Due to its massive popularity, Avito was seeing a huge number of new account registrations—the problem was that only 10% of these new registrations seemed to be from real new users. Because the other 90% of new registrations (called one-time accounts) were mostly being used for fraud and spam, Avito needed a way to detect and block these suspicious accounts. On top of that, Avito was looking for a way to augment their Ad moderation machine learning process with account registration risk information.
After looking at what phone number intelligence local mobile operators could provide, Avito quickly saw that TeleSign’s Score API was exactly what they needed to cut down on fraudulent accounts and secure their platform. TeleSign was able to provide:
- Analysis of users’ phone numbers and associated behavior attributes from digital services around the world, not just the communication behavior data that local mobile operators provide.
- Real-time cross-customer risk evaluation data from many domains beyond just Classifieds, which incorporates irregular online registration activity to help flag risky phone numbers when they were used on Avito’s platform.
- Customized phone number intelligence algorithms to add to Avito’s machine learning models
With Score, automatically classifying accounts as potentially harmful or safe, has quickly become one of the most important features in Avito’s platform machine-learning model, both in user registration and in their content moderation process when determining whether individual ads are harmful. “TeleSign’s Score API helped us to improve the moderation process dramatically and reduce fraud by providing the highest accuracy of important signals needed to ensure account quality,” said Andrey Rybintsev, Trust & Safety Director, Avito.
Since implementing Score, Avito has been able to reduce fraudulent activity on their site by more than 40% as measured by the number of fraud victims per day.