What Causes Fraud Detection False Positives? Can They Be Prevented?
Picture this: you operate an upscale, luxury auto dealership. A person wearing dirty jeans and a ragged t-shirt walks in, saying they’re intent on purchasing a top-of-the-line sports car.
How would you react? Would you automatically turn them away, just based on how they look? Would you let a $250,000 sale walk out the door because you simply don’t believe the customer could be legitimate?
Now, imagine doing that with 30%, 50%, or even 70% of your potential buyers. Sounds ridiculous, right? Well, the truth of the matter is that the majority of eCommerce merchants are losing that much every year to fraud detection false positives.
In this post, we’ll look at false positive rates in the eCommerce space. We’ll explore how false positive fraud detection issues impact online retail, and what you can do to protect your revenue.
Recommended reading
- Card Security Codes: How They Protect Consumers & Merchants
- The Top 10 Fraud Detection Tools You Need to Have in 2024
- ECI Indicators: How to Understand 3DS Response Codes
- Proxy Piercing: How Merchants Can Use it to Prevent Fraud
- Card Verification Values: What Are CVVs & How Do They Work?
- Payment Authentication: How to Verify Buyers Before a Sale
What are False Positives in Fraud Detection?
- False Positive
False positives are results which show a certain condition is present when it actually is not. In fraud detection, for instance, this typically means flagging a legitimate transaction for potentially being fraudulent.
[noun]/fôls • pəs • ə • div/False positives can be an issue in almost any industry. In medicine, for example, a preliminary test might return inaccurate results, necessitating additional rounds of tests to verify results. In fraud detection, a false positive happens when you incorrectly block a legitimate transaction because the sale seems like potential fraud.
It’s possible to manually review all transactions, but this demands substantial time and resources. The problem with computer-based systems, though, is that they are making assessments based on rules or algorithms. Like any other computer program, bad input will lead to inaccurate results. Merchants may overcompensate, setting tighter parameters that catch legitimate sales.
Fraud detection software makes decisions based on rules, typically created by the user. More savvy systems use machine learning to compare hundreds of data points and make decisions based on a bigger, more complex picture.
Click here to learn more about machine learning in fraud detection.
This is a more common problem than you might suspect. Of course, payment fraud is a real and growing threat, costing US merchants an estimated $5.72 billion in 2022. That said, merchants lose 13 times more to false positives than they do from true fraud.
Investing in fraud detection tools and practices is important. But, with $13 lost to false positives for every dollar in genuine fraud, it’s clear that merchants need to look at better ways to balance losses against risk.
What Causes Fraud Detection False Positives?
There are a number of automated fraud detection solutions on the market. Most of them can do what they claim to do with a high level of competence.
Automated systems, however, typically have at least some parameters that must be set by the merchant. If those parameters aren’t correctly calibrated, the tool will end up flagging or declining sales simply because one element — say, the customer’s location, for instance — seems suspect.
Suppose you have a customer that attempts to make a purchase while on vacation. The purchase is legitimate, but the system could kick it back because of geolocation or IP address mismatches, suggesting the sale could be fraudulent.
Location is hardly the only criteria for attempting to catch fraud preemptively. Fraud detection software can have customizable parameters for a number of different factors, including:
- Number of transactions within a set period of time (transaction velocity)
- Number of transactions from the same IP address
- Customers with non-matching billing and shipping addresses
- Sales that vary significantly from the customer’s purchase history
- Customers with CVV codes that don’t match the card
- Orders placed from a banned IP address
In short: inaccurate interpretation of information is triggering fraud alerts and leading to good sales being rejected. At the same time, true fraud could be evading detection due to the same miscalibrated rulesets. Both situations cost you money and resources, as we’ll elaborate on more in the next section.
Why Are False Positives a Problem?
As we mentioned before, you’re rejecting legitimate orders. That’s a tangible, measurable loss, but the costs go deeper than that.
Each false positive, either manual or automated, also results in a long-term revenue hit. You could lose a valued customer, or lower the customer lifetime value (CLV) of an existing buyer. According to Forbes, one in three shoppers receiving a false decline say they will not return to the same merchant.
No single one of them – or even a combination of multiple indicators – is a guaranteed sign of a bad transaction. Even sophisticated fraud management solutions can react to the wrong triggers if not given the correct parameters.
Perhaps worst of all, the problem can actually cause a feedback loop. As we pointed out earlier, many fraud detection models rely on machine learning to increase accuracy over time. Collecting bad data, however, will accomplish just the opposite.
You have bad indicators, leading you to deploy tools in an ineffective way. New transactions are judged by inaccurate criteria, opening the door to more false positives, which leads to more bad benchmarks. The system creates a feedback loop that continually adds corrupted data, so the problem gets exponentially worse over time.
How to Avoid Fraud Detection False Positives: 8 Essential Best Practices
It should be obvious by this point that eliminating fraud detection false positives is a crucial operation for any comprehensive fraud management strategy.
A mix of techniques are required here. Finding the best approach will be based on your situation (location, vertical, sales model, etc.) That said, here are a few suggestions that are generally focused on helping you stem revenue losses due to false positives.
Vet Customers Based on Multiple Indicators
Using a mix of methods to validate customer information can lower the odds of false positive fraud. Prioritize multi-factor authentication. Validating CVV numbers, using geolocation, address verification, and deploying other tools in a coordinated manner may help fraud filters decide borderline cases more effectively.
Establish a Baseline
You need to have an honest, realistic understanding of your capabilities regarding fraud detection and preventing false positives. If you conduct manual reviews of flagged transactions, for instance, try to optimize rules so as to not generate excessive alerts that might overwhelm your team.
Segment Risk Factors
Not all indicators are created equal. Some transactions will obviously carry greater risk than others. You will want to spend more or less time reviewing these transactions based on dollar value, buyer profile, and other factors. A transaction with several “high risk” indicators should be more of a candidate for automatic rejection than a lower-risk sale that does not fly as many red flags.
Listen to Feedback
Customers can feel insulted if legitimate purchase attempts are blocked or flagged. Use real customer feedback — especially any complaints or negative reviews you receive — to help you see your fraud detection results in a different light. Compare multiple cases, looking for anomalies and similarities. Try to find ways to streamline and improve your customer experience.
Deploy the Most Current Solutions
Software, powered by AI and machine learning, can compare more data points and analyze trends more quickly than manual reviews. The technology isn’t perfect, but it’s getting better all the time. That’s why it’s important to implement the most current solutions, and to ensure you’re on top of updates and new plugins, to ensure they’re up to date.
Review Fraud Detection Systems
Remember: fraud detection is not a “set-it-and-forget- it” process. In the case of false positives, it requires ongoing monitoring, comparing chargebacks received against transactions denied or flagged. Fine-tune settings that may be too broad or too strict.
Ongoing Testing
You need to test new rules and adjust over time. Before rolling out new rules in a live context, though, it’s wise to test them in a sandbox setting. Subject orders to multiple different rulesets, evaluate the effect on your false positive rate, and adjust the parameters as needed.
Avoid “Chargeback Guarantee” Tools
Some fraud solutions are more focused on limiting chargebacks than balancing risk. Often, these tools are calibrated to be overly protective so as to lower the chance of chargebacks. If your priority is to stop chargebacks without seeing a surge in false positives, there are solutions at your disposal.Eliminating False Positives: One Step in the Process
Unfortunately, there is no “silver bullet” for reducing false positive rates. The nature of the threat makes it a moving target. All online retailers have to deal with at least some false positives. If you’re not, there’s a good chance you’re about to see a high volume of chargebacks with “fraud” reason codes come pouring in.
The bottom line: true fraud and chargeback prevention requires a more comprehensive approach.
Chargebacks911® offers the only end-to-end technology platform that prevents disputes, wins reversals, and maximizes your ROI, all while providing better data insights to help you eliminate fraud detection false positives. Contact us today to get started.
FAQs
What are false positives in fraud detection?
False positives are results which show a certain condition is present when it actually is not. In fraud detection, for instance, this typically means flagging a legitimate transaction for potentially being fraudulent.
What is false positive and false negative in fraud detection?
A false positive marks a transaction as fraud when it actually is not. In contrast, a false negative is where an actual fraudulent transaction is passed as legitimate. For comparison, true positives correctly show that a given condition is present.
How can false positives in credit card fraud detection be reduced?
The most effective method is a machine learning solution that analyzes trends in data (as opposed to a static, rules-based approach). That said, any automated solution will need ongoing monitoring and adjusting to maintain the balance between fraud prevention and losses to false positives.
What is the formula for false positive rate in fraud?
You can find your false positive rate by taking the total number of “positive” responses generated by fraud detection software, and dividing it by the number of false positive responses generated in the same period.
What are examples of false positives?
From a fraud detection/prevention perspective, false positives are legitimate user actions flagged or denied based on suspicion of criminal activity. Examples would be a legitimate transaction being declined, a failed account log-in, or a valid sale being canceled.
