Model Sandbox
Understand what Kadryn model comparisons establish before rollout and what they do not guarantee.
Model Sandbox supports pre-change comparison. It does not turn simulation into a product-quality guarantee.
What does Model Sandbox establish?
Compare model alternatives using modeled cost and the evaluation evidence available in Kadryn. Simulation does not guarantee equivalent model quality.
The available comparison evidence can support a decision before rollout, but the result must remain scoped to the data, assumptions, and evaluation evidence actually available.
What does it not prove?
- It does not guarantee equivalent model or product quality.
- It does not guarantee a universal savings percentage.
- It does not establish causal ROI before or after rollout.
- It does not replace post-change measurement.
How does this relate to post-change measurement?
A model comparison is pre-change evidence. After implementation, Kadryn keeps applied changes, observed/measured outcomes, and workflow confirmation distinct.
See Savings and post-change measurement for those boundaries.
Safe interpretation
Use the Sandbox as decision evidence, not as a promise that a cheaper or differently configured model will preserve equivalent quality in production.