Section 1 gives the forecast dates and the one number that matters most: the probability of achieving the date currently in the programme. Sections 2 and 3 show the distribution behind those dates and which drivers produce the spread.
Appendix A is the risk driver register, Appendix B the duration-range basis for every activity band, and Appendix C the simulation settings — enough to reproduce the run.
The programme's planned completion of 28 January 2025 has an 11% probability of being achieved. It is not a forecast; it is the best case of a distribution whose centre lies five weeks later.
50% probability of completing on or before. +35 days on plan.
80% probability. +80 days on plan. Recommended commitment date.
90% probability. +125 days on plan.
Chance of completing by 28 January 2025 as programmed.
A single-date programme states one outcome of a network whose durations are uncertain. Where that date sits at the 11th percentile, committing to it is not an ambitious target — it is a commitment that will fail in roughly nine cases out of ten, and every downstream plan built on it inherits the same odds. The gap between P50 and P80 here is 45 days, which is the width of the uncertainty that the single date conceals entirely.
Recommendation. Adopt 18 April 2025 (P80) as the commitment date and hold 28 January as an internal target only. The 80-day difference should be declared as time risk allowance rather than absorbed silently, so that its later consumption is visible as consumption rather than appearing as fresh delay.
Working calendar: North Sea Offshore (UKCS) — 10-hour days, 5 days/week, excluding weather downtime windows (H₃⁄₃ > 3.0 m or wind > 25 kn).
Cumulative probability of completion by date across 10,000 iterations. Each iteration samples every remaining activity's duration from its assigned range and recalculates the network.
Incremental contribution of each risk driver to the P80 duration. A driver's height is the delay it adds at the 80th percentile, not its worst case.
Where the file carries duration ranges — optimistic and pessimistic fields, or a linked risk register — those are used. This file carries none, so ranges were assigned automatically by risk class: each activity is banded from its phase, duration, activity type, remaining-versus-original duration and position relative to the critical path, and given a three-point range for that band. Appendix B sets out every band and its range.
The simulation samples the network as submitted. Its output is therefore only as complete as that network, and the most important limitation of any Monte Carlo forecast is what the model does not contain.
Activities are not sampled independently. Correlation groups are applied within phases, so that a bad weather draw affects every weather-exposed activity in that period together rather than averaging out across them. Sampling independently would understate the spread substantially — on this network by roughly 30% at P80 — because it would treat a systemic driver as if it were 200 unrelated coin flips.
Every limitation above biases the result the same way. The real distribution is wider and later than shown, not narrower and earlier. Treat the P80 as a floor rather than as a central estimate.
Produced from fictional project data for illustration. No real schedule, client, contractor or asset data is used or implied.
Drivers ranked by contribution at P80. “Key activities” gives the activity-ID range over which the driver was applied.
| # | Risk driver | Category | P80 | Severity | Key activities |
|---|---|---|---|---|---|
| R-01 | North Sea weather windows Q1 2025 seasonal restrictions (H₃⁄₃ > 3.0 m) limit offshore crane operations | External | +38d | High | 1,104–1,218 |
| R-02 | Structural steel delivery — Ulsan 26d voyage + 5d port clearance, ±12d variance | Supply chain | +22d | High | 876–891 |
| R-03 | Hook-up labour availability Competing UKCS campaigns constrain specialist crews in the same window | Resource | +9d | Medium | 1,302–1,410 |
| R-04 | Commissioning system turnover Punch-list closure rate uncertainty across 14 subsystems | Execution | +7d | Medium | 1,520–1,634 |
| R-05 | Vendor representative mobilisation Offshore travel and induction lead times for specialist attendance | Supply chain | +4d | Low | 1,455–1,478 |
| Total contribution at P80 | +80d | ||||
Contributions are incremental at P80 and are not additive to the P90 figure — at P90 the weather driver grows disproportionately, which is what produces the tail in Figure 1.
Every remaining activity is banded automatically and given the three-point range for its band. Ranges are expressed as multipliers on remaining duration.
| Risk class | Applies to | Optimistic | Most likely | Pessimistic | Activities |
|---|---|---|---|---|---|
| Weather-exposed offshore | Crane-dependent scopes in the restricted window | 0.95 | 1.00 | 2.10 | 114 |
| Long-lead procurement | Fabricated items with sea freight | 0.92 | 1.00 | 1.65 | 38 |
| Hook-up construction | Piping, E&I and mechanical installation | 0.90 | 1.00 | 1.35 | 241 |
| Commissioning | System turnover and pre-commissioning | 0.88 | 1.00 | 1.45 | 147 |
| Fixed-duration inspection | Statutory and class inspections | 1.00 | 1.00 | 1.10 | 62 |
| Administrative | Approvals, holds and interfaces | 0.90 | 1.00 | 1.25 | 32 |
| Remaining activities ranged | 634 | ||||
Every band is skewed right. Construction activities can finish somewhat early but can overrun substantially, and the ranges reflect that rather than assuming a symmetric distribution around the plan. A symmetric assumption is the most common cause of a Monte Carlo forecast that looks reassuring and is wrong.
| Group | Coefficient | Rationale |
|---|---|---|
| Weather-exposed offshore | 0.80 | One weather state affects all exposed work in the same period |
| Hook-up construction | 0.45 | Shared crews and shared access constraints |
| Commissioning | 0.40 | Common systems and shared vendor attendance |
| Long-lead procurement | 0.25 | Partly common freight route, otherwise independent |
Enough detail to reproduce the run exactly. The simulation is deterministic given the same file, seed and settings.
| Setting | Value | Note |
|---|---|---|
| Iterations | 10,000 | P80 stable to ±1 day beyond ~4,000 iterations |
| Sampling | Three-point, right-skewed | Per risk class — Appendix B |
| Correlation | Applied within phases | Coefficients in Appendix B |
| Calendar | North Sea Offshore (UKCS) | 10h days, 5d weeks, weather downtime excluded |
| Seed | Derived from file hash | Same file, same result — no clock dependency |
| Computation | In-browser Web Worker | File not uploaded, nothing retained |
| Iterations | P50 | P80 | P90 |
|---|---|---|---|
| 1,000 | 2 Mar 25 | 21 Apr 25 | 28 May 25 |
| 4,000 | 4 Mar 25 | 18 Apr 25 | 1 Jun 25 |
| 10,000 | 4 Mar 25 | 18 Apr 25 | 2 Jun 25 |
P50 and P80 are stable from 4,000 iterations; P90 continues to move slightly, as tail percentiles always do. Reporting the convergence rather than only the final figure lets a reviewer judge how much confidence the percentile deserves.
Every report in the set is built from a Primavera P6 or Microsoft Project file, parsed in your browser. Nothing is uploaded and nothing is retained.