KTC prioritizes customer safety with ACI intelligence
One of Thailand’s largest non-banks and its best-known credit card provider, Krungthai Card PCL (KTC), is recognized across Southeast Asia for championing next-generation financial products and services. To protect its wide network of over 3.3 million account holders, KTC leverages ACI payments intelligence integrated with AI to combat fraud, minimize costs, and deliver incremental revenue.
The challenge
Elevate fraud detection in the face of rising levels of sophisticated threats
Prioritizing customer safety, KTC adopted AI models to combat new and increasingly sophisticated threats – from bad actors, high-risk merchants and untrusted merchant category codes/countries (MCC) – in real-time and at scale. The solution also had to distinguish unusual amounts and activities at the customer level to reduce the rates of false positive alerts.
The challenge was to build smarter safeguards without introducing more complexity, coding, or user friction. A smart solution offering an invisible layer of security was needed, one that could be rolled out quickly without requiring extra resources, new system integration or additional skills training.
The strategic decision to harness ACI’s broader payments intelligence portfolio provided KTC with access to the latest multilayer AI technology for advanced, real-time contextual and behavioral decisions.
KTC could easily utilize ACI’s model generator, which creates unlimited adaptive machine learning models quickly and intuitively in just a couple of hours. Offering an easy-to-use intuitive interface, ACI’s model generator enables KTC fraud teams to pinpoint assertive and strong signals of fraud without any complex coding or advanced data engineering skills.
Enhance customer experiences using automation and richer profiles to reduce false positives
In collaboration with ACI, KTC developed a dynamic feature library for multiple use cases. Utilizing powerful tactical scoring models that can be analyzed, trained, tested, and compared within KTC’s familiar proactive risk manager and ACI’s model generator user interface the library includes:
Transactional features — Values of MCC, country, currency, amount, hour of a transaction
Velocity features — Time since last fraud at this merchant, time since last transaction at this card, transaction count at the card in the previous X hours, sum of purchase amounts at the card in the previous several hours
Habitually (aka familiarity or trust) indicators — Genuine customer profiling to reduce false-positive rate: merchant binary checks if the customer has used this merchant in the past; and similarly: MCC/currency/country/postcode familiarity features
Fraud profiling — To increase fraud detection rates – for example, the merchant fraud rate feature is defined as the number of fraudulent transactions divided by the number of all transactions at this merchant in several days – fraud rates could be constructed at the MCC, acquirer, country, currency, and postcode levels
These features reduce effort required of the KTC fraud team. The resulting models automatically score and flag transactions above particular thresholds. As a result, ACI’s model generator rules reduced false-positive rates by filtering out the previously established and trusted spending patterns of genuine customers.
Existing ACI assets accelerated next-generation fraud capability with minimal investment
No extra hardware, software or complex integrations were required, as KTC was able to use its established ACI fraud platform to implement payments intelligence solutions. This led to the rapid adoption of machine learning models without the cost and effort typically associated with risk management transformation.
“ACI’s multilayer AI platform revolutionizes our fraud detection without overburdening our fraud team. Payments intelligence provides ready-to-go tools to tackle the most sophisticated CNP and scam attacks. We are already detecting 1.5 times more cash-out scams than before and can detect fraudulent transactions with optimal false-positive and high detection rates.”
Rywin Voravongsatit Vice President – Head of Operations Control & Merchant Operations Business Unit, KTC
The solution
In just two weeks, ACI’s team developed multiple machine learning models and a comprehensive library within ACI’s model generator of 50 dynamic features to detect and learn emerging threats, and significantly improve KTC’s fraud detection capabilities.
By reducing manual coding and processes, ACI payments intelligence has freed KTC’s fraud team to focus on developing new detection logic and adjusting existing strategies.
A key advantage of the ACI solution are the high levels of automation, including:
Ready-to-use features: Speeding fraud detection and updating models – every merchant now has its self-adjusting false-positive rate, eliminating the need to maintain merchant risk lists or set manual rule thresholds.
On-going refresh: Spotting fraudulent patterns and ready for retraining features daily, eliminating time-consuming manual processes.
Feature and logic threshold adjustment: Accelerating changes across all functional layers – KTC can adjust parameters to improve risk management and operational control.
Outpaced 85% fraud detection rates: KTC has exceeded its goals and created a more secure payments experience for KTC customers.
Delivered a 3:1 precise false positive ratio: Enhanced consumer profiles have resulted in fewer false positives, improved operational efficiency, and smoother customer experiences.
Targeted scope for CNP overseas fraud: Customized models successfully address the challenge of detecting cross-border fraud as consumers shift more of their online spending to global merchants.
1.5 times more effective cash out scam detection: Combining transactional, behavioral and contextual data has increased cash-out scam initial detection rates from 33% to 50%.
Modern, agile and responsive fraud operations: Dynamic features enhance fraud prevention and provide invaluable insights into genuine customer behavior.
Data aggregation delivers greater insight, agility, and customer centricity across KTC’s enterprise.
Model development can be managed by both internal business owners, without the involvement of data scientists. Test results are easier to understand, analyze, and act on, speeding up KTC’s ability to implement new features and manage fraud and customer services more effectively.
The real-time fraud platform provides KTC with an accurate single source of truth about customers and their typical behavior to determine the best course of action to keep user experiences optimized with fewer false positives. By creating separate scopes and models for major merchants, KTC can also more effectively combat overseas CNP fraud.
KTC has achieved its ambition of accelerating AI-powered, dynamic and generative models to fight new fraud patterns as they happen, creating more effective fraud strategies to mitigate risks and safeguard brand reputation and profit.
With ACI’s support, KTC has removed time-consuming data work and turned the fraud function into a valuable enterprise-wide asset capable of maintaining its competitive advantage and improving user experiences across existing and future products and services.
CustomersPartners
