Fraud controls for financial transactions are needed and required by law enforcement agencies to flag suspicious criminal activity. These controls, however, require deeper analysis of the effectiveness and the negative impact for the legal customers. Owing to the intrinsically private nature of financial transactions, this analysis is often performed after several months of actively using fraud controls. In this paper, we present an analysis of different fraud prevention controls on a mobile money service based on thresholds using a simulator called PaySim. PaySim uses aggregated data from a sample dataset to generate a synthetic dataset that resembles the normal operation of transactions and injects malicious behaviour. With technology frameworks such as agent-based simulation techniques, and the application of mathematical statistics, we show in this paper that the simulated data can be as prudent as the original dataset for setting optimal controls for fraud detection.