By Jim Parke |
A major project is approved on a promise: the expected benefits justify the estimated cost, risk and disruption. Years later, its budget may have risen, its design may have changed and its completion date may have shifted. The project team can still be working effectively against its current plan. But someone must ask a different question: does the investment still make sense on the terms now before us?
North East Link in Melbourne illustrates why that question matters. In its February 2025 case study, the Victorian Auditor-General reported an aggregate total estimated investment of $26.21 billion across the road and tunnel packages, up from an original approved $15.64 billion: an increase of 67.58 per cent. The audit recorded a six-quarter delay against the original completion date. Yet both the delivery agency and the auditor rated cost and time green against the current approved budget and schedule. Benefits measurement was amber: some expected benefits lacked sufficiently specific baseline data, and responsibility for measuring longer-term economic benefits was unclear. Tunnelling has since finished, and the government says the tunnels are due to open to traffic in 2028.
The audit reported the delivery agency’s explanation that higher than expected contract costs arising from procurement contributed to the increase. It also recorded design changes and industry capacity and supply chain pricing pressures. The government assessed that the added costs of the design changes were necessary to improve the future asset. That is a reason to examine the revised scope and its benefits, not simply to compare two budget numbers.
Those findings do not establish that North East Link should have been cancelled or that its benefits will not materialise. They show how a change in cost can make the original decision harder to evaluate, particularly when the benefits are not yet fully measurable. The lesson applies to any strategic project whose business case has changed during delivery.
Compare the forecast with the outside view
The estimate for a major project usually begins inside the project. Engineers, contractors and advisers identify work packages, price known risks and model uncertainty. That work is essential, but it can be vulnerable to assumptions shared by the people closest to the proposal.
The outside view, associated with Bent Flyvbjerg’s work on major projects, starts elsewhere: with comparable projects and what actually happened to their costs and schedules. Reference class forecasting places the proposed project against a distribution of past outcomes. Australian Government cost estimation guidance describes the need for a reference set large enough to be meaningful and sufficiently consistent to be comparable. If the class is too broad, it tells us little about the particular project. If it is too narrow, there may be too few observations to draw a reliable conclusion.
The purpose is not to replace engineering judgment with a single uplift. The department treats a deterministic reference class estimate as a possible benchmark, rather than its preferred way to estimate contingency; its funding rules require probabilistic cost estimation for projects seeking Commonwealth funding above $25 million. The outside view asks why this project is expected to perform differently from the historical pattern. What evidence supports that judgment? Are the procurement approach, ground conditions, interfaces and market capacity genuinely comparable? What happened to earlier projects when the same assumptions proved wrong?
An outside view is especially useful before approval, when a sponsor still has room to choose a smaller intervention, a different sequence or no project at all. It remains useful later, when a changed forecast should be tested against both the original estimate and current evidence.
Understand what a probability estimate can tell you
A cost estimate expressed at P50 means that, under the model’s assumptions, there is a 50 per cent probability of the modelled cost not being exceeded. At P90, the corresponding probability is 90 per cent. The difference matters to a sponsor deciding how much funding to commit and what contingency to retain.
But a model is only as informative as its inputs and boundaries. It may combine carefully assessed risks while missing an issue no one identified. Historical data cannot predict extreme outcomes absent from the precedents on which it relies. A model may also reflect optimistic assumptions about market prices, approvals or the coordination of multiple work packages. A sophisticated calculation does not independently validate those assumptions.
A sponsor should therefore ask what is outside the estimate as well as what is inside it. Does the figure include land, financing, interfaces, operations and the full scope needed to realise the benefit? Which risks have been transferred by contract, and which ultimately return to the owner? What changed between the original estimate and the current one? These questions are more useful than treating a P90 figure as a guarantee.
Reopen the benefits case when the project changes
A cost increase does not, by itself, prove that a project has ceased to be worthwhile. Its scope may have expanded, its benefits may have changed, or its risk profile may have improved. Equally, a project can stay within its revised budget while delivering less than was promised at approval.
The business case should therefore be revisited as a whole. Compare like with like: the same scope, price basis, time horizon and treatment of risk. Update the costs, forecast benefits, delivery timetable and realistic alternatives. Distinguish a change to the asset from a change to the value the asset is expected to create. Avoid dividing an old benefits estimate by a new headline budget and presenting the result as a definitive benefit–cost ratio; that shortcut may mix different bases and omit changed benefits or costs.
The North East Link audit offers a second reason to revisit benefits. It reported that some benefits needed more specific baseline data and that it was unclear which agency was best placed to measure longer-term economic effects. It also found that some responsibilities in the benefits management plan still referred to an earlier project office. Those are questions of ownership as much as measurement. Who will collect the evidence after construction? Who is accountable if the expected change in travel, safety or local conditions does not occur? And what decision can be made before the evidence becomes available?
Give the board a real choice
When a project is under way, continuing can feel inevitable. Money has been spent, contracts signed and teams assembled. None of those facts makes the next dollar automatically worthwhile. Nor does a cost increase automatically make cancellation sensible: the cost of stopping, contractual obligations and the value of completed work must also be considered.
At a material change point, the board or sponsor should receive a fresh decision paper. It should set out the expected cost to complete, remaining risks, updated benefits and the consequences of delay. It should compare continuation with feasible changes to scope, sequence or delivery approach, and, where credible, suspension or exit. It should identify what is already committed and what can still be avoided. It should also explain what information would change the recommendation.
That is a governance task, not a rebuke to the delivery team. A team can meet its milestones and still be executing an investment that deserves renewed scrutiny. Independent challenge helps the sponsor separate progress against the latest plan from the continuing case for the project.
Approval is therefore a continuing responsibility. The decision to begin a major project should specify when its case will be reopened and who has authority to recommend a different course. North East Link’s green cost and time ratings measured performance against the current approvals; the same audit recorded a 67.58 per cent increase in estimated investment and a six-quarter delay from the original approvals. Both views are true. The sponsor’s task is to hold them together and test whether the expected outcome still justifies the cost of reaching it.
Further reading
- Victorian Auditor-General’s Office, North East Link (all roads packages): case study (February 2025).
- Victoria’s Big Build, North East Link construction.
- Australian Government, Guidance Note 3B: Deterministic contingency estimation.
- Australian Government, Guidance Note 3A: Probabilistic contingency estimation.