Estimate
Three-point (PERT) estimation, then a reference-class bias check against actuals from comparable finished projects.
- Takes
- An optimistic / most-likely / pessimistic three-point estimate in any unit, plus planned-vs-actual figures from comparable finished projects.
- Returns
- A PERT expected value, standard deviation, and P50/P80/P90 range — then a historical overrun ratio from your comparables and a bias-adjusted forecast.
Deterministic — no model call, no network. Pure function, not a probabilistic rater; the T1–T3 reliability-tier doctrine does not apply.
Will not
- Judge whether your three-point inputs are realistic — garbage in, garbage out.
- Source comparable projects for you — you supply the actuals.
- Guarantee the P50/P80/P90 range: it assumes a roughly symmetric distribution around the PERT mean, a standard approximation, not a certainty.
1. Three-point estimate (PERT)
Expected = (o + 4m + p) / 6. Std dev = (p − o) / 6.
- Expected
- 14.33 days
- Std dev
- 3.67 days
- P50
- 14.33 days
- P80
- 17.42 days
- P90
- 19.03 days
2. Reference-class bias check
Re-anchors the PERT expected value above (14.33 days) against actuals from comparable finished projects.
160%
125%
- Median overrun ratio
- 143%
- Bias-adjusted forecast
- 20.43 days
mean: 143% · n=2
= PERT expected value × median overrun ratio from your comparables