Anomaly detection within software build processes

The building of software packages is a compute-intensive process, especially for relatively large software. Changes that can detrimentally affect the build process are often not detected until resource consumption impact reaches critical levels. These changes become increasingly difficult to detect and isolate as time passes from the date of the change, resulting in capacity crunches and an overall slowdown of the build process. This disclosure provides techniques to automatically detect resource consumption anomalies in the build process either in real time or soon after such anomalies arise. Anomalous events, e.g., relatively large changes in resource consumption not attributable to changes in the number of users or projects, are detected based on time series data for software projects that are being built