Contingency and management reserve are two funds admitting different classes of event under different release authorities, and an enterprise that holds one undifferentiated pot has silently merged a delivery decision with a governance one.
A project director on a network refurbishment opens a switchgear bay and finds the busbar assembly in worse condition than the survey suggested. Replacing it costs rather more than a week of programme spend. She emails the finance controller: draw it from the reserve. He agrees. The exchange takes four minutes, appears in no minute book, and is exactly what a reserve exists for.
Two questions went unasked. Which reserve — money held inside the approved cost baseline against risks the team already named, or money held outside it against events nobody forecast? And on whose authority — the delivery leadership who own the schedule, or the body that owns the mandate? In most enterprises neither has an answer, because there is one pot, and whoever reaches it first defines what it was for.
The consequence arrives not with the first claim but with the fourth. By then the fund meant to stand behind an event nobody could have anticipated has been consumed by events that were anticipated, entered on a register, and simply under-priced. The enterprise still reports a reserve balance. What it no longer holds is the capacity to absorb a surprise.
Reserves are the only pre-authorised capacity an organisation has to act without going back for permission; every other pound is committed. The rules on who may release reserve, and against what class of event, are therefore the operative risk appetite — not the version in the annual report, but the one that decides what happens on a Tuesday afternoon.
The Strategic Context
The cost architecture most large organisations inherit has a specific shape. Estimates roll up through work packages and control accounts into a project estimate. Contingency is added, producing the cost baseline — the figure against which performance is measured. Management reserve is added on top, producing total authorised funding. Contingency sits inside the measurement baseline; management reserve sits outside it.
That geometry is not an accounting convenience. It encodes a delegation. Money inside the baseline is already authorised against work already described. Money outside it is authorised for nothing; it enters only through a change transaction, and that transaction is where someone other than the delivery organisation gets a say.
The earned-value tradition is explicit about the second fund and silent about the first. It requires management reserve to be set aside, its use documented, its release into the baseline routed through change control, every movement logged — and in stricter public-sector formulations forbids the customer from confiscating it or the contractor from using it to absorb contract variations. This is a governance instrument in governance language.
Contingency receives no comparable treatment. It is defined as a reserve for expected variability, sized by an allowance that narrows as the estimate matures, and handed to the delivery manager. Who may release it, against what evidence, and answerable to whom, the surrounding material never says.
Nor is the boundary between the funds stable inside the discipline's own teaching. One strand sets contingency against known unknowns and reserves management reserve for the genuinely unforeseen. Another defines management reserve as covering unforeseen work and known risks that may or may not occur — precisely the class the first assigns to contingency. That overlap is the admission question itself, unresolved, in material presenting itself as settled.
What Leaders Commonly Misread About Held Money
The first misreading is that a reserve is a number. It is a permission with a number attached. Two funds of equal size and different release rules are two instruments; two funds of different size and identical rules are one instrument in two columns.
The second is that the important question is how much. Boards spend considerable time on sizing and almost none on admission. An undersized fund with a clean admission rule fails visibly and early, which triggers escalation. A generous fund with no admission rule fails late and invisibly.
The third concerns how contingency is calculated. The common method multiplies each register entry's probability by its cost impact and sums the results, producing the expected cost of the register — the average across many repetitions of the same portfolio. A portfolio holder running fifty projects lands near that average. A single project runs once; it will meet some subset of its risks and none of the others, and the sum of expected values describes an outcome it will almost certainly not have. That is not conservatism or aggression. It is a category error about what the number means.
The same arithmetic carries a second defect. Where a risk has been treated — a second supplier qualified, an inspection brought forward, a spare held — the treatment carries its own cost, sitting in the baseline as scheduled work. If the contingency sum is still computed from the untreated probability, the enterprise has funded one event twice. The taxonomy of treatments belongs to Article 39 in this collection, the width of the estimate band to Article 37; neither is re-argued here. The point for reserves is sharper: a budgeted treatment and a funded contingency for one event are two payments, and nothing in the cost architecture detects it.
Reframing the Issue
Stop treating reserves as money and start treating them as decision rights.
A reserve is a pre-agreed statement that a named party may commit a defined quantity of the enterprise's resources, without fresh authorisation, when a defined class of event occurs. That is the whole instrument. The balance is only how far the right extends.
Read that way, the funds separate for an obvious reason. Contingency is a delivery decision right: it lets those running the work absorb ordinary variability without stopping to ask, because stopping to ask costs more than the variance. Management reserve is a governance decision right: it preserves the authorising body's capacity to respond to something that changes the undertaking, and preserves it precisely by not delegating it.
Consider a hypothetical electricity transmission and distribution network operator refurbishing an ageing substation fleet. Its contingency should cover the routine and unwelcome: condition worse than survey, an outage window lost to weather, a defect found at commissioning. The enterprise anticipated the type though not the instance, and the delivery director should be able to authorise all of them — a board paper for every corroded busbar would cost more than the risk.
Its management reserve should cover something categorically different: a regulatory determination that changes the required end state, a supplier failure that removes a technology from the market, a systemic defect in a component installed at forty sites. These do not consume the plan; they invalidate part of it. Releasing money against them is re-authorisation, and belongs to whoever authorised the work.
One pot cannot do both jobs, because the jobs require opposite properties. Delivery money must be fast and delegated; governance money slow and reserved. Merge them and speed wins — it always does, because the delivery claim is concrete, urgent and in the room, and the governance claim is abstract, future and represented by nobody.
What Each Fund Is Permitted to Admit
The admission class
Every fund needs a stated class of admissible event and, more importantly, of inadmissible event. The exclusions do the real work. A contingency barred from funding scope additions, estimating corrections or price escalation is a contingency; one with no exclusions is unallocated budget with a respectable name.
That discipline appears in the earned-value tradition and almost nowhere else: management reserve is not to absorb the cost of contract changes. A change has a counterparty who ought to pay, and funding it quietly from reserve destroys the commercial record and the baseline together.
Who may release, and against what evidence
Authority without an evidence rule is discretion. The workable form is a triple: a named releasing officer, a named class of event, and a named artefact that must exist before release. For contingency that artefact is a registered risk whose trigger has demonstrably fired. For management reserve it is a change transaction moving scope and budget together.
That rule also exposes events fitting neither fund. In a hypothetical hospital and clinical services operator, a service relocation might find a ward's medical gas manifold below the standard the new model assumes. That is neither ordinary variability nor a change of mandate; it is a failure of specification, and reserve is the wrong conversation about it. What governs it is the acceptance standard nobody wrote down. This article does not address how an enterprise discovers which benchmarks it has silently authored for itself; that is carried by [Related article: Where the Standard Is Silent, the Benchmark Becomes Yours].
What the register cannot hand to either fund
Both funds are sized from the risk register, so both inherit its limits. The most consequential sits at the bottom of the probability scale: beneath the lowest band any common instrument offers, events cannot be recorded, and an unrecordable event cannot be reserved against. This article takes the register as given and asks only which fund may pay for what it contains; what the instrument cannot record is the subject of [Related article: The Floor Beneath Your Risk Scale].
A further class of decision removes options rather than adding costs. An enterprise committed to zero tolerance of an outcome has abolished a family of responses — transfer and acceptance among them — and the survivors must be funded. This article does not price that abolition; that is the work of [Related article: A Zero-Tolerance Commitment Is a Capital Decision].
Decision Framework
The two-fund test. Apply it to every reserve balance the enterprise holds, before the money is needed. Each fund must return a written answer to all five questions.
| Question | Contingency answers | Management reserve answers |
|---|---|---|
| 1. Admission — what class of event may it fund? | Identified risks of an anticipated type | Events altering the plan's premise or scope |
| 2. Exclusion — what may it never absorb? | Scope additions, estimating corrections, escalation | Contract changes with a counterparty |
| 3. Authority — who releases it? | Named delivery officer, to a stated ceiling | The authorising body, via change control |
| 4. Evidence — what must exist first? | A registered risk whose trigger has fired | An approved change, scope and budget together |
| 5. Replenishment — who restores it, and from where? | The authorising body, on re-assessed exposure | The funding authority, on re-authorisation |
Three failure conditions apply. The same answers: two funds answering questions 1 and 3 identically are one fund with two names — collapse the reporting and stop claiming two lines of defence. The open admission: a fund answering "whatever comes up" to question 1, or nothing to question 2, is unallocated budget and should be reported as such. The misplaced authority: where a fund admitting changes of scope, mandate or end state names a delivery officer at question 3, a governance decision has been delegated to a delivery role; the correction is to move question 3, not the money.
A fourth check runs across funds. Take the ten largest treated risks and confirm each treatment cost sits in the baseline and each provision has been recomputed on post-treatment exposure. Where both are funded in full, the difference is double-counted and should be released.
From Strategy to Execution
Immediate — within the current reporting cycle. Run the two-fund test on the three largest reserve balances. Resize nothing. Produce the five written answers and the officer named at question 3; where a question cannot be answered, that is the finding. Then classify every drawdown of the last twelve months against the admission classes now written. The unclassifiable proportion is the size of the problem.
Medium-term — the next two planning cycles. Split any undifferentiated pot into two funds with distinct release rules, visible in reporting: baseline, contingency inside it, management reserve outside it, each with its own movement log. Introduce the evidence rule before the ceiling rule — a required trigger changes behaviour faster than a lowered limit. Recompute provisions on post-treatment exposure and retire the duplicated funding.
Long-term — the next authorisation cycle. Move reserve rules out of project documentation into the enterprise's delegation schedule, alongside every other decision right the organisation grants. The question then stops being one of project controls and becomes what it always was: which decisions the enterprise is willing to have made without it.
Signals to Monitor
Reserve consumption running ahead of physical progress, indicating a fund absorbing estimating error rather than risk. Drawdowns that cannot be matched to a registered risk with a fired trigger. Management reserve movements unaccompanied by a change transaction. A movement log with gaps, or none. Approvals given verbally and confirmed after the spend. Any period in which the reserve balance falls while assessed exposure does not — the signature of a fund paying for the wrong class of event.
Questions for the Leadership Team
- For each reserve balance we hold, what class of event may it fund, what may it never absorb, and where is that written?
- Which named individual may release each fund, within what ceiling, and to whom do they account if the release proves wrong?
- Of the reserve drawn down last year, what proportion matches a risk that was on the register before the money was spent?
- Where a risk carries both a budgeted treatment and a contingency provision, has the provision been recomputed on post-treatment exposure, or are we funding the event twice?
- Has any reserve movement in the last two years absorbed the cost of a change a counterparty could have been billed for?
- If our largest programme exhausted its contingency tomorrow, who would decide whether to replenish it, from what source, and how long would it take?
Closing Perspective
The four-minute email is the point. Reserves fail not through fraud but through a sequence of individually reasonable approvals, each the fastest way to keep the work moving, none of which anyone was positioned to refuse. That is not a control failure. It is what a system does when the right to spend its capacity for surprise was never allocated, and fell to whoever stood closest to the problem.
An enterprise settles this once, deliberately, in a delegation schedule — or repeatedly, by accident, at the worst possible moments. The choice is not whether to hold reserve. It is whether the organisation knows, before the bay is opened, which of its two funds is answering and who holds the right to say no.
About EraNorth Insights
EraNorth Insights publishes practical analysis on strategy, projects, operations, transformation and decision intelligence for professional and organisational use. About EraNorth.