Why we version every scenario drop

Scenario drift is the silent killer of fair evaluations. When two cohorts see subtly different motion seeds, your metrics tell a story nobody can defend. We treat scenario packages like software releases: semantic versions, checksum manifests, and signed instructor notes.

Versioning also shortens classroom turnover. Instructors swap packages in seconds, confident that downstream analytics will recognize the build. That predictability matters when you are stacking multiple units through the same week.

We still see teams resist the overhead. The compromise we recommend is lightweight tags for experimental vignettes, promoted only after a dry run with analytics attached. Finally, keep a human-readable change log—even bullet points beat a binary-only drop.