Contingency Drawdown: How to Manage Contingency on Capital Projects
How to manage contingency drawdown on capital projects: build the curve from risk retirement, re-test adequacy at milestones and control every transfer.
Twenty months into a four-year, C$1.4-billion processing plant expansion delivered under an engineering, procurement and construction management (EPCM) agreement, the contingency drawdown report shows 55% of a C$160-million contingency drawn at 41% complete. The controls lead calls the position healthy: C$72 million remains, almost double the C$38-million expected value of the open risk register. Both numbers are correct, and neither answers the sponsor’s real question — whether the contingency left can carry the remaining work to completion at the confidence level the project was funded at. Contingency is sized with rigor at sanction and then tracked for years as a declining balance, which is how projects reach 80% complete with a reserve that looked adequate every month until the month it was gone.
This article covers why contingency drawdown is a forecast rather than a spend curve, how to build the drawdown curve from risk retirement, how to re-test adequacy at execution milestones, the transfer and release mechanics that keep the audit trail intact, how contract strategy decides who holds which pot, and the four indicators for the monthly cost review. The aim is a contingency position that can be defended every month — not a balance nobody questions until it runs out.
Contingency Drawdown Is a Forecast, Not a Spend Curve
At sanction, contingency is sized to a confidence level — commonly P80 for a single project — as the gap between the base estimate and that percentile of the total cost distribution. It is not a savings account: AACE International states that “contingency, by AACE’s definition, is a cost expected to be spent” (RP 58R-10, Escalation Estimating Principles and Methods Using Indices, May 2011). A project that finishes with a large unspent balance was not necessarily well run; it may simply have been over-funded.
The most common drawdown report plots percent of contingency drawn against percent complete and calls the project healthy while the first number stays below the second. That says very little, because risk does not retire in proportion to progress. Design growth and procurement pricing risk retire early, at issued-for-construction (IFC) drawings and purchase order award; subsurface conditions, field productivity and commissioning risk retire late. A project at 41% complete may have retired most of its procurement exposure while carrying nearly all of its productivity exposure — so drawn-below-complete can look healthiest exactly as the most expensive risks arrive.
The useful question is adequacy: does the remaining contingency cover what the remaining work requires at the funding confidence level? That is a forecast to be recomputed, not a balance to be watched. The Federal Transit Administration (FTA) builds the principle into its oversight of transit capital projects: Oversight Procedure 40, Risk and Contingency Review (October 2023), has the project sponsor prepare “preserved contingency (or ‘contingency drawdown’) curves that define minimum levels of cost or schedule contingency to be preserved as a function of project advancement” and carry them in its Risk and Contingency Management Plan. The curve is a floor to be defended, not a budget to be spent down.
The question is never how much contingency is left — it is whether what is left still covers what can still go wrong.
Building the Contingency Drawdown Curve from Risk Retirement
The planned drawdown curve comes out of the model that sized contingency at sanction. Take the Monte Carlo cost risk model and attach a retirement window to every source of uncertainty in it — each risk event and each range on the base estimate — marking when it either materializes or can no longer occur. Re-run the model as at each future milestone with retired items removed. P80 of remaining cost minus the base estimate-to-complete is the contingency the remaining work requires; joined up, those points form the required-contingency curve — the floor.
Typical retirement windows on an industrial project:
- Design growth. Retires discipline by discipline as drawings reach IFC status, not at one project-wide date.
- Procurement pricing. Retires at purchase order award for fixed-price packages; escalation-indexed and reimbursable packages carry exposure until final invoice.
- Subsurface and site conditions. Retire when excavation, piling and foundations are complete.
- Field productivity. Retires progressively with earned hours and stays open longest on a labor-intensive project.
- Commissioning and startup. Retires only at handover — the tail that early releases most often starve.
The resulting curve steps down early as definition and procurement risk retire, slopes through construction as productivity exposure is earned out, and ends in a commissioning tail that teams routinely forget to protect. A straight line from sanction to zero is wrong in both directions: it holds contingency the project no longer needs during procurement and releases contingency it still needs late in construction. Plot planned contingency held — sanction contingency less forecast draws — above that floor. Where held sits well above required, there is a release opportunity; where it crosses below, the project is underfunded at its sanctioned confidence level, whatever the balance says.
Re-Forecasting Contingency at Execution Milestones
The plan curve is a hypothesis, tested by re-running the risk model at defined execution gates — after IFC by discipline, after major package awards, at roughly half of construction, at mechanical completion — and whenever a single trend or risk event moves the forecast past a threshold set in the contingency management plan. The U.S. Department of Energy (DOE) sets a minimum cadence for the register itself: “Post CD-1, the Risk Register should be evaluated at least quarterly” (DOE G 413.3-7A, Risk Management Guide, Change 2, November 2021), CD-1 being DOE’s approval of the selected alternative and its cost range. The quantitative model needs its own triggers, because a quarterly register review can update every line and never ask the adequacy question.
A re-run removes realized risks, whose cost now sits in control account forecasts through transfers, and retired ones; adds emergent risks; and re-ranges the remaining base estimate-to-complete using actual performance — the productivity factor the field has achieved, not the one the estimate assumed. Required contingency is the P80 of remaining cost minus the base estimate-to-complete; the adequacy ratio is remaining contingency divided by that requirement.
A worked example
Take the expansion from the opening. At 41% complete the base estimate-to-complete is C$765 million and the re-run puts P80 of remaining cost at C$853 million, so the remaining work requires C$88 million. The project holds C$72 million: an adequacy ratio of 0.82, or C$16 million short of its sanctioned confidence level, despite a balance nearly double the open register’s expected value. The gap is systemic uncertainty on the remaining base — quantity growth and productivity ranges that were always inside the sanction P80 but never appear as register lines. That is why comparing contingency to register expected value flatters the position.
An adequacy ratio below 1.0 calls for a decision, not a footnote: mitigation that retires exposure, a funding request to the sponsor while options still exist, or recovery measures held in reserve for this moment. FTA’s OP 40 calls identifying such secondary mitigation “a prudent, but not required, project sponsor effort.” Nobody requires it on privately funded projects either, which is why a pre-agreed list of deferrable scope, alternative specifications and resequencing options — each with a value and a decision date — rarely exists when needed.
Transfers, Releases and the Audit Trail
Contingency has two legitimate exits. A transfer moves it into a control account when a risk it was sized for materializes. A release returns it to the sponsor, or to unallocated reserve, when a risk retires without occurring. Scope change is the third way contingency gets moved, and the illegitimate one: new scope belongs in the baseline through formal change control, with its own funding decision.
Every transfer should carry five things: the risk ID it draws against, the realized cost and its basis, the receiving control account, the approver under the delegation-of-authority matrix, and the date. A transfer without a risk ID is either unrecorded scope change or proof that the register was incomplete, and it should be reported as one or the other rather than quietly processed.
Releases are where the discipline usually fails. Contingency attached to a retired risk stays in the pot “just in case” and funds the next overrun nobody wanted to report. The rule is mechanical: when the re-run shows the requirement has fallen, release the excess above the floor at the next monthly cycle, with a record. Hoarding hides performance; releasing below the floor starves the commissioning tail.
The earned value world already has the matching rule for contractor reserves: DOE’s earned value management training material states that management reserve “may not be used to ‘pad the future’ to offset previously accumulated overruns or under runs.” Owner’s contingency deserves the same discipline: transfer it when a risk is realized and the forecast moves, and keep reporting variance against the original control budget as well as the current one, so the transfer funds the cost without erasing the performance signal.
Who Holds the Contingency Depends on the Contract
Drawdown management assumes someone owns the pot. Contract strategy decides who, and getting it wrong produces either double-counted contingency or none at all.
- Lump-sum EPC. Under a lump-sum engineering, procurement and construction contract, the contractor prices its own risk into the price; that allowance is invisible to the owner and not the owner’s to draw. Owner contingency covers retained risks only — scope gaps between the contract and the business case, owner-furnished items, interfaces, permits and changes.
- EPCM or reimbursable. The owner carries the cost risk, so the whole contingency is the owner’s even when the EPCM contractor administers transfers; its delegation limits belong in the contingency management plan.
- Target cost or alliance. Contingency typically sits inside the target, with overruns and underruns shared through the pain/gain mechanism, so drawdown reporting has to be joint.
U.S. federal practice draws the line explicitly. On DOE projects, contractor management reserve sits within the contract budget base and is managed by the contractor, while contingency is “a cost reserve owned by the customer,” held outside the contractor’s performance baseline, and drawing on it requires “specific, written DOE approval.” Two pots, two owners, two drawdown curves. Report them separately: a single “contingency remaining” line that blends contractor reserve the owner cannot spend with the owner’s own overstates the cover available. Keep owner contingency out of owner’s cost lines as well — a buried reserve cannot be drawn down transparently because nobody can see it.
Mixed strategies need the most care: where lump-sum civil packages sit beside a reimbursable mechanical contract, size and draw contingency package by package according to who carries the risk, not as a blanket percentage.
Four Indicators for the Monthly Cost Review
Four numbers carry the contingency position, and each should be shown with its trend over the last six reporting periods:
- Adequacy ratio. Remaining contingency divided by required contingency from the latest re-run. Below 1.0 the project is underfunded at its sanctioned confidence, however large the balance looks; set the escalation trigger in the contingency management plan before it is needed.
- Drawdown against plan. Contingency held against the planned curve built from risk retirement, not against percent complete. Draws running ahead of the curve in a phase where little exposure was expected to retire are the earliest signal available.
- Burn velocity. The average monthly draw over the last three months, projected across the remaining duration. If the projection exhausts contingency before mechanical completion, the re-forecast is overdue regardless of the calendar.
- Draw split and unlinked transfers. The share of draws against identified risks versus scope, plus the count of transfers with no risk ID. The second number should be zero every month.
None of this needs new software — just a risk model kept alive after sanction, a transfer log with mandatory fields, and a monthly cycle that asks the adequacy question. Managed that way, contingency stops being the line that quietly absorbs bad news and becomes the most forward-looking number in the cost report.
Key Takeaways
- Contingency drawdown is a forecast of remaining uncertainty, not a spend curve, and percent drawn against percent complete says little about whether the remaining contingency is adequate.
- Build the planned drawdown curve from the retirement window of every risk and range in the sanction model, and treat it as a floor — the principle FTA’s Oversight Procedure 40 applies to transit capital projects.
- Re-run the risk model at execution gates and on trigger events, and compare remaining contingency with the P80 requirement of the remaining work rather than with the expected value of the open register.
- Move contingency only by transfer against a realized risk or release of a retired one, record a risk ID on every movement, and keep reporting variance against the original control budget.
- Contract strategy decides who owns which pot, so owner contingency and contractor reserve need separate curves and separate reporting.
- Report the adequacy ratio, drawdown against plan, burn velocity and draw split every month, with escalation triggers agreed before they are needed.