Run controlled experiments
- Define a starting state and one changed input.
- Record UI, reports, files, database rows and registry/configuration changes.
- Capture network or device traffic with direction and timing.
- Exercise invalid inputs, cancellations, retries and recovery paths.
- Repeat the experiment and retain artifacts.
Build an evidence model
Document workflow, preconditions, state transition, output and confidence. Separate observed facts from inferred rules; one run does not prove universal behavior.
Use executable analysis for static/runtime structure, requirements recovery, black-box migration and the reverse-engineering pillar.