The Green Sheet Online Edition
September 28, 2015 • Issue 15:09:02
Payment fraud prevention through predictive analytics
Feedzai, a data science company established in 2011 with offices in San Francisco, London and Lisbon, develops software tools designed to predict and detect payment fraud before it happens. Feedzai said its curated blend of big data analytics and security best practices has helped its global clientele in the Americas, Europe and Africa identify security vulnerabilities within their existing infrastructures in order to prevent sensitive data from being attacked or compromised.
Feedzai provides an array of resources, insights and organized data to acquirers, issuers, retailers and e-commerce providers. These enhanced reporting tools derived from sophisticated big data mining software and machine-learning algorithms can help teams make informed decisions and then take immediate action to protect and manage critical security and data, Feedzai noted. Reseller opportunities and a developer portal are available.
Fraud prevention tools
Feedzai's risk and fraud software uses intelligent algorithms to examine individual customers as opposed to relying on sampling and segmentation. "Traditional risk mitigation tools lack the analytics resolution needed to distinguish fraudsters from 'friendlies', and cause your frontline operations to mistake good customers for criminals," the company stated.
Big data solutions
The combination of machine learning with human intelligence can provide an in-depth view of live data streams that can be matched against historical data to expose suspicious and irregular activities. Feedzai big data solutions also include:
- Live dashboard interface with hyper-granular data resolution
- Highlighted trends that emerge from live data streams and terabytes of historical data, requiring further analysis
- Sophisticated machine learning algorithms that constantly update predictive models
- Email and SMS alerts for analysts and management that enable them to take immediate action
Risk management solutions for acquirers
Feedzai merchant underwriting solutions are designed to not only prevent fraudulent transactions but also to predict merchant insolvency, collusion and unfunded chargebacks. In the ever-changing world of merchant acquiring, risk profiles can shift rapidly, making it challenging to predict outcomes and update decision rules.
Feedzai provides continuous underwriting oversight to detect unusual patterns that may signal merchant underwriting risk. Configurable business analytics dashboards combined with key performance indicators (KPI) can detect unique data elements in a data stream that can be further analyzed and benchmarked against peer group, season and historical patterns, the company noted.
Examples of available KPIs include:
- Volume and velocity of use
- Excessive changes in sales and refund volume
- Average transaction size
- Chargeback rates by merchant location
- Multiple sales with the same card number
- Suspicious descending sales authorization attempts
- Key entered transactions
- Forced transactions
- Draft retrieval requests
- Credit refunds or reversals
- Large individual transactions
- Rates of card-not-present and international activity
Online fraud prevention
Feedzai provides machine-learning algorithms to retailers through a secure set of application program interfaces (APIs). As the algorithms of the APIs become familiar with the business, they begin to create a series of rankings and explanations related to customer orders. Scores ranging from 0 to 1,000, with customizable cut-off points, are assigned to each order at checkout.
The APIs simplify merchant processing and shopping cart integration, Feedzai said. Merchants, marketplaces, e-commerce system providers and other e-commerce stakeholders can tap into the Feedzai API to facilitate safe and secure online selling.
In-store fraud prevention
Feedzai feels that sophisticated loss-prevention algorithms make omni-channel selling safe. Feedzai Fraud Prevention for retailers includes a detailed evaluation of every sales transaction with an accompanying risk score. Orders are protected during buying and checkout using machine-learning models. Real-time scoring is based on buyers' spending patterns accumulated over several years. Adaptive models detect "first-use" fraud where there is no transaction history.
Notice to readers: These are archived articles. Contact names or information may be out of date. We regret any inconvenience.