Tropic Ridge Wind Farm - technical risk report (synthetic)
Synthetic, illustrative data on a fictional project. This is not a client deliverable and is not for funding decisions.
Run 2026-11-03 · engine commit db2f509 · seed 42 · 100,000 iterations · Back to the dashboard

Tropic Ridge Wind Farm - Integrated Cost and Schedule Risk Analysis

Ridgeline: cost and schedule risk for capital projects, from bid to handover. A CBL Analytics product.

Tropic Ridge Wind Farm is a synthetic project. Its register, schedule and every figure in this report were generated to demonstrate the method; they describe no real project.

Project: Tropic Ridge Wind Farm
Run date: 2026-11-03
Iterations: 100,000
Seed: 42
Day rate: $35,000/wd
Engine SHA: db2f509642eba9ee9b2f7321ceed529aa8c2bd9c

Cross-validation

Check	Result	Detail
Cost marginal	PASS	gap $4,237 (0.052%)
Schedule marginal	PASS	gap 0.000 days
Top-3 cost drivers	PASS	cost-only WT-008, WT-009, WT-012; integrated WT-008, WT-009, WT-012
Top-3 schedule drivers	PASS	standalone WT-021, WT-008, WT-022; integrated WT-021, WT-008, WT-022
Overall	PASS	basis: the Stage 3 replay

Executive summary

Recommended project contingency: $14.34 M (P80, integrated cost plus schedule).

Integrated event-risk contingency, cost plus float-aware delay cost on the Stage 3 basis: P50 $10.21M, P80 $14.34M (recommended), P90 $16.57M. Schedule slip in days follows below.

Headline contingency (P80, Stage 3)

Recommended contingency: $14.34M (P80, integrated cost plus schedule)

Percentile	Integrated contingency ($M)
P50	$10.21M
P80	$14.34M
P90	$16.57M
P95	$18.50M
Mean	$10.53M

Stage 3 CPM replay: one distribution, percentiles of one modelled outcome. The P50-P95 spread is project uncertainty, not a spread between Mode assumptions.

Schedule risk analysis on the Stage 3 CPM replay indicates a mean schedule slip of 70.1 days, equivalent to $8.07M at $35,000 per working day. The Stage 2.5 Mode-scenario comparators span 79.0 to 95.1 days mean; they are shown for continuity with earlier reporting and are not the headline.

The headline is one Stage 3 distribution: each risk's delay is injected at its declared gating activity and propagated through the programme's own logic, and the P50 to P95 spread is project uncertainty rather than a spread between Mode assumptions. The table below is the Stage 2.5 comparator range - the DP Review baseline and the All Series upper bound that earlier reports published as the headline - with the Stage 3 figure placed between them for reconciliation. Cost analysis is independent of Mode classification.

Scenario	Mean schedule slip (days)	Mean cost ($M)	P80 contingency ($M)	Notes
DP Review baseline (Mode as registered)	79.02	8.08	14.86	Conservative starting point
Stage 3 (CPM replay, one distribution)	70.11	8.07	14.34
All Series (all-Series upper bound)	95.14	8.08	15.58	Methodological extreme

Distribution

Total cost - base, event-risk cost and delay cost

The total-cost S-curve puts the base estimate beneath the contingency, in layers on one dollar axis: the base with its systemic (Layer 1) uncertainty, then plus event-risk cost, then plus delay cost. Contingency at any confidence level is the total at that level less the declared base. Read that way it also carries the base's own uncertainty, so it is larger than the event-risk contingency the headline recommends, which this chart does not replace: it shows where that contingency sits on the project's whole cost.

Confidence level	Total cost ($M)	Declared base ($M)	Contingency, P-value less base ($M)
P50	$137.95M	$120.00M	$17.95M
P80	$146.52M	$120.00M	$26.52M
P90	$151.27M	$120.00M	$31.27M

Base: the register's Layer 1 estimate of $120.00M, sampled with its systemic factors on a random stream of its own. The register's shared drivers (table below) link part of it to the schedule: in each iteration a driver's draw moves both the time-dependent cost on its grid rows and the durations of its activities, and this total takes its delay cost from that drivers-on schedule pass. The event-risk contingency above is drawn on the event-only pass and does not include the drivers. The base is illustrative: this is a synthetic project.

Shared drivers: uncertainty that moves cost and schedule together (AACE 57R-09 risk drivers). A productivity driver draws its own range and applies it to its grid rows' cost and its activities' durations; a quantity driver takes its Layer 1 quantity factor's draw and applies it to its activities' durations only. Owner-view P80 finish: working day 420, against 405 on the event-only pass. The last two columns are leave-one-out at P80 on the owner view: each figure is the owner-view P80 with the driver, less the P80 with that driver removed (its cost and its duration change) and every other draw held fixed. They do not sum, as the risk tornado's do not.

Driver	Kind	WBS	Range (P10 / Mode / P90)	Acts on	Mean duration change	Adds to P80 total ($M)	Adds to P80 finish (wd)
D1 Civil productivity and ground	Productivity	10, 20	-5% / +0% / +20%	own range; cost on BOP civils (Labour, Plant)	+6.3%	$0.61M	+6.3
D2 Erection productivity	Productivity	30	-5% / +0% / +15%	own range; cost on Turbine transport & erection (Labour, Plant)	+4.3%	$0.25M	+1.5
D3 Electrical installation productivity	Productivity	40	-5% / +0% / +15%	own range; cost on Electrical BOP (Labour, Subcontract)	+4.2%	$0.27M	+1.2
D4 Grid and substation scope	Quantity	50	from Q_Grid x 1	quantity factor Q_Grid; no added cost	+8.5%	$0.19M	+5.5

Joint confidence - cost and finish together

Each dot is one iteration: its cost on the vertical axis and its finish on the horizontal. The dashed lines are the budget and the P80 finish; the dark dots meet both. The cost axis includes delay cost, because the budget has to fund it; the figure counts iterations that meet both lines and adds nothing across them. The orange line is where 70% of iterations meet both (the NASA reference norm); no Australian guidance requires a joint confidence level, so it is shown beside the headline, not as one.

Joint confidence level 33%: the share of iterations that come in at or under a budget of $134.34M (the $120.00M base plus the P80 contingency) AND finish by the P80 finish, working day 420. On their own the budget is met in 35% of iterations and the finish in 80%. Rank correlation of cost with finish 0.44; meets both 33%, budget only 2%, date only 47%, neither 18%.

Cost and schedule are sampled jointly: a risk that fires raises cost and delays the finish in the same iteration, and each shared driver moves its cost and its activities' durations from one draw. Both axes are the owner view, drivers on. The base estimate's price and rate factors are still drawn independent of the schedule, so they spread the dots up and down without a slope. The budget is the declared base plus the event-risk P80 contingency.

The integrated contingency is the headline distribution: in each iteration, the event-risk cost plus the finish delay that survives float, priced at the day rate. The S-curve gives the probability that the contingency needed is at or below any dollar value, with P50, P80, P90 and P95 marked; the recommended contingency is read off it at P80. The histogram below shows the same distribution's shape.

The build-up draws the same total in layers on one dollar axis: event-risk cost alone, then with delay cost added. The gap between the two curves at any probability is what schedule delay adds at that confidence level. Because the two arms are sampled jointly, the gap at P80 is not the delay cost's own P80; the P80 build-up further down shows the difference as the joint-sampling adjustment.

Confidence curve - integrated contingency by P-level

P80 is the recommended industry-standard contingency level for engineering and construction projects (USACE Cost Engineering DX guidance, AACE RP 41R-08 / 65R-11). Higher confidence increases conservatism; lower reduces it.

Confidence level	Integrated contingency ($M)	Schedule slip (days)	Above naive sum ($M)
P50	$10.21M	67.1	$0.09M
Mean	$10.53M	70.1	$0.00M
P70	$12.77M	87.6	$-0.01M
P80 (recommended)	$14.34M	99.9	$-0.12M
P90	$16.57M	118.0	$-0.33M
P95	$18.50M	133.7	$-0.50M

Above naive sum column shows how much the integrated (joint cost+schedule) contingency exceeds the simple addition of cost-only and schedule-only contingencies at the same P-level. Positive values indicate cost and schedule are correlated through shared tail risks. Schedule slip column is dollarised at the day rate to populate the integrated contingency; raw days shown for engineering reference.

Decomposition of integrated P80 contingency: cost-P80 plus schedule-P80 dollarised at the day rate ($35,000/wd), less the diversification adjustment from joint sampling, equals the integrated P80 reported in the headline.

Event-risk cost distribution

The event-risk cost distribution shows the per-iteration total event-risk cost from the integrated Monte Carlo simulation. The S-curve gives the cumulative probability of total event-risk cost at any dollar value; the density histogram shows the right-skewed shape characteristic of event-risk distributions where most iterations cluster near the mean and a long tail captures rare high-cost scenarios. Reference percentiles P50, P80, P90, and P95 are marked. Read this view alongside the integrated P-level table above to understand both the marginal cost distribution shape and the joint integrated headline figure.

Schedule slip distribution

These charts show the full probability distribution of schedule slip, not just point estimates. The S-curve shows cumulative probability; the histogram shows the right-skewed shape. Cumulative probability runs from 0 (no slip) to 100 (all simulated outcomes are at or below this level). Reference percentiles (P50, P80, P90, P95) are marked as vertical lines.

The right-skew is a direct consequence of low-probability, high-impact risks. Mean is pulled to the right of median by the tail; the distance between P80 and P95 widens as Mode shifts from Parallel toward Series. Selecting a contingency at the mean leaves roughly half the simulated outcomes uncovered; selecting at P80 covers four in five. The choice is risk appetite, not arithmetic.

Cost drivers

Integrated drivers - leave-one-out at P80

Each bar is the integrated P80 saved if that one risk were removed, with every other risk's draws held fixed (100,000 iterations, the run's own seed and joint sampling). It is a leave-one-out figure, not a contribution: the bars do not add up to the total, because removing two risks together is not the same as removing each alone. Each bar is split into the risk's cost and the delay cost its removal saves, which is zero for a risk with no time impact and can be small for one whose delay float absorbs. A chain child keeps the fire state the run gave it when its parent is removed.

The cost tornado ranks risks by their P80 contribution to total cost contingency. Each bar's length is the dollar value at the P80 percentile of that risk's per-iteration cost; longer bars are larger drivers. The label shows the dollar value and the share of total P80 cost contingency.

Risk	Title	P80 contribution ($M)	% of total	Basis of estimate
WT-008	WTG delivery delay (long-lead)	1.59	8.8%	Wind Farm Beta capex basis
WT-009	Main crane mobilisation delay	1.45	8.0%	Wind Farm Beta capex basis
WT-012	Erection weather standby	1.10	6.1%	Wind Farm Beta capex basis
WT-021	Grid connection redesign	0.89	4.9%	Wind Farm Beta capex basis
WT-031	SCADA integration defects	0.86	4.7%	Wind Farm Beta capex basis

A small number of risks dominate cost contingency. Treatment effort concentrated on the top three drivers delivers the largest contingency reduction; the long tail contributes incrementally. This pattern - top-3 dominance - is consistent with industry experience and validates focusing risk treatment plans on the ranked top slice.

Contribution by RBS Level 1 category

Cost contingency concentrates by Risk Breakdown Structure (RBS) category as well as by individual risk. The table below shows P80 contribution aggregated by RBS Level 1, useful for assigning treatment-plan ownership and for stakeholder discussions about which discipline carries the largest contingency.

RBS Level 1 category	Risk count	P80 contribution ($M)	% of total
Turbine Supply & Erection	6	4.84	32.6%
Civil & Earthworks	7	3.16	21.3%
Grid Connection	5	2.30	15.5%
Balance of Plant	5	1.73	11.7%
Commissioning & Controls	2	1.40	9.5%
Electrical Reticulation	4	1.38	9.3%

Schedule drivers

The schedule tornado ranks risks by probability-weighted mean contribution in days, not by P80. For low-probability discrete risks the P80 quantile is mathematically zero; reporting at P80 would obscure the contribution. Probability-weighted mean (firing probability times mean impact when fired) is the appropriate ranking metric and matches the EMV first-principles calculation.

Risk	Title	Probability	Mean impact when fired (days)	Probability x mean impact (days)	Mode	% of additive draws (comparator)	Basis of estimate
WT-021	Grid connection redesign	30.0%	58.48	17.544	n/a	22.2%	Synthetic
WT-008	WTG delivery delay (long-lead)	40.0%	34.74	13.897	n/a	17.6%	Synthetic
WT-022	Network operator approval delay	40.0%	28.36	11.344	n/a	14.3%	Synthetic
WT-028	Power transformer delivery delay	25.0%	33.10	8.274	n/a	10.5%	Synthetic
WT-001	Turbine foundation rework	45.0%	12.91	5.807	n/a	7.3%	Synthetic
WT-031	SCADA integration defects	50.0%	9.88	4.939	n/a	6.2%	Synthetic
WT-024	Grid compliance test failure	25.0%	17.53	4.383	n/a	5.5%	Synthetic
WT-012	Erection weather standby	60.0%	6.51	3.904	n/a	4.9%	Synthetic

Schedule contribution concentrates in a different set of risks than cost contribution. Cost-only analysis cannot infer schedule drivers; integrated cost-and-schedule MC is required to surface the schedule-dominant risks.

An explanatory comparison of the traditional vs probability-weighted risk matrix is included in the appendix.

Schedule contribution by RBS Level 1 category

Schedule slip concentrates by Risk Breakdown Structure (RBS) category as well as by individual risk. The table below shows probability-weighted mean schedule contribution (working days) aggregated by RBS Level 1, useful for assigning treatment-plan ownership and for stakeholder discussions about which discipline carries the largest schedule exposure.

RBS Level 1 category	Risk count	Probability x mean impact (days)	% of total
Grid Connection	3	33.27	42.0%
Turbine Supply & Erection	3	20.33	25.7%
Balance of Plant	3	9.44	11.9%
Civil & Earthworks	3	9.39	11.9%
Commissioning & Controls	2	6.70	8.5%
Electrical Reticulation	1	0.00	0.0%

The 12-scenario counterfactual table quantifies how the headline mean schedule slip would shift under each plausible single-risk Mode reclassification. Each row shows mean slip if that risk were reclassified to Series while all other classifications remain at DP Review. Rows with delta > 1 day from baseline are the high-leverage Mode decisions; resolving these classifications narrows the headline range substantially.

PM review should prioritise the rows highlighted above. The highest-leverage classifications are the ones where reclassifying to Series shifts the mean by more than one day; these are the decisions that move the contingency. Risks whose Mode shift produces sub-day movement contribute to range width but not to the central estimate.

Two risk matrices compared side by side. Left: traditional probability vs mean impact (days when fired), the matrix engineers see in standard QRA. Mean impact when fired is the PERT mean of the register Optimistic/Most Likely/Pessimistic days and is independent of Mode classification. Right: probability vs probability-weighted impact (probability times Mode-multiplied mean impact), the matrix that ranks risks the way Monte Carlo does. Markers on the right panel reflect the Stage 3 Mode scenario; the schedule-drivers table immediately above uses the same scenario for consistency, so right-panel x-coordinates match the table's Probability x mean impact column. Each marker is a risk; the top contributors are labelled with their risk IDs.

Compare the position of low-probability, high-impact risks across the two panels. On the traditional matrix they sit at low probability and high mean impact - easy to deprioritise in standard prioritisation. On the probability-weighted matrix they rank higher because the integrated mean impact captures the tail. This differential ranking is the visual case for MC over standard matrices.

Reconciliation

How does this MC result compare to first-principles and gut-feel methods that the reader is likely familiar with? The table below lists six approaches that all aim at the same question - how much contingency? - and shows what each one produces against the same risk register.

Approach	Schedule contingency (days)	Cost contingency ($M)	Notes
Engineering judgement (10% of programme)	30.50	NaN	Common rule of thumb
Engineering judgement (20% of programme)	61.00	NaN	Conservative gut-feel
Sum of pessimistic impacts (worst case, no probability)	558.00	45.84	Adds all worst-case impacts; ignores probability
EMV (sum of probability x ML impact)	84.20	6.62	First-principles probability-weighted estimate (deterministic)
MC mean, Stage 3 headline (recommended central estimate)	70.11	8.07	Reconciles with EMV; adds tail visibility
MC P80, Stage 3 headline	99.92	10.97	Useful for contingency floor
MC P95, Stage 3 headline	133.71	14.32	For risk-aware contingency

MC mean reconciles with EMV (sum of probability times mean impact) to within MC noise - this is a validation check that the simulation is implementing the underlying first-principles calculation correctly. Beyond that reconciliation, MC adds tail visibility (P80, P95) that deterministic methods cannot provide. Engineering judgement bounds (10-20% of programme) bracket the MC result, confirming the result sits within the range a seasoned engineer would recognise.

How is the integrated cost-plus-schedule P80 constructed? The waterfall below traces the components: cost-only P80, then schedule slip dollarised, then the diversification adjustment (or anti-diversification) under joint sampling. The two scenarios shown are the headline and the All Series upper bound.

Under the headline distribution, integrated P80 may exceed the naive sum of cost-only P80 and schedule-only P80. This is a counterintuitive but real result of joint sampling: when a low-probability, high-impact tail risk couples cost and schedule simultaneously, the joint P80 captures correlated worst-case behaviour the marginal P80s miss. Under All Series, joint sampling produces a small diversification benefit because cost-driver and schedule-driver risks fire independently.

Actions and implications

What is the ROI on risk treatment? The table below pairs each top driver with an illustrative treatment and the contingency reduction that would follow. Treatment effects are scenario estimates for prioritisation purposes; actual treatment effectiveness depends on project-specific factors.

Illustrative only, treatment definitions and effectiveness factors require workshop input. Rows are populated from the strawman Treatments sheet; remaining rows await workshop input. The methodology supports per-risk treatment descriptions and effectiveness estimates: populated rows render the workshop description and the computed reduction (driver value × supplied effectiveness %); rows without a workshop-supplied effectiveness value render 'Workshop input required' in the treatment-description and estimated-reduction columns.

Risk	Type	Current contribution	Illustrative treatment	Estimated reduction
WT-008	Cost	$1.59M (P80)	Workshop input required	Workshop input required
WT-009	Cost	$1.45M (P80)	Workshop input required	Workshop input required
WT-012	Cost	$1.10M (P80)	Workshop input required	Workshop input required
WT-021	Schedule	17.54 days (prob-weighted mean)	Workshop input required	Workshop input required
WT-008	Schedule	13.90 days (prob-weighted mean)	Workshop input required	Workshop input required
WT-022	Schedule	11.34 days (prob-weighted mean)	Workshop input required	Workshop input required

Treatment effort directed at the top drivers reduces contingency more than effort spread across the long tail. The ranking gives the project team a defensible prioritisation for treatment plans and a quantified sense of the contingency at stake against each treatment decision.

Recommendations

• Prioritise risk-treatment plans for the top schedule drivers: WT-021, WT-008, WT-022. Resolution of these risks narrows the headline contingency materially.
• Treatment effort focused on WT-008 ($1.59M P80 contribution) reduces cost contingency more than effort spread across the long tail.
• Use the per-risk diagnostic outputs (audit trail appendix) to drive targeted risk treatment plans for the top-3 cost and schedule drivers.

Methodology and caveats

• Joint sampling mechanism. The integrated Monte Carlo engine shares a single severity uniform per dual-coded risk per iteration, driving both the cost dispatch and the schedule PERT draw. This produces correlated cost and schedule severities for the 15 risks flagged as time-impacting in the project risk register. Cost-only risks use independent severity uniforms. The mechanism preserves marginal calibration with the standalone cost and schedule engines while enabling diversification benefit measurement at the joint percentile.
• Schedule classifications. Series, Hybrid, and Parallel classifications for the dual-coded risks reflect the conservative starting-point informed by current schedule structure. Sensitivity of the headline to classification choice is bounded and documented in the appendices.
• Full-network CPM replay. The schedule analysis replays the programme's full activity network on every iteration: each activity's duration is sampled, a forward and backward pass is run over the whole network, and the finish date is read from that pass. The critical path is therefore an output of each iteration rather than a subset fixed in advance, and an activity off the planned critical path contributes to finish-date slip whenever sampling puts it on the path. The deterministic critical path quoted in the appendices is the activity set with zero float under a single zero-risk pass; where chains tie on float, that set is a superset of any one longest path. Series, Hybrid and Parallel classifications are carried from the strawman and reported for context; under the full-network replay they do not multiply a risk's impact, because the network decides how a delay propagates.

Distribution choice per risk reflects the shape of the underlying uncertainty:

• PERT (default), three-point Optimistic / Most Likely / Pessimistic where the Most Likely value carries more weight than the extremes; the standard for contingency analysis.
• Triangular, three-point with no central-tendency weighting; used where the Most Likely value is genuinely uncertain.
• Lognormal, right-skewed continuous; used for cost categories with fat-tail exposure (e.g. commodity price spikes).
• Binary, fires fully or not at all; used for irreducible step-change events.

The methodology aligns with the following standards and recommended practices: AACE RP 44R-08 Risk Analysis and Contingency Determination Using Range Estimating; AACE RP 65R-11 Integrated Cost and Schedule Risk Analysis Using Risk Drivers and Monte Carlo Simulation of a CPM Model; AACE RP 57R-09 Integrated Cost and Schedule Risk Analysis Using Monte Carlo Simulation of a CPM Model (cited as future scope for full CPM-based replay); AACE RP 113R-21 Integrated Cost and Schedule Risk Analysis - Sampling and Computational Aspects; PMI PMBOK 7th edition (Project Risk Management); ISO 31000:2018 (Risk management guidelines).

Delay cost and risk cost impacts

Risk cost impacts exclude project time-related costs (preliminaries and site overheads). The model prices those once, through the day rate, on the finish delay that survives float in each iteration. Event-specific standby (idle plant or crew, demobilise / remobilise) is a direct cost and stays in the risk's cost impact. The day rate itself covers preliminaries and site overheads only, with no standby, so the same cost cannot enter through both the day rate and a risk's cost impact.

Left out of the base layer because they price project duration, which the delay cost already prices through the schedule replay: Duration (rows: Turbine transport & erection, Commissioning, Owners costs & PM; about $0.61M at the mean). The register is unchanged.

Convergence diagnostic: 95% bootstrap CI on the integrated P80 cost across iteration counts 1k → 100k. Band tightens as iterations grow; convergence is reached when further iterations no longer materially shift the band.

Monte Carlo simulation produces a sample, not the truth. The convergence diagnostic above shows how the headline figure stabilises as iteration count grows; a tight 95% bootstrap confidence band at the iteration count this run used indicates the simulation has run enough iterations for the headline to be reliable to within the band's width. The diagnostic measures Monte Carlo error only, it does not measure model uncertainty in the inputs (probability estimates, three-point distributions, Mode classifications) or in the methodology itself.

• The cost Monte Carlo engine groups risks by their underlying driver, common conditions that increase the joint probability of multiple risks firing together. A tight construction market that drives commodity prices up simultaneously raises the probability of multiple downstream risks; modelling those as independent would understate the joint tail exposure at exactly the contingency-relevant probability levels. Risks within a Correlation Group fire together more often than chance per the coefficient shown; risks across groups fire independently.
• The table below maps each active Correlation Group to its driver description, the member risks, and the joint-firing coefficient. Together with the per-risk Probability and three-point distributions in the register, the CG mapping is the third leg of the cost MC's correlation structure, the other two being within-CG joint sampling (Cholesky-decomposed correlated uniforms) and the shared severity uniform on the dual-coded risks for joint cost-and-schedule sampling.

Correlation group (CG) drivers

CG ID	Driver description	Member risks	Member risk count	Coefficient
CG01	Civil & Earthworks	WT-001, WT-002, WT-003, WT-004, WT-005, WT-006, WT-007	7	0.6
CG02	Turbine Supply & Erection	WT-008, WT-009, WT-010, WT-011, WT-012, WT-013	6	0.5
CG03	Electrical Reticulation	WT-016, WT-018, WT-019, WT-020	4	0.5
CG04	Grid Connection	WT-021, WT-022, WT-023, WT-024, WT-025	5	0.4
CG05	Balance of Plant / Substation	WT-026, WT-027, WT-028, WT-029, WT-030	5	0.5
CG06	Commissioning & Controls	WT-031, WT-032	2	0.5

Limitations of current analysis

• Per-party views are deferred to the next stage. The current deliverable presents the whole-project view only.
• Criticality assessment uses the conservative starting-point classification informed by current schedule structure
• Per-activity attribution of the finish to individual risks is deferred. The replay reports the project finish and each activity's criticality; the per-risk driver ranking beside it is the additive comparator basis, not the replay.
• Sensitivity analysis on individual risk probabilities and impacts (one-at-a-time, top-driver) is deferred to the next stage.
• The headline presents the event-risk contingency (Layer 2) on an input base estimate. The total-cost S-curve adds that base with its systemic uncertainty (Layer 1), sampled independently of the event risks; the base estimate itself remains an input, not an estimate this product makes.

None of these are limitations of the methodology; they are inputs not yet available or capabilities deferred to the next stage. When inputs land, the analysis extends without re-engineering.

What does one Monte Carlo iteration actually compute? The walkthrough below traces a single iteration from the simulation in concrete terms. Read it as one of the simulated futures the analysis explored; the distribution charts above are the ensemble of all such futures. Some risks listed in this iteration may sit outside the Page 3 top-cost-driver tornado - the walkthrough captures every firing risk in the simulated iteration regardless of overall ranking. Iteration-specific firings reflect the random draw, not the headline driver hierarchy.

Monte Carlo iteration #28647 (one of 100,000 simulated iterations, sitting at the 95th percentile of the schedule distribution): total cost $9.16M, schedule slip 133.71 days.

Risks that fired in this iteration:

• WT-001: cost $0.93M
• WT-002: cost $0.21M
• WT-003: cost $0.42M
• WT-005: cost $0.36M
• WT-006: cost $0.25M
• WT-007: cost $0.23M
• WT-016: cost $0.25M
• WT-019: cost $0.20M
• WT-020: cost $0.56M
• WT-021: cost $2.26M
• WT-022: cost $0.93M
• WT-023: cost $0.64M
• WT-025: cost $0.42M
• WT-031: cost $0.80M
• WT-032: cost $0.70M
• WT-001: schedule contribution 14.56 days
• WT-002: schedule contribution 1.87 days
• WT-003: schedule contribution 2.86 days
• WT-021: schedule contribution 77.43 days
• WT-022: schedule contribution 34.04 days
• WT-031: schedule contribution 9.77 days
• WT-032: schedule contribution 3.72 days

This is one of many simulated futures the analysis explored; the headline range and distribution charts above are the ensemble of all 100,000 such iterations.

Audit trail

Item	Value
Risk register file	wind_register.xlsx
Strawman methodology workbook	wind_register.xlsx
Run date	2026-11-03
Iterations	100,000
Random seed	42
Day rate	$35,000/wd
Reporting P-level	P80
Package version	0.1.0
Engine commit	db2f509642eba9ee9b2f7321ceed529aa8c2bd9c

Internal cross-validation: all four gates passed at the canonical 100,000-iteration production run. cost marginal gap $4,237 / 0.052%; schedule marginal gap 0.000d; top-3 cost contributors consistent (WT-008, WT-009, WT-012); top-3 schedule contributors consistent (WT-021, WT-008, WT-022).

To regenerate this deliverable: restore the listed register and strawman files to the versions specified in the audit trail, set SR9_SEED to the seed recorded in this run's provenance and execute RUN_FINAL.bat. Outputs reproduce within MC noise tolerance. This deliverable reflects the input data as recorded in the audit trail. Inputs evolve as the project progresses; re-running the analysis against current inputs produces an updated deliverable.

Input file hashes (SHA-256)

Item	Value
Risk register	a973af5d83e1298827e453261a0875ccda4376c4a381c3277786182e388ba135
Strawman	a973af5d83e1298827e453261a0875ccda4376c4a381c3277786182e388ba135
Schedule (P6)	f6c55af3331f2156138d371a366c78f25951308bac9c6205a1b299b2ef7f746b

Appendices

Mode-classification sensitivity

Methodology

Caveats

• The headline is a single Stage 3 distribution. Mode classifications play no part in it; they remain visible only in the Stage 2.5 comparator range beside the reconciliation.
• Schedule analysis operates on the full programme logic exported from the planning tool. Delays propagate along that logic; a gating activity outside a risk's cohort is refused before any iteration runs.
• Liquidated damages and IDC layers are deferred to a later stage pending commercial input confirmation.

References

• AACE 57R-09: Risk analysis and contingency determination using expected value (cost engine basis).
• AACE 65R-11: Integrated cost and schedule risk analysis using Monte Carlo simulation.
• PMI Practice Standard for Project Risk Management.

PERT distributions for cost and schedule severity per risk (opt/ml/pess from the project risk register). Lognormal available for selected event risks where right-tail behaviour matters. Bernoulli gates for risk firing. Cholesky-decomposed correlated uniforms for within-CG correlation.

Audit trail

Item	Value
Risk register file	wind_register.xlsx
Strawman methodology workbook	wind_register.xlsx
Run date	2026-11-03
Iterations	100,000
Random seed	42
Day rate	$35,000/wd
Reporting P-level	P80
Package version	0.1.0
Engine commit	db2f509642eba9ee9b2f7321ceed529aa8c2bd9c

Disclaimer

• Joint sampling mechanism. The integrated Monte Carlo engine shares a single severity uniform per dual-coded risk per iteration, driving both the cost dispatch and the schedule PERT draw. This produces correlated cost and schedule severities for the 15 risks flagged as time-impacting in the project risk register. Cost-only risks use independent severity uniforms. The mechanism preserves marginal calibration with the standalone cost and schedule engines while enabling diversification benefit measurement at the joint percentile.
• Schedule classifications. Series, Hybrid, and Parallel classifications for the dual-coded risks reflect the conservative starting-point informed by current schedule structure. Sensitivity of the headline to classification choice is bounded and documented in the appendices.
• Full-network CPM replay. The schedule analysis replays the programme's full activity network on every iteration: each activity's duration is sampled, a forward and backward pass is run over the whole network, and the finish date is read from that pass. The critical path is therefore an output of each iteration rather than a subset fixed in advance, and an activity off the planned critical path contributes to finish-date slip whenever sampling puts it on the path. The deterministic critical path quoted in the appendices is the activity set with zero float under a single zero-risk pass; where chains tie on float, that set is a superset of any one longest path. Series, Hybrid and Parallel classifications are carried from the strawman and reported for context; under the full-network replay they do not multiply a risk's impact, because the network decides how a delay propagates.
• Multi-party views deferred. The current deliverable presents the whole-project view only. EPC Contractor, Wind Farm Owner, and Shared stakeholder views are deferred to a later stage pending confirmation of the risk-share parameters and counterparty allocation rules.
• Day-rate basis. The day rate used to dollarise schedule slip is sourced from the Risk Register Variables sheet via the strawman methodology workbook. The same value is used by the standalone schedule engine, the integrated engine, and the executive summary so all reported dollar conversions are consistent.
• Strawman provenance. The strawman methodology workbook is the single source of truth for Mode classifications, day rate, commercial reference values, and run configuration. Its version date appears in the audit trail appendix. When the strawman is reissued, this deliverable re-runs against the amended state and supersedes the prior version.
• Engine versioning. The cost Monte Carlo engine, schedule MC engine, integrated cost+schedule engine, and final composition layer carry version stamps recorded in the audit trail appendix. The run date and seed are recorded so the deliverable is reproducible from the strawman snapshot it was generated against.

Artefacts written by this run

• BETA_executive_summary_2026-11-03.docx
• sra_stage3_standalone_PROD_DPReview_2026-11-03.docx
• sra_stage3_standalone_PROD_DPReview_2026-11-03.html
• the run folder (state cache and provenance)
• engine_log.txt
• run_diagnostic_2026-11-03.txt
• provenance.json

Provenance: input digests are per file, so a same-filename re-issue of any one input is diagnosable. Full paths are in the run's provenance.json.

• register: git blob id da4c09c5b498a8ea...
• xer: git blob id 73e5c92cfd2df97b...
• strawman: git blob id da4c09c5b498a8ea...

Cross-validation passed. All artefacts above are clear for review.