Retail Needs A New Approach To Fraud: Enter Artificial Intelligence
We’ve seen many new fraud and authentication techniques and point solution providers enter the market over the past several years, but have we really made any improvement against fraud, particularly in the payments space? Apparently not.
Account Opening Fraud
Account opening fraud is a rapidly increasing challenge for issuers due to the plethora of identity data available to fraudsters. The 2018 Identity Fraud Study by Javelin Strategy & Research shows that the number of identity fraud victims increased by eight percent in 2017, with the amount stolen totaling $16.8 billion.
Account Takeover Fraud
Account takeover, where a fraudster gains access to a victim’s account, typically leads to unauthorized fraudulent transactions. Account takeover fraud (ATO) is still trending upward, especially in the financial services sector. According to Javelin, existing account takeover fraud tripled in 2018 to 1.5% of all US-based consumers.
Key Gaps in the Fraud Ecosystem
Some of the top financial institutions employ specific and often expensive point solution providers for device risk, behavioral risk, mobile phone intelligence, social reputation, email reputation, call center fraud defense, bot and malware detection. And each of these providers typically provides a risk score or a rules-based approach, and a potentially long list of data attributes.
But this approach creates an issue and an opportunity. It isn’t necessarily a bad investment to add new point solution or data providers as long as you are getting value out of these investments. However, that is often the hardest determination to make. Personally, I have worked in large organizations where we invested in the latest and greatest data source, but never fully realized the potential of what we had purchased.
To complicate this further, often there is no cross-channel communication for authentication or identity decisions. Each channel is working in its own silo as is each point solution provider with their scores and rule sets. And from a consumer view, they must prove their identity or authenticate across different lines of business or for different types of transactions. It starts to be obvious why we are losing the battle against fraud with this approach. Many systems, many rules, many scores, no central decision maker or analysis to determine the optimal blend of accurate decision, cost, and performance.
With several point solution providers, each with its own rules and scores, there is often no unified approach across all of these to make a final real time decision on a given use case. There needs to be a single decision system that can ingest all the data from these platforms into a single risk analysis, at a specific point in time, for a specific use case. And that system can determine when call outs for additional data are necessary.
Artificial intelligence – Creating a competitive edge
Artificial intelligence (AI) is creating a competitive advantage for early adopters, by adapting faster to new fraud techniques and creating a better customer experience resulting in more approvals and fewer false declines.
In addition, AI can ingest existing large data sets and point solution provider scores, and in turn provide enhanced real time decisioning. Specifically, to combat multiple types of fraud while balancing false declines, AI is able to solve for the typical silo-based view of customer interactions across channels and the industry, connecting authentication, device, behavior, and transaction data for modeling or rule creation across customer interaction points.
Use Artificial Intelligence To Take Stock In What You Have
It’s possible you lack the real-time ingestion of all of these scores, rules and data sources to provide more insight into consumers, merchants, and devices.
And, what is also often the case is the lack of intelligence to take action on that data. Given whatever data signals and risk systems you have in place, you should use AI to analyze that data and apply historical known fraud or failed authentication attempts against those scores and data elements.
Apply tools such as artificial intelligence to assess what data elements and risk scores correlate most significantly to fraud. You are essentially using the power of AI to determine what data is useful and what is not. As you assess the correlation of risk scores to each other, you may find that some point solution providers are adding very little lift to the decisions, which can result in a cost savings by offboarding these vendors. In the end you need to determine the appropriate set of authentication mechanisms given financial costs and benefits, particular to each use case and transaction type.
And of course, AI is able to incorporate and analyze new data sources to determine if they provide more insight into risk decisions. But before adding another point solution provider, another risk score, another data source, have you made an honest assessment and exhausted the data you already have? Before you add another point solution provider or data source, can you test the providers “lift” over your other sources in a proof of concept using AI? And can you create a flexible contract structure to only leverage certain data sources when they have proven to add to a quality decision?
It’s easy to be attracted to the allure of a new innovative sounding solution when you are experiencing fraud issues. But often the issues are foundational, a lack of integration, a lack of end to end data analysis, and a lack of knowledge of how to optimize the data you already own and understanding the new data sources that truly provide the ability to mitigate fraud.
-Michael Lynch, Chief Strategy and Product Officer, Deep Labs