- Indonesia's leading travel platform for flight and hotel bookings
- Growing rapidly in Southeast Asia
- Payment fraud and ATO were damaging user trust and brand reputation
- Rules couldn't keep up with an increase in customers and evolving fraud tactics
- Two separate custom machine learning models for payment abuse and ATO, respectively
- Faster, more efficient, proactive fraud detection
- Now accepting 2x the orders
- Reduction of traffic to 3D Secure by 3x
A smart booking platform for savvy travelers
Jakarta-based Traveloka is Indonesia’s number one platform for booking flights and getting great deals on hotels. With an ever-growing number of visitors to the site, this company has grown to offices in Thailand, Malaysia, Singapore, Vietnam, and the Philippines. Traveloka’s business is booming in the Southeast Asian market and – following on the heels of legitimate customers – fraudsters are creeping into the fold.
Account takeover and too much friction degraded trust
As a Sift customer, Traveloka’s volume of fraud is miniscule, and the Traveloka team is committed to keeping that fraud rate low. Traveloka sees two main types of abuse: 1) payment fraud coming from stolen credit cards and 2) account takeover (ATO) stemming from stolen credentials and social engineering schemes. Both these problems lead to financial loss but, more importantly, damaged user trust and brand reputation. In order to combat these problems, Traveloka dedicated an internal team to fraud and risk, developing a series of elaborate fraud rules that attempted to provide an automated first screening of all orders.
However, as the range of customers on the site changed, Traveloka’s rules-based system couldn’t keep up. When it came to payment fraud, they experienced many false positives that were blocking good customers and their orders, leading to poor customer experience. On the ATO side, static rules were missing a lot of cases, weren’t able to adapt quickly enough to emerging trends, and resulted in a lot of false positives, blocking legitimate users from accessing the site. Traveloka wanted to maintain their low fraud rate for both payments and ATO while also reducing friction for legitimate customers falsely caught in the rules filters. This led Traveloka to search for a flexible and adaptive solution.
“When our anti-fraud rules rejected too many legitimate transactions, we engaged with Sift to reduce that number of false positives and stay ahead of real fraudsters.”
Custom machine learning models and behavioral data for multiple types of abuse
In keeping with Traveloka’s focus on smart solutions and innovation, they began investigating machine-learning based solutions to replace their rules-based system. Big data was already an integral part of Traveloka’s customer service, marketing, and fraud operations. And now the product team – headed by Wayan Perdana – was tasked with finding an adaptive solution that reduced false positives, identified more ATO incidents, and could increase conversions. He turned to Sift because of its sophisticated machine learning platform that scales with growth, adapts to new fraud patterns, and accurately separates good users from bad.
Traveloka integrated with Sift to detect both types of fraud. Traveloka has two separate, custom machine learning models that leverage behavioral data – one for payment abuse and the second for ATO – to identify suspicious cases. Traveloka saw accurate results within weeks, giving them the ability to reduce friction for legitimate, paying customers and preemptively identify more ATO attempts at the point of login. Additionally, the Sift Console provides Traveloka’s team an easy-to-use interface for investigating both payment fraud and compromised accounts. They can work faster and be more productive thanks to the holistic picture it paints of the user’s activity.
“We believe in an adaptive machine-learning approach to fraud management. Sift's web interface and API were quite simple and straightforward.”
Faster checkout for happy customers; fewer ATO incidents
With Sift, Traveloka was able to accept twice the amount of orders that were previously blocked by their rules system. They also crafted a better customer experience that reduced traffic to 3D Secure by 3x. On top of that, they have seen fewer ATO cases overall thanks to Sift’s detection abilities.