As computer applications grow in complexity, ensuring that the performance of an application does not regress becomes more and more difficult. Most performance regressions are usually detected after apps are deployed to users, hurting the experience of many users. Once such performance regressions are detected, it takes a lot of effort and time to fix. This talk introduces a technique to detect potential performance regressions early in the development cycle before they even reach users. The technique is based on a static resource analysis which automatically infers asymptotic complexity of a program’s resource usage and is implemented in Facebook’s Infer static analysis tool. This talk will delve into how the underlying static analysis works, how the tool is deployed, and used in combination with dynamic profiling tools to detect performance regressions.