Fraud is a universal challenge that is constantly evolving, meaning merchants and issuers must always stay one step ahead. The complexity of threats continues to escalate, too. And when every transaction carries a risk, each split-second decision carries the question: is this transaction legitimate or fraudulent?
The volume of global digital commerce requires data-driven oversight, not manual guesswork. The cost of getting it wrong can be huge. In the U.S. alone, cyber-enabled fraud losses amount to about $13.7 billion a year.1 Meanwhile, declining valid transactions — known as false declines — to avoid fraud is costly. Optimizing their authorization rates is now rated as “extremely important” by more than 40% of merchants globally.2 They must find ways to combat sophisticated attacks without adding more friction that will ultimately drive customers away.
Instead of waiting for a crime to happen and fixing it later, advanced analytics let you identify elevated risk earlier in the transaction lifecycle. Tools like machine learning and generative AI act like digital detectives, scanning billions of transactions to learn exactly what good and bad spending looks like. This massive visibility helps inform your decisions and interventions, like blocking scammers instantly in real time, while rolling out the red carpet for legitimate customers — closing the gap between risk and revenue.
When it comes to fraud risk scoring and analytics, VisaNet processes more than 269 billion transactions annually,3 offering a massive dataset for benchmarking and detection. Visa has an added advantage in being positioned on both sides of the transaction, allowing us to connect merchants and financial institutions to share data securely. This means that financial institutions have access to more higher-quality data to feed into their models, allowing them to more accurately calculate important analytics like fraud risk scores.
By understanding how fraud scoring works and how it can help reduce fraud, you can help stop bad transactions, approve more good ones and ensure a frictionless experience.
Fraud scoring is a way to rate transactions based on their level of risk. Calculated in real time, it ensures that each payment is evaluated to detect the likelihood of a customer being genuine, risky or fraudulent. A numerical risk value, or score, is used — from 0 (very safe) to 99 (high risk).
So how does fraud scoring reduce payment risk? Utilizing automated AI models, fraud scoring happens seamlessly behind the scenes within milliseconds, and the decision is used to identify genuine transactions while declining fraudulent ones. This not only helps to reduce fraud but also streamlines the customer experience by keeping every payment as frictionless as possible.
There are two common approaches used together in fraud risk management: rules-based controls and risk-based, model-driven scoring. Machine learning models analyze transactions and generate risk signals that reflect changing customer and ecosystem behavior. Rules then apply thresholds or policies to those risk signals to inform what action to take, such as allowing a transaction to proceed or triggering additional checks.
Generative AI takes fraud scoring and analytics a step further and helps merchants and issuers stay at the cutting edge in terms of countering emerging threats posed by fraudsters. Generative AI can support investigators by summarizing risk signals and historical context, helping increase efficiency in review and documentation.
Fraud scores help inform decisions, which are configured by issuers, acquirers or merchants — and ultimately lead to individual interventions. This is why having more reliable, better-quality data is so important.
Different types of data are used to calculate a score, and they can carry a different weighting, too. For example, an unexpected customer location is likely to matter less than multiple failed transaction attempts. And because businesses all look at the data differently, depending on factors like their risk appetite and what models they use, the same transaction can receive multiple scores. Additionally, having access to rich, accurate data that allows for continuous learning is increasingly valuable, so your business is able to adapt to the latest methods used by fraudsters.
Featurespace, a Visa solution, uses Adaptive Behavioral Analytics and Automated Deep Behavioral Networks to provide the insights you need, including early detection of scam patterns, so you can take the decisions that are right for your business.
Here’s an overview of how Visa and Featurespace’s fraud scoring solutions calculate fraud scores and help your business balance risk:
- For merchants, transaction control through intelligence: fraud scoring takes place before the authorization stage. This process is like an automated gatekeeper, analyzing data points including device intelligence, geolocation and transaction history to filter out bad orders before they reach the bank. Merchants can take control with Visa solutions such as Decision Manager (DM) and Visa Protect Risk Insights (VPRI).
- For issuers, making the decision: the scoring element is used to approve or decline a transaction and this takes place during authorization. In mere milliseconds, tools such as Visa Advanced Authorization (VAA) analyze the probability of fraud to help the bank approve the transfer of funds between the two parties. Analytics can provide dynamic visibility to behavior and anomalies. Financial institutions use dashboards to compare their fraud rates against industry peers and regional benchmarks. Identify specific spikes in fraud types, such as card-not-present versus card-present or high-risk merchant category codes (MCCs).
- The data engine: risk scores are generated by comparing the customer’s current transaction with billions of historical network transactions to establish validity.
Pioneered by Featurespace, a Visa solution, Adaptive Behavioral Analytics (ABA) are industry-leading models where a customer’s transactions are analyzed to recognize patterns of “good” or “normal” behavior and therefore identify suspicious or abnormal behavior. By understanding what “good” transactions look like rather than simply flagging “bad” activity, ABA helps financial institutions catch scams as they surface, improving fraud detection while reducing false alarms that frustrate customers.
| How the data engine works | How it works for merchants | How it works for issuers |
|---|---|---|
| Scores are generated in milliseconds. | Scoring occurs before authorization. | Scoring occurs during authorization. |
| The current transaction is compared with billions of historical network transactions. | It acts as an automated gatekeeper, analyzing data points such as device intelligence, geolocation and history. | Tools like Visa Advanced Authorization (VAA) analyze the probability of fraud. |
This helps to establish validity by creating a fraud score. ABA helps financial institutions catch scams as they surface, improving fraud detection while reducing false alarms that frustrate customers. | Merchants identify good transactions and decide which ones reach the bank. | This helps the bank approve the funds transfer. |
“Payment fraud undermines customer trust and brand integrity. The most successful businesses use AI to turn fraud prevention into a growth enabler, approving more legitimate transactions with confidence.”
Michele HerronHead of Value-Added Services, North America, Visa
- Shift to automation: Merchants, issuers and financial institutions are now able to automate decisioning by shifting from manual review to machine learning-driven scoring — which can lower manual workloads by 25% or more.4
- Multi-player defense: Merchants can screen orders at checkout to increase approvals and help with chargeback liability; financial institutions, meanwhile, are able to use scoring to optimize their authorization strategies and monitor risk in real time.
- Leveraging the Visa EMV 3-D Secure solution: With the 3DS solution, data can be shared between both merchant and issuer, which improves the accuracy of the risk score. This has been shown to lead to a 9% lift in approval rates and a 45% reduction in fraud.5
- Utilize consortium data: Decision Manager’s Unified Consortium Model gains insights from billions of global transactions from across networks and merchants to deliver enhanced risk-scoring accuracy.
- Implement AI/ML scoring: Merchants can adopt predictive scoring to reduce manual reviews and increase automation. Visa can analyze network-wide patterns to help your business score authorization requests in real time.
Visa Provisioning Intelligence (VPI) for banks and issuers is a smart, proactive tool designed to support real-time risk assessment. It assesses requests when someone tries to add a payment credential to a digital wallet or device, a process known as “provisioning.” By assessing the probability of fraud during the token provisioning phase, VPI provides real‑time risk scoring that helps banks and issuers identify potentially fraudulent provisioning activity. This enables issuers to block high‑risk token requests before a digital token is created, adding an important layer of security at the point of entry.
Visa Consumer Authentication Service (VCAS) for issuers is a data‑driven, hosted solution designed to make online shopping safer and smoother. It supports issuers by managing their authentication strategies within 3-D Secure programs. By leveraging enriched data and a dedicated VCAS risk score, the service enables issuers to apply real‑time, rule‑based decisioning to determine whether transactions can proceed frictionlessly or require additional authentication. This helps enable seamless, risk‑based authentication and delivers a simple and secure digital payment experience.
Visa Deep Authorization (VDA) for financial institutions is one of the risk-scoring tools specifically tailored for the card-not-present landscape. Using AI and deep-learning models, VDA analyzes historical cardholder and merchant transaction behavior to model evolving spending patterns. This technology generates a transaction risk score during the authorization process, helping issuers enhance fraud detection and decisioning. Currently, this powerful solution is available for the U.S. market.
The Featurespace Platform is an AI-powered fraud prevention and financial crime detection platform with a suite of solutions designed to help financial services companies prevent fraud. It evaluates payment transactions, provides instant risk scores and enables immediate intervention across a wide range of payment types, including credit and debit cards (card-present and card-not-present), ACH, Faster Payments, wire transfers, checks and person-to-person (P2P) transfers.
Using its proprietary Adaptive Behavioral Analytics (ABA), Automated Deep Behavioral Networks powered by machine learning, and Large Transaction Model (LTM), the Featurespace Platform helps identify risk, catch new fraud attacks and identify suspicious activity — to include fake account openings and complex merchant fraud — in real time, all while ensuring legitimate transactions continue to flow smoothly.
Visa A2A Protect for financial institutions applies adaptive, real‑time risk scoring to help assess fraud and scam risk in account‑to‑account payments. Designed to deliver value from day one within a single institution, it uses each bank’s own transaction data and customer behavior to establish a baseline of normal activity and identify anomalous real-time payments (RTP) activity as payments are initiated. This supports faster, more informed responses to emerging risks without relying solely on manual review and while keeping control firmly with the institution.
When institutions choose to opt in, Visa A2A Protect can also incorporate intelligence across banks, to enrich those risk signals with broader contextual insight. This additional layer helps financial institutions understand how emerging scam patterns may be evolving beyond their own environment, supporting more resilient and informed fraud strategies over time.
Visa Protect Authentication Intelligence (VPAI) for banks and issuers provides a risk-based authentication score to help assess fraud risk in online payments. A key tool within the Visa authentication suite, it’s designed to enable seamless, data-led decisioning during authentication. By using this intelligence, banks and issuers can better determine whether an authentication request can proceed frictionlessly or if it requires additional verification — helping deliver a simple and secure digital payments experience.
Visa Analytics Platform (VAP) for issuers is a data-driven analytics platform that provides industry benchmarks and insights to support fraud monitoring and optimization.
Visa Advanced Authorization (VAA) for issuers is one of the tools that generates a transaction risk score by analyzing network‑level data to identify emerging fraud patterns.
Visa Risk Manager (VRM) for issuers is one of the tools that applies configurable, rule‑based controls to transaction risk scores to support authorization decisioning. For example, VRM can be used alongside Visa Advanced Authorization (VAA), which generates a transaction risk score during authorization.
Decision Manager for merchants automates fraud detection using advanced machine learning and data-driven insights from billions of transactions.
Staying on the lookout for fraud is a constant battle but with Visa fraud scoring and analytics solutions you can use data as your secret weapon — saying no to scammers, while saying yes to more genuine customers.
Featurespace x Eika
Before implementing Featurespace’s platform, the Eika team relied on mainly manual checks and email notifications. Featurespace helped Eika Gruppen reduce its phishing losses by almost 90%,6 bringing all customer transactions into a single, comprehensive view and helping them automate processes across their fraud monitoring.
Featurespace x NatWest
Featurespace delivered its real-time fraud detection Platform with adaptive machine-learning models as NatWest’s first line of defense. The Platform improved the value of fraud detected by 57% and the value of scams detected by 135%, while reducing the overall false positives by 40%.
How do enterprises integrate fraud scoring with payment systems?
By using a solution like the Visa Acceptance Platform, businesses can establish a single connection that links their payment processing directly to fraud prevention tools. This allows them to turn specific capabilities on or off — like tokenization or account protection — without having to rewire their entire system every time. Once connected, enterprises can integrate tools like Decision Manager or Visa Advanced Authorization (VAA) to analyze transactions in real time during processing.
How accurate are fraud scoring models for online payments?
Visa is able to draw on its vast network of payments data to understand buying behaviors. When using authentication technology like Visa Secure with EMV 3DS, fraud can fall by 45% compared with non-authenticated transactions.7
How do financial institutions benchmark fraud scores?
Financial institutions benchmark their fraud scores effectively by comparing their own vital signs against vast networks of global data and the performance of their peers. It's like a health check-up for their payments system to ensure they aren't being too strict (declining good customers) or too lenient (letting fraud in).
They use powerful tools like the Visa Analytics Platform (VAP) to access industry benchmarks. This allows them to compare their specific stats — such as authorization rates, fraud rates and liability rates — against similar institutions across different channels and geographies.Can fraud scoring reduce chargebacks for merchants?
Yes, absolutely. Think of fraud scoring as a highly trained digital “bouncer” for your online store. By helping to spot bad transactions at the door, you can then prevent the very sales that would eventually turn into costly chargebacks. Fraudsters are getting smarter, but so are the scoring models.
What is the difference between a credit score and a fraud score?
Think of a credit score and a fraud score as two different types of “background checks” that happen at very different times and for very different reasons. While one looks at your financial health over years, the other is looking at a single moment in time.
Credit score: measures your financial reliability. It asks, “If we lend this person money, will they pay it back?” Financial institutions use it to set credit limits or approve loans based on your history of repaying debts.
Fraud score: measures transaction risk. It asks, “Is this transaction actually being made by the cardholder, or is it a thief?” It calculates the probability that a specific purchase is illegitimate based on patterns and behaviors.What is fraud analytics and how do enterprises use it?
Fraud analytics is the process of using sophisticated data analysis to spot suspicious patterns and help businesses make smarter decisions in real time, which can ultimately lead to them blocking bad transactions. While the AI does the heavy lifting to identify risky transactions, businesses can still set their own rules. For example, a tool might suggest a rule based on recent data, like “Block all orders over $500 from this specific suspicious IP address”, allowing the merchant to adapt quickly to new threats. Businesses aim to lower their false positive rate — when a legitimate customer is accidentally declined. By using reliable data, they can ensure fewer loyal customers get rejected at the checkout, keeping satisfaction and revenue high.
How do banks apply machine learning in fraud analytics?
Unlike old models that need to be updated manually, modern machine learning models are self-adapting. They constantly ingest new data (like truthful reports on chargebacks) to learn about emerging threats — such as adaptive malware or new scams — and update their defenses automatically. By looking at billions of transactions across the entire payment network, these models can spot a new type of attack hitting one bank and instantly learn how to protect others from it. Beyond reducing fraud, these tools recognize positive behaviors to help banks identify and roll out the red carpet for their best, most trusted customers.
What’s the difference between fraud analytics and fraud detection?
Think of fraud detection as the security guard at the door, and fraud analytics as the strategist in the back office reviewing the stats. While one acts in the heat of the moment to detect a specific crime, the other looks at the big picture to improve the whole system.
Fraud detection assigns a specific risk score (often 0 to 99) to an order. If the score is too high, the detection system may decline the transaction to prevent loss.
Fraud analytics looks at wider patterns, trends, and the overall picture after transactions have occurred. It takes all that messy data and turns it into a report card for your business.
See how Visa fraud management solutions can help support your business
Sources/Footnotes/Disclaimer
- FBI. (2024). FBI Internet Crime Report 2024.
- The Merchant Risk Council (MRC), Verifi, Visa Acceptance Solutions, and B2B International. 2025 Global eCommerce Payments & Fraud Report. (2025).
- Visa. (2023). VisaNet data (fiscal year 2023).
- Based on data collected from Decision Manager clients moving to actively using Decision Manager’s Identity Behavior Analysis.
- As of Visa Q4 FY23, authenticated transactions saw 11 bps of fraud whereas non-authenticated ecommerce transactions saw 20 bps of fraud (45% reduction in fraud between authenticated and non-authenticated ecommerce).
- Coull, Gavin (Featurespace - SME), Macfarlane, Alasdair (NatWest - Fraud Prevention COE Digital X). (2023, March 1). NatWest at the frontier of scam detection.
- Visa. (2023). VisaNet data: 3DS Visa Secure (2023).