Finding fraudulent activity amidst legitimate business transactions is like finding a needle in a haystack- you have to comb through millions of data points to find the few that are harming your business. These few needles could potentially be costing you millions.
Most solutions can only sift through and alert you to outliers, but what if you could know when and where that needle was dropped so you could catch it as it falls? What if you could not just catch it, but
predict, stopping fraud before it occurs?
The recent webinar,
“Unleash the Power of Predictive Analytics to Mitigate Risk and Fraud,” answers these questions with examples of innovative solutions achieved with IBM SPSS Predictive Analytics. Hosted by SPSS Portfolio Marketing Manager Luciane Ellis and Scott Mutchler, VP of Advanced Analytics Services at IBM Business Partner QueBIT Consulting, this presentation highlighted how predictive analytics techniques help businesses not only spot fraudulent transactions, but also identify deceptive vendor networks.
Not just a toolEllis kicked off the webinar by describing how predictive analytics is more than just a tool to wield: “Predictive analytics is a technique, it’s a concept, it’s a way of looking at data to uncover value. It’s connecting data to effective action and applying those insights throughout your business.”
Since the earlier you detect fraud the more you can save, being able to detect anomalous data in real time can help your business take action to mitigate fraud immediately. Employing predictive analytics to tap into structured and unstructured data to find fraud before it happens means your business can spend more time on building and maintaining strong relationships with legitimate customers and vendors.
How does predictive analytics detect fraud?The presentation walked attendees through how predictive analytics uses techniques such as identity resolution, business rules, segmentation, and predictive modeling to detect fraud.
Identity resolution is where unique persons, claimants, and entities (like vendors and companies) are identified, linking the relationships between identities. This can help identify those who are falsifying information and creating multiple aliases/companies.
Business rules are the requirements that you want to be met as part of your solution, such as identifying claims or purchases that are above a certain price point or frequency. These business rules can also combine with predictive analytics to give human understanding.
Segmentation is the grouping of similar customers and claims, which allows the system to build predictive models that are specific to customer, claimant, or vendor segments.
Predictive Modeling is the ground on which this solution is built, making predictions based on known cases of fraud in the past and detecting outlier behavior in real-time.
These are some of the most commonly-used fraud detection techniques, but IBM SPSS software provides a range of capabilities that can be applied to meet specific business needs.
Fraud detection in the real worldQueBIT worked with a mid-sized insurance company that had been targeted by organized fraud, and needed a predictive analytics solution that would enable it to:
• Automate claim review and identify fraud early in the lifecycle.
• Spot vendors who were performing fraudulent activity as well as suspicious claims.
• Explain exactly what was fraudulent and what was not.
By utilizing predictive models and applying business rules (such as dollar amount or frequency of claims by claimant), the company can now calculate a propensity score to identify the likelihood that any particular claim is fraudulent. A claim score of 89 percent would rank as a high probability of fraud, saving fraud investigators from manually examining each claim for possible deception by highlighting the most likely ones. The results would also include the specific reasons for why the claim may be fraudulent or not.
The IBM SPSS platform delivers these capabilities through a convenient graphical user interface that allows users to define business rules and easily build models that can look for patterns and trends in large volumes of data quickly. The models can be applied to past claims determined as fraudulent to find others with similar characteristics that are also likely to be suspect, enabling companies to predict where and when fraud might occur next, and who is likely to be responsible for it.
By identifying and blocking fraudulent claims, the IBM SPSS predictive analytics platform found $25 million in savings in the first year alone for the insurance company, significantly improving efficiency and reducing loss, cost and waste.
Want to find out more? View the on-demand webinar.See other presentations in the “Predictive Analytics in Action” series:
Democratizing data science: Advanced analytics as a team sport Retain valuable customers and grow your businessLearn how you can get started on your predictive analytics journey. #Frauddetection#Frauddetection#SPSSStatistics