An estimate is a measure of how well the scope is defined

In its 2024 annual report, Saudi Arabia's Public Investment Fund recorded a write-down of roughly 8bn dollars against its giga-projects. The portfolio's carrying value fell to SAR 211 billion from SAR 241 billion a year earlier, a reduction of more than 12%, which Reuters and CNBC reported in August 2025 as reflecting cost overruns, delays and changed market conditions.

A write-down is the moment an organisation formally accepts that a number it has been carrying no longer describes the thing it refers to. The interesting question is what the original number was ever claiming to be, rather than the size of the correction.

Most organisations answer that question wrongly, and the standards they cite in support answer it for them.

What an estimate class actually declares

The AACE International classification system is the one most commonly invoked in tender documents and board papers across the region. Its building and general construction guidance, Recommended Practice 56R-08, sets out five classes. The important sentence is the one nobody quotes: the maturity level of project definition is the sole determining characteristic of class. Everything else (the estimating method, the purpose, the accuracy range) is secondary, and merely correlates with it.

Expected accuracy range by estimate class for building and general construction. Class 5, at 0% to 2% definition, runs from minus 30% to plus 50%. Class 4, at 1% to 15% definition, minus 20% to plus thirty. Class 3, at ten to forty, minus fifteen to plus twenty. Class 2, at thirty to seventy-five, minus ten to plus fifteen. Class 1, at sixty-five to one hundred, minus five to plus ten.

Item	Scope definition	Expected accuracy range
Class 5	0 to 2%	-30 to +50%
Class 4	1 to 15%	-20 to +30%
Class 3	10 to 40%	-15 to +20%
Class 2	30 to 75%	-10 to +15%
Class 1	65 to 100%	-5 to +10%

The class is set by the second column, not the first. An estimate cannot be promoted by spending longer on it, only by defining more of the project. Ranges are quoted at an 80% confidence interval, after contingency.

Two features of that table repay attention. The first is the asymmetry. At every class, the high side is longer than the low side. The classification is stating in print that an estimate is more likely to be exceeded than undershot, and that this remains true even at Class 1.

The second is subtler. 56R-08 notes that although maturity is expressed as a percentage, it is the maturity of the defining deliverables that determines the class, not the percentage itself. "30% design" is not a class. A named set of drawings, specifications and basis-of-design documents, at a stated status, is a class. This is why an estimate cannot be promoted by asserting that design is further along than the deliverables show.

The small print that changes the answer

Three qualifications sit inside these documents and rarely survive the journey into the papers that cite them.

The first is scope. 56R-08 states that it does not address transport infrastructure, dams, reservoirs or tunnels, and that it does not cover one-of-a-kind projects, naming concert halls, sports stadiums, research buildings, health facilities, science laboratories and high-technology manufacturing. The process industries equivalent, 18R-97, excludes commercial building construction and transport infrastructure in turn. A large mixed-use development is substantially a portfolio of the excluded categories. The tidy accuracy range quoted against it frequently has no standing in the document it is drawn from.

The second is direction. AACE numbers its classes downwards, so Class 5 is the least defined and Class 1 the most. The American Society of Professional Estimators numbers its levels upwards, so Level 1 is the least defined and the numbers rise as definition improves. AACE's own comparison of classification practices places the two systems side by side and the reversal is plain. Two estimators can each report "a one" and mean opposite ends of the same scale.

The third is confidence. The 56R-08 ranges are quoted at an 80% confidence interval, after contingency. Other classification documents, including earlier revisions, use different bases. A range without its confidence interval is only a pair of numbers.

Contingency is not a safety margin

56R-08 describes contingency as typically set to achieve roughly equal probability of overrun and underrun. Read plainly, that means a properly contingent estimate is expected to be exceeded about half the time. This is the design intent of the method, not evidence that the method failed. An organisation that treats a P50 estimate as a ceiling has misunderstood what it commissioned.

The practical damage comes from layering. When a designer adds an allowance, the estimator adds contingency, the programme adds a reserve and the sponsor adds a further margin, each in ignorance of the others, the total corresponds to no stated confidence level at all. It is neither conservative nor optimistic. It is unmeasured, and it cannot be defended in a challenge because nobody can say what it represents.

There is a harder number underneath this. 56R-08 records that for weak project systems, and for complex or otherwise risky projects, the high ranges may run to two or three times the tabulated high range. A Class 5 estimate nominally topping out at +50% can, on that guidance, credibly reach +100% to +150%.

An estimate inherits every basis inside it

Alongside definition maturity, 56R-08 lists the systemic risks that drive accuracy. Among them are the quality of assumptions, market and pricing conditions, currency exchange, regulatory and community risk, and third parties, including utility owners.

That last one deserves separating out, because it is where good estimates are quietly ruined. The cost of diverting existing services is usually supplied by the utility owner. It arrives as a single figure, prepared on that organisation's own basis, priced at its own date, against its own understanding of what is being moved. Placed into the project estimate it looks like a priced line item. It is in fact a separate estimate, of a different class, at a different confidence level, possibly for a different scope. An estimate assembled from several such contributions has no single class, however carefully the rest of it was built.

The same document names something else that is rarely written down in project papers: organisational pressure for a predetermined value, and political risk and bias, as drivers of estimate accuracy. That is in the standard. It is not an editorial observation.

First cost is not the cost

TOTEX is often used loosely as a synonym for total cost of ownership. The two are related, and they are not the same thing.

TOTEX is a regulatory allowance framework, developed by economic regulators such as Ofwat and Ofgem, which combines capital and operating expenditure into a single allowance. Its purpose is to remove a distortion: where a regulated company earns a return on capital but merely recovers operating cost, it has a structural incentive to solve problems by building something. TOTEX is a remedy for a bias created by how spending is remunerated.

Whole-life cost, life cycle cost and total cost of ownership are analysis methods. They ask what it will cost to acquire, operate, maintain, renew and eventually dispose of an asset across its service life, discounted to a common basis. TOTEX corrects an incentive. Whole-life costing answers a question.

The distinction matters at inception because an estimate that stops at practical completion is answering a question nobody needs answered. 56R-08 carries a section on classification for long-term planning and asset life cycle cost estimates precisely because the two are different exercises. The commitments made in the first are inherited, without renegotiation, by whoever runs the asset for the following forty years, which is also where the cost of a weak handover, read: Handover is where the cost lands, and the Gulf is now taking deliveryHandover is where the cost lands, and the Gulf is now taking delivery and the condition of the asset register, read: Accrual conversion gave three years. It did not give anyone a register.Accrual conversion gave three years. It did not give anyone a register. come due.

Why a lifecycle cost ratio is unusable without its scope and its period

, read: A lifecycle ratio nobody can check, and a conclusion that does not need itA lifecycle ratio nobody can check, and a conclusion that does not need it

What actually narrows the range

Only better definition narrows the range. Everything else changes how the range is described.

Requirements have to be confirmed rather than assumed. AACE lists the quality of assumptions as a systemic driver of accuracy, and the useful discipline is procedural: an assumption that has been recorded, owned and dated is a managed risk, while an assumption that was never written down is a liability that surfaces as a claim.

Contract form has to be settled early, because it changes what the estimate must contain. Lump sum, remeasured and turnkey arrangements allocate quantity risk differently, and an estimate built for one is not valid for another. The AACE maturity matrices treat contracting strategy as a deliverable in its own right, moving from assumed at the earliest classes to defined at the latest.

Work packages have to align with the funding cycle. A package that straddles two budget years carries escalation and the risk of an approval gap, and an estimate that ignores the boundary understates both.

Finally, the basis of estimate has to be written. 56R-08 gives it a section of its own. An estimate separated from its basis cannot be checked, cannot be updated when scope moves, and cannot be defended when it is questioned years later by people who were not in the room.

The estimate range check applies the Table 1 ranges above to a figure of your own, for the class its basis declares, and shows where the approved number sits inside the band. It takes the class as given, because the RP says a percentage must not set it and the deliverables that do are set out in the RP's Table 3.

---

Sources. AACE International Recommended Practice 56R-08, Cost Estimate Classification System: As Applied in Engineering, Procurement, and Construction for the Building and General Construction Industries, revision of 7 August 2020, Tables 1 and 3 and sections 3, 7 and 9. AACE International Recommended Practice 18R-97, Cost Estimate Classification System: As Applied in Engineering, Procurement, and Construction for the Process Industries, 2 February 2005, Figures 1, 3a and 4. Public Investment Fund 2024 annual report, as reported by Reuters and CNBC, August 2025. Ofwat and Ofgem price control methodologies for the TOTEX framework. Where a publisher hosts a free extract, that is what the link points at.