A construction cost range forecast an investor should ask for is not a percentage band around a total. It is a list of named exposures, each with a low and a high, added up. The first is decoration. The second can be checked.
Why the point estimate keeps being wrong
A project is reported at $64.2 million. It finishes at $67.9 million. Nobody lied and no single decision caused it.
The total is a sum of hundreds of items, most of them settled and a few of them not. The settled ones do not move. The unsettled ones move a lot, and they move in one direction more often than the other, because an allowance is set at a figure somebody thought was reasonable and the realistic outcomes above it are less constrained than the outcomes below it.
A point estimate hides that structure. It presents a number built from a mix of certainties and guesses as though all of it had the same standing.
The conventional fix is to add a contingency percentage, which is a range in disguise and a poor one. Four percent of the total is spread evenly across everything, including the parts that cannot move, and no part of it is attached to a cause anybody can name.
What you want instead is an explicit range whose width comes from the items that are genuinely open.
How a construction cost range forecast an investor can check is built
From the bottom, not from the top. Five categories, each producing a low and a high.
Unresolved allowances. For each, the carried figure and a realistic landing range based on the current specification and current pricing. This is usually the largest single contributor and the most tractable, because every item is named in a schedule.
Uncapped exclusions. For each, the probability it is encountered and the cost if it is. An exclusion with a quantity cap contributes a bounded amount. One without contributes a much wider band.
Unbought packages. The dollar value of work still priced from an estimate, and a reasonable band for market movement between the estimate and the award.
Outstanding owner decisions. Every decision not yet made is a cost not yet fixed, and late decisions price worse than early ones.
Expected change volume. Based on the pattern already established on this job rather than on an industry rule, which means it improves as the job proceeds.
Add the lows, add the highs. The result is a range with a reason behind every dollar of its width.
Why a bottom up range behaves better
Three properties, and all three matter to somebody holding capital.
It is checkable. Anybody can open the allowance schedule and disagree with a specific line. Nobody can disagree with four percent, which is why four percent survives scrutiny it does not deserve.
It narrows over time in a way you can watch. As allowances resolve and packages are bought, items drop out of the list and the band tightens. A percentage does not narrow, it just gets applied to a smaller remaining balance, which tells you nothing.
It tells you where to spend attention. If two items account for 60 percent of the width, those two items are the project risk and everything else is noise. That is an actionable statement in a way that a contingency percentage never is.
There is a fourth property that matters in a committee. A range built this way survives being asked where it came from, which is the question that ends most presentations built on a percentage.
The same discipline applies to the allowance schedule itself, which is where most of the width originates and where it is most often understated. That is covered in why an allowance is not a budget line.
What the range should not do
It should not be symmetric. Cost outcomes on construction projects are not normally distributed around the estimate, and a band of plus or minus five percent is describing a world that does not exist. The downside is bounded by what the work costs. The upside is bounded by nothing in particular.
It should not be presented without the drivers. A range with no named items is a percentage wearing different clothes, and it invites exactly the same unanswerable question.
It should not be confused with contingency. The range is the demand side, which is what the project might need. Contingency is the supply side, which is what is available. Presenting them together is the whole point, and presenting either alone is why most cost reporting fails to inform a decision.
It should not be static. A range produced at signature and never updated is worse than no range, because it carries an authority it has stopped earning. Rebuild it when the allowance schedule moves materially, which on most jobs is two or three times.
It should also not be rounded into comfort. A band of $1.6 million to $4.2 million says something a band of $2 million to $4 million does not, which is that somebody added up specific items rather than picking edges that looked sensible. Precision that comes from arithmetic is worth keeping even when it looks odd on a slide.
Finally, it should not carry a probability it cannot support. Labeling the high end a worst case implies a distribution nobody has modeled. The honest description is a reasonable high given the items currently open, which is a smaller claim and a defensible one.
A worked example
Illustrative figures. Not taken from any client project and not a quotation.
A $59 million guaranteed maximum price on a data center at 45 percent complete. The report shows cost at budget and contingency remaining of $2.4 million.
Built from the bottom: fourteen unresolved allowances totaling $6.2 million carried, with a realistic range of $6.6 million to $8.1 million, contributing $400,000 to $1.9 million. Two uncapped exclusions contributing $0 to $1.1 million. Unbought packages of $9 million contributing $270,000 to $720,000 at reasonable market movement. Outstanding owner decisions on electrical capacity contributing $500,000 to $1.4 million. Established change volume contributing $430,000 to $580,000.
Total open exposure $1.6 million to $4.2 million against a fund of $2.4 million.
The report said the project was at budget with comfortable contingency. The range says the fund covers the middle of the distribution and not the top of it, and that two items, the allowances and the electrical capacity decision, account for most of the width. One of those two is a decision the owner controls and has not yet made.
What to ask for
- A range rather than a point estimate, with the low and high stated separately.
- The five contributors listed, each with its own low and high.
- The three largest contributors to the width named, with the driver behind each.
- The range shown against the available contingency, not instead of it.
- An asymmetric band, because the downside and upside are not symmetric in practice.
- A rebuild whenever the allowance schedule or the buyout position moves materially.
- Confirmation that whoever built it has no position in the outcome.
Item three is the one that changes behavior. Once two items account for most of the width, the project has a short list, and short lists get acted on in a way that percentages never do. The review packages describe how the range is produced and what it is built from.
What we do
We build the range from the allowance schedule, the exclusions, the buyout position and the established change pattern, and we show the arithmetic. Every contributor is named, every low and high has a basis, and the whole thing is set against the available fund so the comparison is on one page. We do not produce a competing estimate of the work. The reading behind it is the readiness review.
Questions people ask
Is a range just a contingency percentage presented differently?
No, and the difference is checkability. A percentage is applied to a total and has no cause behind it. A range is built from named items, each of which somebody can open the schedule and argue with. One survives a committee asking where it came from and the other does not.
How wide should the range be?
As wide as the open items make it, which is the point. A project with fourteen unresolved allowances and two uncapped exclusions has a wide band, and narrowing it for presentation removes the information. The band should tighten because items resolve, not because somebody preferred a tidier number.
Who should build the range?
Somebody with no position in the outcome. A contractor building it will lean low because the ceiling is theirs to defend. A sponsor building it has an audience to reassure. The arithmetic is not difficult, which is precisely why the independence of the person doing it is the part that matters.
This is general information about construction contracts and is not legal advice.