The DHHS-funded Health Profession Opportunity Grant Program (HPOG) training program impact evaluation provides an opportunity to create new methodologies for  getting “inside the black box” of what makes employment programs successful by building on  Bloom, Hill & Riccio through a focus on individual-level variation in program experience within sites.  It introduces new analytic methods that capitalize on random variation in the exposure of individuals to different program features through three-way random assignment within a site.  The existence of exogenous variability in the parts of the treatment applied to different individual participants provides the opportunity to:  test the reliability of Bloom, Hill & Riccio’s inferential approach against an experimental benchmark; and extend the focus on individuals to model endogenous choices of service receipt within site as a further check of the reliability of the cross-site inference at the heart of Bloom, Hill & Riccio.