Reviewing a calculation requires knowing which inputs, rules and engine version produced the result. Reproducing it later also requires preserving the execution conditions that affect the calculation.

Records and integrity checks

A calculation record can include versioned inputs, the selected contract profile, rounding rules, engine version and output. A recorded hash can detect a change in serialised inputs or results. Identifying the cause requires comparing the versioned records; the hash alone does not identify which input changed.

Repeatability tests

A repeatability test runs a fixed dataset under defined conditions and compares the result with an expected output. Expected-output changes require a reviewed explanation. Cross-platform reproducibility needs checks on the supported platforms as well.

Fix accumulation order where floating-point rounding can affect results.
Define rounding points and serialisation rules.
Record the engine version and relevant runtime dependencies.
Distinguish equal numerical results from byte-identical PDF or spreadsheet files, which may contain changing metadata.

Product context

Road Designer’s bundled sample provides an input dataset and example deliverables for evaluation. The sample alone is not evidence of byte-identical output across machines. Maître d’Ouvrage’s versioned payment-calculation workflow is a development objective, not a currently deployed capability.

Reproducibility and correctness

A reproducible result can still be wrong. Engineering correctness requires validation against reference cases, applicable rules and the project’s assumptions. SmartBar’s provisional checks therefore remain subject to technical validation and structural-engineer review.