Due to Scalarr personalized machine learning model 3x more fraud was prevented compared to the previous solution.
More than 151K fraudulent installs were detected over 3 months and rejected, and as a result, the significant part of the marketing budget was saved.
The prevalent type of fraud ZiMAD has is Click Injection (89%), which occurs mainly in incentive CPA campaigns with a CPA rate range from $3 to $10.
Smart Bots were detected in the same incentive traffic as well. Fraudsters emulated all events, even including in-app purchases. Events were emulated in abnormal numbers - about 100% of paid events, and abnormal time periods - in a very short time, not typical of real users.
Fraudsters try to attack more often during active phases of UA campaigns.