“Our business is time-sensitive, and we needed to find a solution that quickly and effectively helped reduce fraud issues within our system. Fraud used to keep me up at night – now it doesn’t.”
Sam Shank, CEO and Founder
Mobile travel booking application
Global and growing user base
Preventing global and dynamic mobile-only fraud
Requires immediate decision at time of transaction
Leveraged customized machine-learning model tailored to their business
Improved accuracy continuously by sending relevant data and feedback
Saw accurate results within the first 6 weeks
50% reduction in chargebacks since using Sift Science
Implemented auto-banning logic within app which reduced manual reviews
Delivering last-minute hotel deals
HotelTonight was built to make traveling smarter and easier. Through the HotelTonight mobile app, customers are able to find and book last-minute deals with top-rated hotels across the globe. Whether the need is a same-day booking or a reservation as far in advance as a week, the HotelTonight app gives users convenient and immediate access to exclusive travel offers.
Prior to implementing Sift Science, HotelTonight had begun quantifying the impact of fraud and found that it was a real and growing problem for their bottom-line that needed to be fixed. They didn’t have a fraud solution in place nor know how to start tackling the problem. Meanwhile, the company was seeing an increasing number of chargebacks with each passing month. To help reduce their burgeoning fraud problem, Kenneth Sung, Sr. Fraud Analyst at HotelTonight, looked to Sift Science as the solution.
Predicting fraud before it occurs
The core foundation of HotelTonight’s business is based on allowing users to find and quickly make last-minute hotel reservations. Their booking model means that a user might book a room and immediately use that reservation to check-in. That same immediacy makes it impossible for HotelTonight to manually assess each transaction for fraud - there simply isn’t any time. For HotelTonight, a rules-based fraud solution would have been ineffective due to the spontaneous nature of transactions and the dynamic and rapid-fire nature of their fraud problem. Rules just couldn’t keep up.
Thus, the challenge for HotelTonight was to predict fraud before it happened rather than reacting to it afterwards. Once a hotel reservation is approved and booked, the funds transfer immediately, making researching or canceling the transaction nearly impossible. HotelTonight’s fraud problem was further complicated by the fact that they were experiencing multiple types of fraudulent behavior, ranging from chargebacks and credit card testers to referral code fraud.
“Sift synchronizes all the pertinent information into an easy-to-read profile so that we can quickly make educated decisions and impact legitimate guests as little as possible” - Kimberly Sutton, Trust and Safety Manager”
Continuous learning and improvement
When Kenneth took on HotelTonight’s fraud problem, he worked with a single engineer to fully integrate Sift Science into their app. In addition to analyzing typical data fields like order and transaction information, HotelTonight also sent attributes and events custom to their business, allowing Sift Science to learn and detect the fraud unique to HotelTonight. Within weeks, Kenneth had implemented a powerful machine learning fraud prevention system with Sift Science that was customized for HotelTonight and worked seamlessly with their internal fraud review team.
After just six weeks of using Sift Science, HotelTonight’s fraud analysts were seeing amazing results. During that time, Kenneth observed Sift Science’s fraud prediction accuracy improve as the HotelTonight fraud team helped teach the machine learning system about their fraudulent activity. Over time, HotelTonight’s Sift Scores became so accurate that they were able to effortlessly determine if a user was bad or good, making it easy for their team to take quick and definitive action.
Confident and efficient automation
With every passing week, the HotelTonight team grew more efficient as Sift Science honed in on their unique fraudsters. Sift Science became so accurate that Kenneth felt confident enough to automatically ban and decline users without any manual review by integrating Sift Science’s APIs into their app. This allowed their team to save precious time, as they now confidently blocked 6.5x more fraudulent orders automatically.
Sift Science is now fully incorporated into HotelTonight’s fraud management practice and is the primary solution for managing their fraud review process. HotelTonight uses Sift Science’s Console and Custom Lists feature to prioritize and expedite manual reviews based on customizable criteria. Since implementing Sift Science, HotelTonight has seen their chargebacks drop by over 50% while also drastically reducing the time spent on manual reviews.
“We definitely saved a lot of time with Sift Science. Our fraud management resources are so much more efficient, saving us time and energy to focus on growing our product.”