The Basics of Fraud Scoring: How Quantifying Your Risk Helps Prevent Fraud
Every eCommerce transaction is unique. At the same time, many elements of a transaction will remain consistent for single buyers.
Typically, indicators like IP address, ZIP code, and shipping preferences remain more or less constant from one purchase to the next. If an order comes in that completely bucks an established buying pattern, savvy merchants know to take a closer look, as it may be a case of fraud.
What about orders that straddle the fence, though? Orders that raise a few flags, but maybe not enough to reliably say it’s a case of fraud? You have to make a judgment call. But, even that can be very tricky.
If you’re too lenient, you may end up accepting a fraudulent order. If you’re too careful, you end up rejecting legitimate customers.
Fraud scoring can help you make smart, consistent choices about whether to accept or reject orders. This article will examine how fraud scoring works, and weigh some of the pros and cons of using this key fraud prevention tool.
Recommended reading
- What is 3-D Secure? Fraud Prevention Solution Explained
- What are Velocity Checks? How Do They Stop Fraud Attacks?
- ECI Indicators: How to Understand 3DS Response Codes
- Visa Secure: What This Tool Does & How to Get Started
- Account Takeover Protection: Best Practices for Businesses
- How Digital Risk Protection Protocols Keep You Safe
What is Fraud Scoring?
- Fraud scoring
Fraud scoring refers to the process of quantifying the level of risk involved in a transaction. Machine learning technology examines each transaction based on dozens of different indicators, then assigns a simple numeric score representing the transaction's risk level.
[verb]/* frôd • skaw • ruhng /
Fraud scoring is also known as transaction risk scoring. That’s actually a more accurate term.
Much like a consumer credit rating, a fraud scoring system assigns different values to different elements of a transaction. Points are added or subtracted according to predetermined rules. A final score is calculated, which helps quantify the amount of risk a transaction presents.
Many payment gateways offer built-in fraud scoring, but third-party options are also available. Businesses and verticals vary, however. Make sure you look carefully at things like customization capability and cost-per-score when considering different service providers.
How Does Fraud Scoring Work?
With a fraud scoring system, numeric values are assigned to various elements of an eCommerce transaction. Clear fraud indicators, such as a missing CVV code, are weighted more heavily than something like a misspelling on the street name.
Basic rules for creating these values may be inherent in the software or script. You can tweak the specifics, though, choosing different elements and weighting them to reflect your business or industry.
Fraud decisioning is based on factors including:
- Dollar value of sale
- AVS response
- Category of merchandise
- Customer IP address
- Order history
- Shipment method
- Social media presence
- Time Zone/Time of order
- ZIP code
There can be many other elements. The main point, however, is that the final transaction score is not a random guess. It doesn’t hinge on any single piece of information. Rather, the risk is predicted based on a mathematical model.
How is Fraud Scoring Used?
If the final score is high, it means there is a good chance the order is fraudulent. The lower the score, the safer the order will be. Of course, if all cases were clearly either fraudulent or genuine…you wouldn’t really need scoring at all.
Fraud scoring is most helpful for cases that are not clearly fraudulent, but not clearly legitimate, either. Consider this example: imagine a regular customer lives in Iowa, but is staying in Singapore for business. The buyer makes a purchase using their Singapore address. Generally speaking, this would be a clear red flag for fraud.
With the benefit of fraud scoring, though, this will be weighed against other indicators. By looking at the transaction holistically, the tool determines that the purchase is valid and can go ahead.
Fraud scoring can help you eliminate pre-transaction fraud threats.
...but what about post-transaction fraud?
In contrast, you may have a transaction that seems okay at first glance. However, fraud scoring parses the transaction more closely, generating a score of 75 based on factors that aren’t immediately obvious. It exposes fraud that you might not have otherwise noticed.
Here’s a typical fraud-scoring model that demonstrates how scores might be broken out:
In the initial set-up process, you’ll usually decide how much risk you’re willing to accept. You set thresholds for when an order should be:
Real-Time Decisioning
Fraud scoring automatically checks sources much faster than could be possible with manual review. This reduces friction in the checkout process, ensuring that fewer orders get abandoned by shoppers.
Adjustable Parameters
Every merchant is different. Fraud scoring offers dynamic parameters, so you can decide how much weight should be assigned to each element. You can later adjust as necessary.
Accuracy
Fraud scoring software draws data from multiple sources. It returns unbiased, objective scores for more accurate risk mitigation that’s not colored by subjective impressions.
Dynamic Options
The best fraud scoring providers may offer an additional tier of decisioning. The system could request secondary verification (such as texting a one-time code) from a customer whose score is marginally risky.
What Are the Challenges of Fraud Scoring?
Fraud scoring is a useful tool, but it shouldn’t be the deciding factor in every case.
No technology is foolproof, and false positives are still possible. You need to calculate what each score ends up costing, and make sure you’re saving more than you’re spending.
Here are some other fraud scoring challenges to consider:
Automatically approved
Automatically rejected
Flagged for manual review
What Are the Benefits of Fraud Scoring?
The most obvious advantage of using fraud scoring is simple: it saves time and money. But there are other benefits to consider:
Bad Data
Like most computer programs, the fraud scoring results are only as good as the data they’re based on. To get accurate scores, you’ll need to make sure the system can access accurate information right from the start.
“Rule Creep”
Unfortunately, every rule you create or adjust will affect the other rules in the model. As time goes on, you may spend more and more resources tweaking, adding to, and/or recalibrating your ruleset to maintain effectiveness.
Limited Decisioning
Even the best systems often limit the number and complexity of rules you can set up. Plus, rules can’t account for context or extenuating circumstances.
The Human Factor
Algorithms typically work on “if/then” statements...but people don’t. This can lead to some false positives, and computer decisions that don’t always make sense.
Ongoing Upkeep
Your fraud scoring is not a “set it and forget it” solution. Fraudsters are constantly updating their tools. You must continually look for new patterns and tweak rules to counter them in response.
Fraud Scoring: Crucial, but Not Enough on its Own
Fraud scoring is one of the most powerful tools you can have in your arsenal. As effective as it is, however, its use is limited to preventing fraudulent orders. That won’t help with friendly fraud.
Friendly fraud is post-transactional. There’s no reliable way to stop it from happening before the sale. And, even if you could, friendly fraud usually comes from seemingly-legitimate customers. Fraud scoring would have no effect.
For effective overall fraud and chargeback prevention, you need a multi-tiered strategy that combats fraud both before and after the transaction. Chargebacks911® offers transparent, end-to-end fraud and chargeback management, going beyond prevention to revenue recovery and future growth.
Whatever you need to combat fraud, we can help. Contact us today for a free demo.
