Cost Contingency in Project Estimates: How to Manage Uncertainty in Capital Projects

No capital project estimate is fully certain at sanction. Engineering is incomplete, commodity markets move, productivity deviates from plan, and site conditions reveal themselves slowly. Cost contingency is the structured allowance that prices this residual uncertainty into the budget, making the estimate honest about what is known and what is merely expected. Handled well, it is the single most important line item separating a defensible capital cost from a wishful one.

Contingency is also one of the most abused allowances on a project. It gets trimmed to hit an affordability target, drawn down to fund new scope, or applied as a flat percentage without any link to the underlying risk profile. This article covers what contingency is (and what it is not), how it differs from management reserve, the three main sizing methods and when each applies, how P50/P80 framing drives funding decisions, and the governance that keeps contingency from quietly disappearing during execution.

Contingency vs Management Reserve

The first conceptual mistake most project teams make is treating contingency and management reserve as interchangeable pots of money. They are not. Contingency is an allowance for identified uncertainties within the defined scope — design development, productivity variance, commodity movement, weather, site conditions. It is part of the cost baseline, is controlled by the project team, and is drawn down against the specific risks it was sized for.

Management reserve sits above the baseline. It funds events outside the defined scope — a major regulatory change, a step-change market disruption, an unforeseen scope expansion — and is released only by the sponsor or steering committee. The two allowances live in different columns of the funding stack and answer to different authorities for a reason: mixing them destroys the integrity of both.

A $216m project might carry $36m of contingency inside the baseline and a further $20m of management reserve above it. The baseline — the performance reference — is $216m. The funding envelope approved by the board is $236m. Variance reporting is always against the baseline, not the envelope, because the envelope is not a plan.

Three Ways to Size Contingency

Three methods dominate practice, and the right one depends on project size, stage, and the cost of being wrong.

The percentage method applies a rate to the base estimate — typically 30–50% for a Class 5 concept estimate, dropping to 5–10% at Class 1. It is fast, easy to benchmark, and defensible on small or well-understood projects. Its weakness is that a flat percentage hides the risk profile: two projects of the same size can carry very different risk exposure, and a single number tells you nothing about where the uncertainty actually sits.

The expected-value method works from a risk register. Each risk carries a probability and a cost impact; the product is its expected cost, and the sum across the register is the contingency. A risk with a 30% probability and a $10m impact contributes $3m. This is the first step toward a risk-informed allowance and forces the team to articulate what it is holding contingency against. It is stronger than a percentage but still deterministic — it produces a single number and no view of the distribution.

The probabilistic method — Monte Carlo simulation over a cost model with ranged inputs and risk events — produces a full cost distribution. From that distribution you read off P50 (the median outcome), P80 (80% confidence of not exceeding), and P90 (conservative). Contingency is set as the gap between the P50 base estimate and the chosen confidence level. Most mining, energy, and infrastructure sanction processes now require a probabilistic analysis above a threshold project size.

P50, P80, and the Funding Decision

Confidence-level framing matters because it turns contingency from a number into a choice. P50 is the median — the expected outcome if the project ran many times. Funding at P50 means an (approximately) equal probability of overrun and underrun, which is acceptable across a large portfolio but rarely at a single-project level. P80 means there is a 20% probability of exceeding the approved budget: the standard most capital-heavy industries sanction against. P90 is conservative and typically reserved for projects where overrun is existential.

The practical implication is that the same project legitimately has different contingency numbers depending on the question being asked. A sponsor asking “what should I sanction?” wants P80. A portfolio manager asking “what should I hold across twenty projects?” cares about the P50 aggregate. A lender asking “what debt capacity is safe?” may require P90. Contingency is not one number; it is a point on a curve, and the estimate package should show the curve, not just the point.

Contingency should also decrease as design matures. The same project that carried 30% contingency at Class 4 should carry 10–15% at Class 2, because uncertainty has been retired — not because the risk magically vanished. Contingency that does not decline with definition is a signal the team has not actually closed out the risks it was holding money against.

Governance, Draw Rules, and Reporting

Sizing contingency correctly is only half the problem; the other half is defending it through execution. Without governance, contingency is drawn down not against risks that materialised but against scope change, estimate errors, and general overrun — and by the time anyone notices, there is nothing left to cover the real tail events in the last six months of the project.

Three rules keep contingency intact. First, draw only against logged risks. Each drawdown must reference a specific risk or risk event from the register; if no line-item exists, the team is either funding scope change (which belongs in the baseline via formal change control, not in contingency) or acknowledging the register was incomplete. Second, retire contingency as risks close. When a risk is retired without occurring — the weather window passed, the permit was granted, the soils report came back clean — the associated contingency is released back to the project or the sponsor, not quietly reallocated to cover something else. Third, track draw split: the ratio of contingency drawn against identified risks versus drawn against scope change. A high scope-change share is the clearest early signal that the change-control system is broken.

These rules tie directly to the change-management discipline. Scope additions should never be funded from contingency; they should raise a formal change and update the baseline through the delegation-of-authority matrix. Contingency exists to cover the risks inside the estimate basis, not to finance new work.

Key Takeaways

• Contingency is an allowance for identified uncertainties inside the defined scope; management reserve sits above the baseline and covers events outside it.
• Percentage sizing is defensible early; expected-value from a risk register is stronger; probabilistic simulation with P50/P80/P90 is the sanction-grade approach for large projects.
• P80 is the typical single-project funding point; P50 is appropriate at portfolio level; the estimate should present the distribution, not one number.
• Contingency should decline as definition improves — a flat contingency over time usually means retired risks are not being released.
• Draw only against logged risks, retire contingency when risks close, and track scope-change draws separately — this is how contingency survives execution.

Related reading

• Change Management and the Cost Baseline: How to Control Scope Change on Capital Projects
• Risk Registers in Cost Estimating: How to Translate Risks into a Defensible Contingency