arrow left facebook twitter linkedin medium menu play circle
January 25, 2023 - Fang Yu

3 Tips to Enhance Fraud Detection While Minimizing Costs

Want to build an effective fraud detection strategy? It all hinges on how well you can craft multiple layers of defense. As you might expect, having the right data to support detection at each of those layers is a critical piece in making them as strong as possible.

Let’s say, for example, you’re a bank wanting to collect signals from end devices such as cell phones and laptops to monitor for fraudulent loan application activity. You’ll need to first collect data points at the root—things like user type and behavioral data. Then, to determine if a loan application is fraudulent, you’ll need bureau information and user history data.

Why buying data alone doesn’t work to enhance fraud detection

While it’s tempting to simply purchase as much data as possible so you have what it takes to cover all types of use cases and transactions, the data-buying route becomes expensive quickly. What’s more, having too much data – or irrelevant data – can overwhelm and confuse your fraud analysts.

On the other hand, a lack of data can make decisioning a challenge, leaving the door open for fraud exposure. To complicate things even further, you have multiple signal providers all trying to convince you to purchase their data. How do you know you’re buying the data you need?

The whole process of balancing good data signals with affordability gets complicated quickly. Fortunately, you can strike that balance—provided you take the right approach.

Here are three tangible tips to help you choose the right signals without overextending your fraud budget.

3 tips for building an effective and affordable fraud detection

1. Segment users first before choosing the signal

Dividing your users into segments makes it much easier to identify the types of data you’ll need to evaluate and approve transactions. If you’re a new business, registration and Know Your Business (KYB) data will be most relevant when segmenting.

To demonstrate the benefits of segmenting, let’s say for example that you have a group of users who have a high FICO score, but your performance detecting fraud for that segment isn’t very good. Third-party fraud may be the culprit, as fraudsters are likely using stolen credentials or IDs—information that can be found on the dark web.

In this case, dark web data can be useful. If you’re seeing more fraud among younger users, you may need to request more documentation for those users, which gives you the data you need to enhance your models. If you’re seeing a lot of transaction fraud, you may need to refine your merchant onboarding process.

By segmenting your users, you can apply the right signals to the right transactions for a more refined approach.

2. Buy only what you need, only when you need it

Most data vendors charge by volume. So if you set out to buy all the data up-front, you’ll most likely end up overpaying. Plus, some data vendors ask users for consent to sell certain data via SMS messages which degrades the customer experience.

Using just the signals you need reduces customer friction. Segmentation helps you decide what data you need for which transactions and when, so you don’t have to spend a lot of your budget up-front on data you may not ever need.

If you choose a fraud detection vendor with the capability to use rules and models in detecting fraud, you’ll have much better capability to make a first-line decision before you need to go to external data.

3. Re-evaluate your strategy continuously

Once you have the data you think you’ll need, run your models and evaluate the results. You may discover that your models could be enhanced with more or different data. Or, you may find there are user behavior changes that require you to revise your strategy.

For example, if you are seeing a lot of third-party fraud, you might consider purchasing data on social security numbers that are being traded on the black market. If you have an uptick in first-party fraud, you might choose to purchase bureau data. Applying fraud intelligence across your data sources will help you identify gaps and understand what other data you should purchase to achieve more accurate results.

Ideally, you’ll want to construct a centralized decision flow ion your fraud solution that visualizes external data sources by when and where they’ve been called on. That way you can easily manage and replace data sources when necessary.

Key takeaways

The main key takeaway here is that buying data for data’s sake and using it for making decisions about every type of traffic across applications isn’t an effective strategy. Instead, it’s important to be selective and only apply certain signals to decisions about specific user segments.

Determining what those segments and associated signals are will help you choose which data vendors to partner with, thereby reducing overall costs while being more efficient about your use of data for detecting fraud.

DataVisor’s fraud and risk platform provides decision workflows to help you segment and profile your users, and devise effective strategies for using data signals to prevent fraud. Our Insights Center can help you identify gaps in your fraud strategy, and determine what new data you should purchase to augment your rules and models.

Learn how Datavisor can help you leverage the right data with its automated investigation and decisioning framework, and request a demo today.

about Fang Yu
Fang spent 8 years at Microsoft Research developing big-data algorithms and systems for identifying various malicious traffic such as worms, spam, bot queries, hijacked accounts, and fraudulent financial transactions across a wide range of Microsoft products.
about Fang Yu
Fang spent 8 years at Microsoft Research developing big-data algorithms and systems for identifying various malicious traffic such as worms, spam, bot queries, hijacked accounts, and fraudulent financial transactions across a wide range of Microsoft products.