There are numerous applications of human performance measurement during training. These include supporting the post-training debriefing; estimating rates of skill decay for use in scheduling refresher training; making data science-based simulator acquisition decisions, and; quantifying training Return on Investment. While much has been written about human performance measurement in the abstract (Beaubien et al., 2017; Dwyer et al., 2001; Rosen et al, 2008), researchers have yet to formally specify the critical High Level Architecture (HLA) or Distributed Interactive Simulation (DIS) data elements that are required to support these specific applications in practice. For example, if the goal is to facilitate a single post-training debriefing, one only needs to record the learners’ performance score. If the goal is to compute skill decay rates, one must also collect each learner’s unique identifier, each training trial’s unique identifier, and the unique timestamp. Finally, if the goal is to make data-science based simulator acquisition or ROI decisions, one must also collect metadata about each simulation platform’s fidelity cues and costs. The purpose of this paper is to specify the critical data elements that are required to support these three applications. While some of the data elements are common across all three applications; others are application-specific. Similarly, some are transmitted across the HLA (or DIS) data bus, while others are accessed via other means. Finally, some are specific to Human Performance Measurement Language (HPML), while others generalize to the Experience API (xAPI). The paper will conclude with best practices and lessons learned for researchers, developers, and engineers about how to systematically collect, annotate, aggregate, archive, model, and visualize human performance data during training to support these three applications.