No assessment sees everything. Foundations are buried. Structural timbers sit behind ceilings. Concrete hides its reinforcement. Cladding hides its fixings. Records go missing. In a desktop read, most of the building is unseen.
There are two common bad responses. One is silence: leave out what cannot be confirmed, and present a tidy number that understates the risk. The other is false precision: give the unknown a single confident figure as if it were known. Both mislead.
Four tools
A capital plan has four tools for carrying uncertainty honestly. An investigation allowance pays for finding out: a roof survey, a structural assessment, a hazardous materials survey. A range shows the spread of plausible outcomes rather than a single point. A conditional allowance carries works that will only be needed if the investigation finds a problem, kept distinct from anticipated expenditure. And a confidence rating tells the reader how firm each line is.
Used together, they let an owner see the likely capital position and, separately, the size of what is not yet known.
How wide a range should be
Cost engineers have a long-standing way of saying how wide an estimate’s range should be: tie it to how much is known. AACE International’s cost estimate classification sets out the expected accuracy of an estimate by the maturity of the information behind it.
| Class | Information available | Typical low range | Typical high range |
|---|---|---|---|
| Class 5 | Concept only, very little definition | −20% to −50% | +30% to +100% |
| Class 4 | Feasibility; limited definition | −15% to −30% | +20% to +50% |
| Class 3 | Budget authorisation; meaningful definition | −10% to −20% | +10% to +30% |
| Class 2 | Control estimate; substantial definition | −5% to −15% | +5% to +20% |
| Class 1 | Bid or tender check; full definition | −3% to −10% | +3% to +15% |
AACE International Recommended Practice 18R-97, cost estimate classification. Ranges are typical, not limits.
Two lessons carry straight into a capital plan. A range that is too narrow for the evidence is not more accurate, only more confident. And the ranges are asymmetric: when less is known, the upside risk is larger than the downside. That is why a register built from early evidence carries wide, upward-skewed ranges on the lines that matter, and why investigation is often the cheapest way to narrow them.
The heritage example
The fire station in The Economics of Heritage shows the pattern clearly. A significant share of near-term spend was investigation rather than remediation: foundations, roof structure, fire compliance. Beside it sat explicit worst-case ranges for the risks those investigations would resolve. The building was not presented as cheaper or more expensive than it was. It was presented as partly unknown, with a plan to find out.
The cost of an investigation is usually small beside the cost of being wrong. Pricing it explicitly is how a plan earns trust.
Keeping the base honest
Conditional allowances need care. If every possible problem is added to the base programme, the plan becomes a worst case and loses credibility. If none is shown, the plan becomes a best case. The register keeps them separate: anticipated works in the base, conditional works beside it, each with its trigger. This is the same discipline as Timing Is Not Risk: checks can sit in the near term while the works they may lead to sit later.
A worked uncertainty
A 1960s office has precast facade panels whose fixings cannot be seen. The register carries three lines. An investigation item in the first year: inspection of a sample of fixings from access equipment, with a modest cost. A base-programme item in years two to five: cleaning, sealant renewal and minor concrete repair, priced with a range. And a conditional allowance beside the programme, not inside it: remedial fixing works if the inspection finds corrosion, priced as a wide range with its trigger stated.
The owner sees the likely programme, the size of the unknown and the small cost of resolving it. After the inspection, the conditional line either disappears or moves into the base with a narrower range. Either way, the plan stays honest.
Uncertainty in the rating
Uncertainty also belongs in the building’s overall position. In the MR Index, an element that needs a check, or a hazard suspected but not yet seen, is carried as an open question until it is resolved. A building with many open questions reads differently from one whose questions have been answered, and resolving them moves the band in whichever direction the evidence points.
What an owner gets from this is a clear list of what to investigate first. Resolving the largest uncertainties early usually narrows the range of the whole programme and makes every later decision easier.
Questions
What is the difference between a provisional sum and a contingency?
Terms vary by market, but generally a provisional or conditional allowance is tied to a specific identified uncertainty, while a contingency is a general margin across a project. A capital plan is clearer when specific uncertainties are named.
Why not include every possible defect in the capital plan?
Adding every possible problem to the base turns the plan into a worst case and undermines its credibility. Keeping conditional works separate, with their triggers, shows the likely position and the risk around it.