Build unrisked baselines and supportable distributions
The cost baseline may be one aggregate estimate or itemized estimate elements. The schedule baseline may be one aggregate duration or a network of activities. In every case, confirm explicitly that the baseline excludes contingency, risk allowances, management reserve, andāwhere applicableāschedule reserve. Keep escalation and other known scope provisions visible rather than hiding them in an uncertainty range.
Base uncertainty represents estimating or duration variation that remains even if no named risk event occurs. For each aggregate, item, or activity, record the point value, basis, evidence links, assumption status, and optional distribution. Avoid applying a broad percentage range to detailed items while also modeling the same uncertainty at aggregate level; that double counts exposure.
Distribution choices
- Fixed records one supported value and contributes no variance.
- Uniform gives equal density between supported minimum and maximum bounds.
- Triangular adds a most-likely value; use it only when all three points have a coherent basis.
- Beta-PERT smooths a bounded three-point judgment and requires a declared shape parameter.
- Bounded normal needs mean, standard deviation, and defensible truncation bounds; an unbounded normal may imply impossible negative values.
- Lognormal is positive and right-skewed; its arithmetic mean and standard deviation must be interpreted correctly.
Unknown is null, not zero and not a fixed distribution. Do not fit a distribution by intuition merely because its shape looks professional. Record estimate class, sample provenance, elicitation method, and why the chosen family represents the mechanism. Invalid ordering or non-finite parameters block simulation. Next, model risk profiles and responses. ## 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.