This candidate research release asks a practical question: when demand forecasts change, how long can a production planner safely preserve the near-term structure of an existing lot-sizing plan? In the classical Wagner–Whitin model, the paper defines a weighted commitment radius: the largest floor-respecting change in forecast demand for which every optimal revised plan retains a chosen prefix of setup decisions.

The candidate derives exact distances to a tie with any fixed competing path and a polynomial frozen-comparator lower certificate. The released software computes certified outputs in exact rational arithmetic, attaches machine-checkable witnesses, and includes a separately implemented producer-side checker. An explicit three-period example shows why the lower certificate can be conservative: its frozen radius is three halves, while the exact strong radius is two.

The evidence package includes source code, a pinned environment, full and quick replay commands, adversarial regression cases, exact enumeration benchmarks, a thirty-page manuscript, machine-readable claims and assurance records, and an illustrative UCI demand-data case. That real-data case uses reconstructed forecast vintages and stylised costs, so it is not field validation.

This remains an unrefereed candidate. It has passed producer-side replay and GitHub Actions, but it has not been independently reproduced, formally verified, conventionally peer reviewed, or tested in an operating supply chain. Novelty also remains conditional because the full text of a directly relevant 1989 paper could not be obtained in the documented bounded search.

This is an Evidence Press audio briefing, released 8 August 2026. The paper, archive, replay commands and open follow-up problems are linked on this page. The voice is AI-generated, and the audio is not additional mathematical evidence.