Method

Evidence-first synthesis

Preserve the path from source material to finding, decision, and action—without confusing automation with researcher judgment.

Connected MyKMHub content

When to use it

When research must connect to a complex operating context

Evidence-first synthesis is useful when findings must remain traceable across interviews, observations, policies, services, systems, or business-process steps. It is especially valuable when decisions will be reviewed, audited, challenged, or handed to another team.

Principles

What must remain true

  1. Preserve the source

    Keep enough reference information to return to the original quotation, observation, document, or record.

  2. Separate evidence from interpretation

    Record what was observed before describing the pain point, implication, or recommendation derived from it.

  3. Connect evidence to context

    Map findings to the process step, policy, role, system, journey stage, or decision they help explain.

  4. Make confidence explicit

    Distinguish direct evidence from reasonable inference. Do not present both with equal certainty.

  5. Represent absence honestly

    “None identified” means the available research does not support a finding. It does not prove that no problem exists.

  6. Keep humans accountable for synthesis

    Automation can reconcile and organize evidence. Researchers remain responsible for interpretation, validation, ethics, and decisions.

Workflow

A minimum evidence-first workflow

  1. Prepare and sanitize source material.
  2. Define the contextual structure before mapping findings.
  3. Extract evidence with stable source references.
  4. Describe supported findings separately from quotations.
  5. Assign confidence and priority using documented criteria.
  6. Mark unsupported areas explicitly.
  7. Review mappings with researchers and subject-matter experts.
  8. Use the matrix to create insights and future-state artifacts.

Quality checks

Review before using the results

  • Can every explicit finding be traced to a source?
  • Are inference and direct evidence visibly distinguished?
  • Are contradictory observations retained rather than averaged away?
  • Are evidence gaps represented without being treated as proof?
  • Have affected people or domain experts reviewed interpretations?
  • Are accessibility, privacy, and policy constraints carried into recommendations?