“It’s a plus or minus ten percent number” is the sentence most sanction meetings turn on, and it is almost always the wrong sentence. The accuracy range is the least important thing an estimate class tells you. What it actually tells you is how much of the project has been defined, and therefore how much of the number is engineering and how much is assumption wearing an engineering hat.
The class is a definition statement
The public estimate classification practice, AACE International’s 18R-97 and its process-industry siblings, ranks estimates from Class 5 to Class 1 by the maturity of the project definition behind them. A Class 5 estimate sits on a few percent of definition: a capacity, a block flow, an analog. A Class 3 estimate sits on a defined scope: process flow diagrams issued, major equipment sized and listed, a plot plan, a piping and instrumentation basis, an execution plan. A Class 1 is a check estimate against bids.
The accuracy ranges published alongside each class are consequences of that definition level, and they are wide bands, not the tidy plus-or-minus most slides quote. The point of the class is not the band. The point is that you cannot buy a Class 3 number by hiring a better estimator. You buy it by doing the definition work, and the class is the receipt.
The ladder by gate
A working owner’s process ties the class to the gate, so that a gate is asking a question the estimate can actually answer:
- Feasibility gate: Class 5. Enough to decide whether to spend definition money, and nothing more.
- Definition gate: Class 4, or Class 3 on a small project where the definition is quick. Enough to release detailed-definition funding.
- Sanction gate: Class 3, or Class 2 on a major project where bids are in hand. This is the control estimate. From here on, every change is measured against it.
- As bids close: Class 1 check estimates, which test the control estimate rather than replace it.
Read the ladder from the other direction and the trouble shows up. A project sanctioned on a Class 4 number has been sanctioned on a definition level the process said was enough to fund detailed definition, not execution. Whatever the slide said about accuracy, the owner has committed execution money to a scope that is not yet defined to execution depth. That is not a risk to be managed. It is a decision to be made knowingly, or not made.
Contingency is not a percentage
The second habit the class exposes is the flat contingency. “Add ten percent” is a number chosen for being a round number. Contingency in a working process is set by risk analysis against the estimate class: the identified risks in the register, quantified, plus the systemic uncertainty that the class itself implies. A Class 3 estimate with a thin risk register and a ten percent contingency is telling you the register is thin, not that the project is safe.
The practical test is simple. Ask what the contingency is for. If the answer is a list of risks with a number against each and a residual for what the class cannot see, it is contingency. If the answer is a percentage, it is a rounding convention, and the sanction is being made on the base estimate with a comfort blanket on top.
The basis of estimate is the deliverable
None of this can be checked from the number. It is checked from the basis of estimate, the document that says what the estimate covers, what it assumes, what it excludes, what definition it rests on, and how the contingency was derived. An estimate without a basis is a number; an estimate with one is a claim that can be tested. At the sanction gate the basis of estimate is the deliverable, and the number is its summary.
Which is why a sanction readiness review spends more time on the basis than on the total. The total will be wrong by some amount; every total is. The question is whether the project knows what it does not know, has written it down, and has priced it. A Class 3 estimate is a commitment to having done that. Whether the project has actually done it is what the gate is for.
The standard behind this is FE-04, Estimate Classification and Basis of Estimate, one of the forty-seven documents in the Capital Project Management System. It is public-framework work end to end, which means your estimators can check it against the practice they already cite.