AbstractEvidence-based SE is an extension of model-based SE that emphasizes not only using SysML or other system models as a basis of program decisions, but also the use of other models to produce evidence that the system models describe a feasible system. Such evidence is generally desired, but often is not produced because it is not identified as a project deliverable in a Data Item Description (DID). Going forward with such unproven solutions frequently leads to large program overruns.Based on experience in developing and using such a DID on a very large project, we summarize the content and form of such a DID, and a rationale for its use. Its basic content is evidence that if a system were produced with the specified Architecture, it would:.•Satisfy the specified Operational Concept and Requirements;•Be developable within the specified Budget and Schedule;•Provide a superior return on investment over alternatives in terms of mission effectiveness; and•Provide satisfactory outcomes for the system's success-critical stakeholders.One key factor of the DID is that the content of the evidence be risk-balanced between having too little evidence (often the case today) and having too much (analysis paralysis). Thus, it is not a one-size-fits-all DID, but one which has ways to be tailored to a project's most critical evidence needs.