Setting — and defending — your contingency

Ask almost anyone building a house what their contingency is, and the answer comes back instantly: ten percent. Ask why ten percent, specifically, rather than seven or fifteen, and the honest answer is usually that it's what everyone says. It's not wrong as a starting point — but it's a convention borrowed from nowhere in particular, applied to a project whose actual risk profile was never looked at, and that's the gap that tends to show up at exactly the wrong moment.

A contingency figure that was never actually calculated has two problems. It might be too small for a project with real, specific exposure — a tricky plot, an unsurveyed existing structure, a first-time self-builder managing trades directly. Or it might be comfortably oversized for a straightforward new build on a clean, well-surveyed plot, which isn't a problem in itself except that the extra margin tends to get treated as spare budget rather than a reserve, and spent accordingly.

Where the ten percent figure actually comes from

It's not arbitrary, exactly — it's a reasonable average across a huge range of projects, which is precisely why it's a poor fit for any single one of them. A simple, well-documented renovation of a structurally sound house carries meaningfully less uncertainty than a new build on a plot where the ground hasn't been investigated, or a renovation where the extent of what's hidden behind the plaster is genuinely unknown until it's opened up. Using the same percentage for both treats very different levels of risk as if they were identical, because the number was never actually derived from either project's specific risks — it's inherited from habit.

Sizing contingency against the risks you actually have

A better starting point is the risk register itself (see last week's article on building one) — if the project's genuine risks have been named, scored and costed individually, contingency stops being a single percentage and becomes the sum of what those specific risks would actually cost if they materialised, weighted by how likely each one is. A plot with unknown ground conditions and no trial pits dug yet deserves a meaningfully larger allowance against groundworks and foundations than one where investigation has already ruled out the worst cases. An unsurveyed existing structure earns a larger allowance for the renovation elements than a new build with no hidden conditions to discover at all.

This doesn't have to be elaborate. Even a rough pass — walking down the risk register and asking "what would this actually cost if it happened, and how likely is that" — produces a figure that's grounded in the specific project rather than copied from a rule of thumb, and it's usually a more persuasive number to hold the line on later, because it can be explained rather than just asserted.

What eats contingency first, almost always

Across the projects I look at, the pattern is remarkably consistent. Contingency gets spent in a fairly predictable order, and it's worth knowing that order before it starts happening to you:

• Groundworks and foundations, almost always first and often by a wide margin — ground conditions that differ from what was assumed, drainage runs that need rerouting, a retaining structure that wasn't fully priced
• Services connections — a grid connection quote that comes in higher than budgeted, a water or sewer connection that needs more excavation or a pumped solution nobody priced for
• Structural surprises in a renovation — a lintel that turns out to be inadequate, timber that's worse than the survey suggested, a wall that isn't doing what the drawings assumed
• Specification creep on finishes, which rarely looks like overspend in the moment — it looks like a long sequence of individually reasonable upgrades
• Professional fees for additional surveys or redesign work triggered by something found on site

Scope creep is budget creep in a different outfit

The specification creep item above deserves its own attention, because it's the one that doesn't announce itself. Nobody sits down and decides to blow the contingency on finishes. It happens as a sequence of small, reasonable-sounding upgrades — a better worktop, uprated windows, a slightly larger rooflight — each decided in isolation, each genuinely small, and each drawn from the same pool of money as the groundworks risk it was actually meant to cover. By the time the ground conditions problem shows up, the contingency that was meant to absorb it has already quietly gone on kitchen upgrades.

The fix isn't refusing every upgrade — some of them are worth having. It's tracking spend against contingency explicitly, as its own running total, rather than letting it blur into the general project account. A spreadsheet that shows contingency draining against named items, updated as each one is agreed, makes the trade-off visible at the point it's made, rather than only visible in hindsight when the ground conditions bill arrives and the reserve meant to cover it isn't there any more.

Defending the line once it's set

Holding a contingency line is mostly a matter of process rather than willpower. Every request to draw on it — whether it's an unforeseen ground condition or a desired upgrade — goes through the same short discipline: what is it, what does it cost, is it a genuine unforeseen risk materialising or a scope addition, and does the remaining contingency still cover the risks still outstanding on the register. A request that can't answer those questions cleanly is usually the one worth pausing on, not because the upgrade is unreasonable, but because deciding it without reference to what's left in reserve is how the reserve disappears.

What a contingency plan that holds up actually contains

• A figure derived from the project's own risk register, not a flat percentage applied without thought
• A running total tracked separately from general project spend, so drawdowns are visible as they happen
• A named owner — usually whoever holds the budget — who approves every draw against it, however small
• A clear test for each request: genuine unforeseen risk, or scope addition masquerading as one
• A standing check, at each stage payment or site visit, of what's left against what's still outstanding on the risk register
• A separate conversation, early, about what happens if contingency runs out before the project does — because "it won't" isn't a plan