Software project telemetry is a new approach to software project management in which sensors are attached to development environment tools to unobtrusively monitor the process and products of development. This sensor data is abstracted into high-level perspectives on development trends called Telemetry Reports, which provide project members with insights useful for local, in-process decision making. This paper presents the essential characteristics of software project telemetry, contrasts it to other approaches such as predictive models based upon historical software project data, describes a reference framework implementation of software project telemetry called Hackystat, and presents our lessons learned so far.