Report a reproducible, reviewable result

A professional report begins with the decision question, project boundary, source model and version, analysis date, currency, duration unit, baseline exclusions, and estimate or schedule maturity. It then shows the selected scenario and profile, seed, requested and completed iterations, engine version, input fingerprint, correlation state, validation findings, and convergence status.

For each selected dimension, report the unrisked baseline and distribution summary, policy percentile, raw delta, non-negative contingency or reserve, existing amount, and gap or surplus. Show confidence at the entered budget or deadline only when the boundary exists. In integrated analysis, label marginal percentiles separately from joint confidence. Include bounded charts with an authoritative table path and never omit inconvenient tails or warnings to simplify the story.

Traceability and limitations

Summarize cost items, activities, risk profiles, responses, dependencies, scenarios, evidence gaps, assumptions, and exclusions. Driver rows need their metric and direction. References establish provenance, not truth. The human recommendation appears as a distinct signed decision record, not as a simulation output.

Do not include raw samples, private prompts, service URLs, internal keys, or unrelated workspace evidence. Export must bind the exact model and version, recompute from persisted inputs when required, and show when a previous result is stale. State version 1 boundaries: no working calendars, resource loading, inflation model, non-Gaussian copulas, Bayesian calibration, automatic distribution fitting, or ERP/P6 import. Continue to assistant safeguards and proposals. ## Interpretation discipline

Keep inputs, derived results, and human decisions visibly separate. Every material number needs linked evidence or an explicit assumption flag and basis. Unknown is not zero. A percentile is conditional on this model, scenario, seed, engine version, and source version; it is neither a guarantee nor an approval. Any material model edit makes an earlier result stale.