Layer 1 copy

Sift Science is a real-time learning system that adapts quickly and gives us accurate results. With Sift Science, you can discover trends quicker.

Gaspar SalvaRevenue Protection Analyst

Overview

  • Spain-based OTA operating internationally

  • Expanding into new markets

Challenge

  • Various fraud types on the platform

  • Rules-based solution couldn’t scale

Solution

  • Automation with Sift Scores

  • Investigate orders in the console

Results

  • Streamlined and data-driven decisions

  • Utilize Lists, network visualization, and Score API

Overview

A front-row seat in ticket sales

Logitravel is an online travel agency (OTA), selling various travel-related tickets and product bundles, including cruises, flights, hotels, cars, trains, ferries, and shows. The business has a robust foundation in Spain, Portugal, Italy, and Mexico, and the popularity of their dynamic packaging options has meant that iteration is key as they grow into Argentina, Russia, and other countries.

Logitravel was started by technically-minded individuals and thus much more focused on the systems than the product. However, as they sought to expand into new markets, they needed a tool to help them understand the trustworthy and legitimate users as well as efficiently identify the scammers, particularly with the recent launch of their Android and IOS applications. Currently, 95% of customers come via website.

Challenge

With a small and agile team, Logitravel was more focused on spending their time growing the business rather than building a fraud solution. However, as a business that operates over borders and across continents, Logitravel was vulnerable to fraud from several different directions. Whether organized crime fraud rings, people who capitalizing on a lack of business rules in a certain geography, friendly fraud in contested ticket sales, or phishing and account takeover, bad users were persistent and clever. In addition to the price of chargebacks, the pain felt by the Logitravel team and their legitimate customers was damaging to the brand and their resources.

After a particularly damaging barrage of fraud, the Logitravel team knew that their third party rules-based fraud solution wasn’t able to keep up with their growing business and evolving markets. Gaspar Salva, a 4-year veteran of Logitravel, was tasked with managing the fraud issue. Although Logitravel had already began a preliminary integration with Sift Science, no one had yet to own the machine learning solution and leverage its findings against fraud.

“We process thousands of transactions every day; if we didn’t have Sift Science and our other systems, we would need at least 30 people reviewing daily.”

Solution

Workflows that fuel automation

Once Gaspar took the helm on fraud management, he and his team trained the algorithm with historical data and sent labels to teach the machine what good and bad customer behavior looked like. Since then, the system has been running quietly and efficiently in the background. Using Sift Scores to automate allows for a smooth fraud management process. Suspicious orders are automatically held for review, with the Sift Science Console providing a data-rich pop-up of investigation tools for the team. The APIs allow Logitravel to easily take advantage of the machine learning findings, thus making every reviewed transaction more effective.

“Without Sift Science, our fraud review process couldn’t operate as it does now. This information that we provide to our agents allows them to make smart decisions fast.”

Results

Scoring high with fraud managers

For Logitravel, the best part about using Sift Science is its intuitive interface and actionable insights. It’s easy for Gaspar and his team to find and investigate transactions and activity. The network visualizations offer a powerful lens into how suspicious users are connected, allowing Logitravel to preemptively flag potentially linked users, saving time when those accounts surface. Using the Lists function enables Gaspar to customize a workflow to track various attributes in real-time and ultimately gives him the ability to explore fraudulent trends.

“We trust Sift Scores – if Sift Science gives someone a score of 90, it’s definitely fraud. The machine learning system learns very quickly.”