Case Study – Cost and Schedule Overrun on an Australian Infrastructure Project

Introduction

Cost and schedule overrun remains the most persistent performance problem in Australian public infrastructure delivery. Grattan Institute analysis found that transport projects announced before a business case or published cost estimate overran their initial budgets by an average of 24 per cent (Terrill & Danks, 2016). Audits by the Australian National Audit Office (2022) and the Victorian Auditor-General’s Office (2021) identify weak early estimating, immature scope definition and passive contract administration as recurring rather than isolated causes.

This case study examines the Bunyarra Corridor Upgrade, a hypothetical A$180 million regional road and rail project delivered by a state transport authority in regional Victoria. The project is 18 months into a 36 month program and is both late and over budget. The analysis quantifies performance at the month 18 data date using earned value management (EVM) as described in the PMBOK Guide (Project Management Institute [PMI], 2021), diagnoses the divergence from baseline, and proposes corrective actions and estimating reforms. Figures are invented but internally consistent.

Project Background and Governance

Scope, Funding and Delivery Model

The Bunyarra Corridor Upgrade duplicates 42 kilometres of regional arterial highway, grade separates three level crossings, and constructs a six kilometre rail passing loop with an upgraded station. Funding is shared equally between the Commonwealth, under the Infrastructure Investment Program, and the state, and the sponsor is the state transport authority, referred to here as the Authority. The business case applied Infrastructure Australia’s assessment framework, although the project fell below the threshold for mandatory independent evaluation, and was approved with a benefit cost ratio of 1.42 and a P50 contingency of 8 per cent.

Delivery is by a single design and construct contract awarded to a tier two contractor on amended AS 4000-1997 general conditions (Standards Australia, 1997). The contract transfers design, ground risk above a nominated baseline and program risk to the contractor, while the Authority retains land acquisition, utility approvals and rail possession scheduling. That allocation matters, because the retained risks are precisely those that have since driven the schedule slip.

Governance and Control Environment

Governance is exercised through a project control group chaired by the Authority’s executive director of infrastructure, with quarterly board reporting. Risk is managed under a register maintained in accordance with AS ISO 31000:2018 (Standards Australia, 2018), and performance is reported monthly using EVM. A gateway review was completed at investment decision, but the readiness for service review was deferred twice. Infrastructure Australia (2021) argues that assurance adds little value where reviews become reporting obligations rather than decision gates.

Performance Status at Month 18

Table 1 presents the month 18 status by work package against a budget at completion (BAC) of A$180.0 million. Planned value (PV) is the authorised budget for work scheduled by the data date, earned value (EV) the budgeted cost of work completed, and actual cost (AC) the cost incurred for that work (Larson & Gray, 2021; PMI, 2021).

Table 1: Earned value status at the month 18 data date, by work package (A$ million)

Worked Earned Value Calculations

Applying the standard EVM formulae to the project totals in Table 1 produces the following position, with each calculation showing the formula, the substitution and the result.

• Cost variance: CV = EV – AC = 91.6 – 118.3 = -A$26.7 million
• Schedule variance: SV = EV – PV = 91.6 – 110.5 = -A$18.9 million
• Cost performance index: CPI = EV / AC = 91.6 / 118.3 = 0.774
• Schedule performance index: SPI = EV / PV = 91.6 / 110.5 = 0.829
• Estimate at completion: EAC = BAC / CPI = 180.0 / 0.774 = A$232.6 million
• Variance at completion: VAC = BAC – EAC = 180.0 – 232.6 = -A$52.6 million
• To complete performance index against the original budget: TCPI = (BAC – EV) / (BAC – AC) = (180.0 – 91.6) / (180.0 – 118.3) = 88.4 / 61.7 = 1.433
• To complete performance index against the revised forecast: TCPI = (BAC – EV) / (EAC – AC) = 88.4 / (232.6 – 118.3) = 88.4 / 114.3 = 0.773

Two ratios sharpen the interpretation: progress in budget terms is EV / BAC = 91.6 / 180.0 = 0.509, while budget consumed is AC / BAC = 118.3 / 180.0 = 0.657. With exactly half the duration elapsed, the project has converted two thirds of its money into slightly more than half its scope.

The TCPI results are the most diagnostically useful figures in the set. Recovering the original budget would require the remaining A$88.4 million of work to be earned for A$61.7 million, an efficiency of 1.433 against a demonstrated 0.774. PMI (2021) treats a TCPI more than roughly 0.10 above cumulative CPI as evidence that a baseline is unachievable. The TCPI against the revised EAC is 0.773, effectively identical to the current CPI, confirming A$232.6 million as the internally consistent forecast.

Table 2 disaggregates the variances and indices, because a composite CPI conceals the loss.

Table 2: Variances, performance indices and forecast outturn by work package (A$ million)

Earthworks and geotechnical treatment alone account for A$10.2 million, or 38 per cent, of the cost variance and 40 per cent of the schedule variance, despite representing only 25.6 per cent of the BAC. Rail systems and signalling record the worst SPI at 0.729 while showing the smallest cost variance, the signature of a work package that has barely started rather than one executed inefficiently. The work package forecasts sum to A$232.3 million against the project figure of A$232.6 million, a rounding artefact.

Figure 1 illustrates the divergence between the three cumulative curves, showing the classic diagnostic pattern: actual cost above planned value, earned value below it, so the project is spending faster and producing slower.

The schedule position requires more care than SPI alone conveys. A naive duration forecast divides baseline duration by SPI: 36 / 0.829 = 43.4 months, implying a 7.4 month delay. However, SPI is a cost denominated measure that converges mathematically to 1.0 at completion regardless of lateness, so it degrades as a forecasting instrument (Vanhoucke, 2018). Earned schedule corrects this by expressing progress in time units. Planned value reached A$85.0 million at month 15 and A$110.5 million at month 18, so the point at which the baseline planned to have earned today’s A$91.6 million is found by interpolation: ES = 15 + 3 x [(91.6 – 85.0) / (110.5 – 85.0)] = 15.8 months. The time based variance is SV(t) = ES – AT = 15.8 – 18.0 = -2.2 months, the index is SPI(t) = ES / AT = 15.8 / 18.0 = 0.878, and the duration estimate is 36 / 0.878 = 41.0 months. Reporting a five month delay rather than seven, with each assumption stated, is far more defensible to a control group.

Root Cause Analysis

Scope Creep and Uncontrolled Change

Fifty-one variations totalling A$14.6 million have been approved since award, of which only nine were assessed against the baselines beforehand; the rest were authorised at site level under a delegation intended for minor works. Love et al. (2016) demonstrate that scope change is rarely a single large decision, but accumulates through many small, individually defensible approvals whose aggregate effect is invisible until it surfaces in the actual cost. Without the integrated change control required by the PMBOK Guide (PMI, 2021), no register reconciled approved changes to the earned value baseline, which had quietly ceased to describe the contracted scope.

Optimism Bias in the Original Estimate

The business case adopted a P50 contingency of 8 per cent, low for a project combining brownfield rail interfaces with 42 kilometres of greenfield earthworks. Flyvbjerg (2021) identifies optimism bias and strategic misrepresentation as the dominant behavioural causes of overrun, the first unintentional and the second a rational response to a funding process that rewards low estimates. Both are visible here: the estimate was built bottom up from a concept design at roughly 20 per cent completion, and the submission was prepared to an announced political commitment rather than a tested cost range. Escalation compounded this, with Australian Bureau of Statistics (2024) producer price indexes for road and bridge construction inputs rising well above general inflation, against a contract indexing only 30 per cent of the sum.

Geotechnical Risk and Latent Conditions

Site investigation before award comprised 34 boreholes across 42 kilometres, well below industry practice for variable basalt and expansive clay. Rock encountered shallower than the geotechnical baseline report predicted across chainages 12 to 19 forced a change from bulk excavation to drill and blast, and a subgrade redesign. Under clause 12 of AS 4000-1997 the contractor is entitled to costs arising from latent conditions that could not reasonably have been anticipated (Standards Australia, 1997), so the saving from curtailing investigation returned many times over as a client borne variation. Transferring ground risk contractually does not eliminate it; pricing it without adequate data simply converts it into a claim (Infrastructure Australia, 2021).

Contractor Claims and Contractual Behaviour

The contractor has lodged claims totalling A$18.4 million, of which A$4.1 million is certified and sits within the actual cost. The unresolved A$14.3 million is not captured in the CPI derived EAC, so the downside forecast is 232.6 + 14.3 = A$246.9 million, giving VAC = 180.0 – 246.9 = -A$66.9 million, an overrun of 37.2 per cent. Superintendent responsiveness contributes: clause 41 of AS 4000-1997 requires timely assessment, yet average turnaround is 68 days against a contractual 28, allowing claims to consolidate into a global position that is dearer to resolve.

Stakeholder Driven Change

Two local councils and a community reference group secured changes to shared user paths, noise treatment and landscaping after the design was nominally frozen, adding A$3.1 million and eleven weeks of rework. The Victorian Auditor-General’s Office (2021) finds repeatedly that consultation conducted after rather than before design freeze converts legitimate community input into rework.

Corrective Action and Risk Response

Table 3 sets out the plan endorsed by the project control group, using the treatment vocabulary of AS ISO 31000:2018 (Standards Australia, 2018) and prioritising actions by their contribution to the variances in Table 2.

Table 3: Corrective action and risk response plan from month 19

Four of the seven treatments are governance rather than engineering interventions, reflecting the diagnosis: the corridor’s technical difficulty was real, but the loss was crystallised by control failures around it. CA3 deserves emphasis, since reporting against a baseline that a TCPI of 1.433 shows to be unattainable damages the credibility of every other number in the report, and the Australian National Audit Office (2022) criticises agencies for retaining superseded baselines.

Lessons Learned and Reference Class Forecasting

The central lesson concerns estimating method rather than execution. Reference class forecasting sets a budget from the distribution of outturns on comparable completed projects, displacing the inside view built from a project’s own assumptions (Flyvbjerg & Gardner, 2023). Suppose the reference class of 42 Australian regional road and rail upgrades shows a mean overrun of 26 per cent and an 80th percentile overrun of 41 per cent. The uplifted estimates would be:

• P50 budget = base estimate x (1 + mean uplift) = 180.0 x 1.26 = A$226.8 million
• P80 budget = base estimate x (1 + P80 uplift) = 180.0 x 1.41 = A$253.8 million

The CPI based forecast of A$232.6 million sits between these values, and the downside forecast of A$246.9 million just below the P80 figure, so the project is performing almost exactly as the outside view would have predicted at approval. The failure was not that it overran; it was that it was funded at a number nobody had reason to believe. Had approval been sought at a P80 envelope of A$253.8 million, the present forecast would fall inside budget and the governance conversation would concern delivery rather than funding rescue.

Three further lessons follow. First, geotechnical investigation should be budgeted as risk reduction rather than preliminaries, at 1 to 3 per cent of capital expenditure where ground is variable; the A$1.9 million notionally saved preceded a A$10.2 million cost variance. Second, change control must operate as a gate with mandatory earned value impact assessment, not a register maintained after the fact (Larson & Gray, 2021). Third, gateway reviews must be decision points with authority to halt, since deferring a review on a project with a CPI of 0.774 removes the only independent check when it is most needed (Australian National Audit Office, 2022).

Conclusion

At month 18 the Bunyarra Corridor Upgrade has earned A$91.6 million of value for A$118.3 million of expenditure, giving a cost variance of -A$26.7 million, a schedule variance of -A$18.9 million, a CPI of 0.774 and an SPI of 0.829. The estimate at completion of A$232.6 million implies a variance at completion of -A$52.6 million, and a to complete performance index of 1.433 confirms the original budget is unrecoverable. Earthworks generated most of the loss, while signalling represents the principal remaining schedule threat. The causes were an underinvestigated corridor, an optimistically contingent estimate, delegated change approval without baseline discipline, slow claim assessment and post freeze stakeholder change, none unique to this project. The remedy is equally consistent across the Australian audit literature: fund infrastructure at a probabilistically defensible estimate drawn from comparable completed projects, invest early in the site information that prices the dominant risk, and give assurance gates authority to act on what the indices show.

References

Australian Bureau of Statistics. (2024). Producer price indexes, Australia. ABS.

Australian National Audit Office. (2022). Governance and delivery of major infrastructure investment programs. ANAO.

Flyvbjerg, B. (2021). Top ten behavioral biases in project management: An overview. Project Management Journal, 52(6), 531-546.

Flyvbjerg, B., & Gardner, D. (2023). How big things get done. Macmillan.

Infrastructure Australia. (2021). Delivering outcomes: A roadmap to improve infrastructure industry productivity and innovation. Infrastructure Australia.

Larson, E. W., & Gray, C. F. (2021). Project management: The managerial process (8th ed.). McGraw-Hill Education.

Love, P. E. D., Ahiaga-Dagbui, D. D., & Irani, Z. (2016). Cost overruns in transportation infrastructure projects: Sowing the seeds for a probabilistic theory of causation. Transportation Research Part A: Policy and Practice, 92, 184-194.

Project Management Institute. (2021). A guide to the project management body of knowledge (PMBOK guide) (7th ed.). Project Management Institute.

Standards Australia. (1997). General conditions of contract (AS 4000-1997). Standards Australia.

Standards Australia. (2018). Risk management: Guidelines (AS ISO 31000:2018). Standards Australia.

Terrill, M., & Danks, L. (2016). Cost overruns in transport infrastructure. Grattan Institute.

Vanhoucke, M. (2018). The data-driven project manager: A statistical battle against project obstacles. Apress.

Victorian Auditor-General’s Office. (2021). Delivering major transport infrastructure projects. Victorian Auditor-General’s Office.