A method and device for suppressing threshold alerts in a telecommunication fraud control system is disclosed. Threshold alerts are generated when the count of a certain category of call exceeds a certain threshold. These counts are maintained in relation to particular accounts. A fraud analyst determines whether or not a particular account will have alert suppression enabled, based on the type of account and its history. Once alert suppression is enabled, the count is multiplied by a coefficient before determining whether to issue a threshold alert. If the multiplied count still exceeds the threshold, an alert is generated. If not, no alert is generated.