Know what the result can support
Schedule Forensics produces computations and clause findings for a reviewer. It does not approve schedules, establish delay responsibility, determine entitlement or supply an expert opinion.
Numerical agreement is measured, not guaranteed
The completed minute-mode snapshot matches 92.85% of date/day comparisons and 91.91% of exact timestamp/minute-float comparisons. Both use 584,687 fields. Read the measurement and its population limits.
The public demonstration uses the CLI's default day-based calculations. Those recorded outputs are not a demonstration of minute-mode parity. Calendar boundaries, progress options, constraints and incomplete inputs can change computed dates and floats. Reconcile calculations with the source scheduling application before relying on them.
Actual Dates progress handling has implementation and format tests, but no new licensed P6 recalculation establishing every case is claimed. Conflicting stored export fields cannot all be reproduced simultaneously.
Coverage depends on supplied evidence
A passing clause does not mean the schedule as a whole conforms. Many requirements depend on contract terms, project records or a person's judgment. The report distinguishes passed, failed, observed, referred and unevaluated clauses and shows the coverage denominator.
The tool does not read the complete project record from a schedule file. Narrative content, correspondence, drawings, meeting records and the truth of supplied contract terms require separate review. A term that is present but incorrect can produce an incorrect finding. Missing and explicitly empty inputs are deliberately different.
Implemented rule coverage is documented in the rule reference. A referenced standard is not a promise that every clause is automated. Cost/resource data availability and reader support do not establish coverage of every cost/resource requirement.
Analysis choices remain material
The engine implements nine AACE MIPs and prospective time impact analysis. That does not establish each method against every project or an independently adopted expert analysis. The analyst must choose an appropriate method, supply the event evidence and read the method-specific qualifications.
Concurrency theory, criticality, float ownership, pacing, prior awards and period selection can change the answer. Do not combine results expressed on different day-count bases or treat a per-event computation as an allocation of responsibility. The sample reports expose their particular assumptions.
Formats have unequal evidence
Readers support XER, Primavera XML, Microsoft Project XML and USACE SDEF; binary .mpp is unsupported. The retained external comparison corpus is XER. Own tests of another reader are not the same evidence as independent application exports.
Source scheduling options and the calculation mode belong with the result. Default or missing settings must be reviewed, especially for multi-calendar work. An exported or modified schedule is not automatically an acceptable contractual submission.
Deployment requires acceptance checks
The current browser service is local to one computer. Organization-wide hosting, user authentication, access control, support and release acceptance must be established for the agreed deployment. No cloud authorization, security certification, insurance coverage or customer endorsement is claimed by this site.
Cross-platform and cross-version reproducibility must be demonstrated for the package and environment being delivered. A successful replay on one machine does not establish either. Accessibility checks and semantic reports do not replace an end-to-end review with the assistive technology your team uses.
A verification record has boundaries
Reproduction does not establish the authenticity of the submitted file, the truth of the assumptions, the correctness of the engine, or when a record was created. There is no external timestamp authority in the shipped workflow. Any legal certification text is a draft for a qualified person, not a certification by the software.
Inspect the verification example or discuss your acceptance requirements. The historical limitation reference retains the earlier investigations and their original scope.
Source: web/pages/limitations.md. Source commit date: 2026-09-10.