A method and system for testing the end-to-end performance of cloud based applications. Real workload is created for the cloud based applications using synthetic users. The load and length of demand may be adjusted to address different models allowing the measurement and analysis of user performance metrics under desired conditions. Web applications and other cloud services may be tested from multiple cloud-based concurrent geographic locations in the world. The method may include generating controlled and complex web load from multiple concurrent geographic locations. The method may also include splitting traffic among multiple geographic locations. The method may also include processing test results and generating performance metrics per location as well as comparing locations. The method may also be applicable to cloud services such as mobile apps and API endpoints.