Fingerprint's new fraud tool learns from each business's own cases

Fingerprint, a company that helps online businesses recognize devices and spot fraud, has launched a tool called Fraud Score. It learns from each customer's own confirmed fraud cases. The company says that in one head-to-head test it caught nearly 50% more fraud than that customer's own in-house model.

What changed

Fingerprint already identifies a device across visits and gathers clues about web and mobile activity. Fraud Score adds a prediction on top. The business sends in cases it has confirmed as fraud, through a dashboard, an API or S3. Fingerprint trains and maintains the model, then retrains it as new outcomes arrive. Predictions come back through Fingerprint's API responses.

The key claim is about balance. Fingerprint says the tool found nearly 50% more fraud while blocking the same level of legitimate users. That matters because any system can catch more fraud by turning away more traffic, which also turns away real customers.

Who it affects

The tool is aimed at online businesses and their fraud teams. Fingerprint lists several targets it can be trained on:

• Payment fraud
• Account takeover
• Multi-accounting and fake account creation
• Promotion abuse
• Platform abuse and content abuse

Each one looks different. A tool for promotion abuse might look for repeated attempts to claim an offer through many accounts. An account takeover tool tries to stop unauthorized access without bothering the real owner. Flink, a customer, said blocking visitor IDs alone was not enough against more sophisticated abuse.

What to watch

The tool is only as good as the records it learns from. If a business's fraud labels are incomplete or inconsistent, the predictions can reflect that. Some outcomes, such as a payment later disputed, may only become clear after the fact, so tracking confirmed cases and checking the model's performance over time stays part of the job.

The 50% figure also comes from one customer, as reported by the company. The gain measures extra fraud detected. What that is worth in dollars depends on the transactions involved and how a business uses each prediction.

What happens next

Fraud Score is available now through Fingerprint's Fraud Score page. Businesses can use its predictions directly or feed them into their existing decision systems to approve, verify or review an interaction. The announcement does not say how the results hold up across other customers, and it does not give pricing.