses and reducing false positives, thereby maintaining a satisfying customer experience and reducing operational costs.

  3. Foresight Score works within the existing TSYS fraud management system and provides the best overall results when used as a complement to existing consortium-based fraud scoring tools. And all of your data will remain secure with TSYS, as it always has.