How we reduced ATO and account opening fraud for BlaBlaCar

BlaBlaCar is one of the most relevant examples of filling the gaps in a demanding market. Frédéric Mazzella, the company's founder, recognized the pressing need for increased transportation options when he found himself in a challenging situation during the busy Christmas season while planning a long-distance trip to his parents. Faced with fully booked trains and congested roads, he had the idea of connecting with drivers heading in the same direction, and sharing fuel expenses in exchange for an available seat. There was no online platform to facilitate such arrangements back then, so this is how the idea of a carpooling business emerged. 

BlaBlaCar is now the world’s leading community-based travel network enabling over 26 million unique members to share a ride across 21 markets. Through its platform, BlaBlaCar leverages technology to fill empty seats on the road with 1 million bookings per week, connecting members looking to carpool, and making travel more affordable, pleasant, and convenient.

Before using Nethone’s solutions, Blablacar faced a challenge with Account Opening (AO) fraud and Account Takeover (ATO). 

The fraud prevention mix we developed for this use case is based on custom machine-learning (ML) models, in-depth user profiling and device fingerprinting data.

We initially implemented on BlaBlaCar’s platform a Proof of Concept (POC) model customized to their specific needs for detecting ATO. The POC proved to be successful, detecting a significant number of ATO cases. Yet fraud trends evolve rapidly, so when new patterns emerged, we proactively responded to the changing landscape, by quickly adjusting the model to maintain high precision of ATO detection. 

Our ML-based fraud detection setup has proven to be highly accurate in identifying fraudsters, while minimizing false positives. This has been instrumental in addressing the ongoing challenge of scammers attempting to exploit BlaBlaCar platform. In a specific instance, our solution detected a single user/device attempting to open 536 different accounts over the course of a month, using various techniques to hide their identity. We were able to identify and block all these fraudulent attempts, protecting the experience for BlaBlaCar’s legitimate users. This particular incident highlighted our unique features and our proactive approach to mitigating fraud risks.

Throughout our ongoing collaboration, we’ve consistently provided support in improving BlaBlaCar’s fraud detection models. Our client greatly appreciates the direct interaction with the data scientist responsible for creating and fine-tuning the customized models to meet their specific requirements. 

One notable achievement is the significant reduction in scam rates over time. Through the implementation of our fraud prevention solution, our client has seen a significant decrease in scam occurrences. Furthermore, the ATO rate has been successfully mitigated, dropping to very low levels.

An important aspect of our fraud setup is the ability to detect approximately 70% of all fraudulent cases in the initial stages of the fraud prevention process. This early detection allows for swift action, preventing fraudulent activities from progressing further. It is noteworthy that this high detection rate is achieved while maintaining a minimal impact on legitimate user traffic.

Overall, the collaboration with BlaBlaCar has yielded tangible results: a substantial decrease in ATO incidents and scams, and the early detection of a majority of fraudulent cases.