Cost estimating

AACE 18R-97 vs CIQS Class A–D: How the Two Estimate-Class Systems Map

I hold cost credentials in both worlds — CCP from AACE International, PQS from the CIQS. On a single infrastructure program I will meet both classification systems in the same week, and I have watched more than one boardroom assume a "Class 3" and a "Class C" are the same estimate. They are cousins, not twins.

By Sharib Maradukhel · MRICS · MCIOB · PQS · CCP · Director, Electrical Cost Management

Two estimate-classification systems dominate the capital work I do. On industrial, energy and increasingly on major infrastructure programs, the language is AACE International — the five-class system defined in Recommended Practice 18R-97. On Canadian public buildings and architectural work, it is the Class A to Class D framework of the Canadian Institute of Quantity Surveyors, mirrored in the estimate definitions used by Public Services and Procurement Canada.

I carry credentials in both — CCP from AACE, PQS and CEC from the CIQS — and the most common mistake I see is not a wrong number. It is a category error: treating the two as one scale with different labels, and translating a number straight across to a letter. They line up, but not the way most people assume.

They count in opposite directions

Start with the source of most of the confusion: the two systems run in reverse of each other.

AACE counts down as an estimate improves. Class 5 is the roughest — a concept screen off almost no definition — and Class 1 is the most definitive, a check estimate against near-complete documentation. The CIQS runs up the alphabet: Class D is the order-of-magnitude concept estimate, and Class A is the tender-ready one.

So the tightest estimate in each system — AACE Class 1 and CIQS Class A — do describe the same thing. But the numbering is inverted, and the instinct to read "Class 1" as the first, roughest step is exactly backwards. More than one schedule has been built on that misread.

They are not measuring the same thing

The deeper point is that the two systems classify on different axes.

AACE 18R-97 classifies an estimate by the maturity of project definition — how complete the deliverables and input data are, expressed as a percentage from 0 to 100 — and by the estimating method that maturity supports. Crucially, the class is a property of the estimate, not the project. A quick screen of a fully designed building is still a Class 5 estimate if the inputs used were order-of-magnitude.

The CIQS framework, as applied through PSPC, ties each class to specific design deliverables and to a procurement or approval gate. A Class D is prepared from a functional program; a Class B from design-development drawings and outline specifications, in measured elemental format, sufficient to obtain project approval; a Class A from completed tender documents. It is a deliverables-and-gate system, expressed in elemental cost analysis.

Put simply: AACE describes the inputs and the method; the CIQS describes the deliverables and the decision they support. Most of the time they move together — better deliverables permit better methods — which is why they map at all. But they answer different questions, and on a program that runs both, that difference is exactly where reconciliations go wrong.

The practitioner's mapping

Here is how they line up in practice. Treat it as approximate: AACE has five classes and the CIQS has four, so the concept end is not one-to-one — Class D absorbs both of AACE's early classes.

The pairing that matters most on a funding decision is the middle row. An AACE Class 3 — the control-budget estimate that anchors most final investment decisions — corresponds to a CIQS Class C. Not Class B, and certainly not its numeric twin. When a process-side team hands over a "Class 3" and a quantity surveyor hands over a "Class C," they are describing the same maturity. When someone instead lines up "Class 3" against "Class B" because the digits and letters feel adjacent, the contingency conversation starts from the wrong place.

They even express accuracy differently

The two systems don't just label maturity differently — they state its consequences differently too.

The CIQS and PSPC tradition frames accuracy as a design allowance: a ceiling on the contingency the class should carry. A Class D should need no more than roughly a 20% design allowance; a Class B, no more than about 10%. It is a forward-looking instruction — carry this much, because this is how incomplete the design still is.

AACE frames it as an expected accuracy range: a low-and-high percentage band, quoted at roughly 80% confidence, that widens sharply at the concept end — a Class 5 can run from around −30% to +50%, and considerably wider on weak or complex scope. It is a backward-looking prediction — this is the band your final cost is likely to land in.

But both traditions end at the same, non-negotiable place, and it is worth stating plainly: the class does not set the accuracy — the risk analysis does. AACE says so explicitly; a class is a starting expectation, and the real range for a specific project comes from a quantitative risk assessment. A CIQS design allowance is a default, not a substitute for pricing the named risks on the job in front of you. Any estimate whose stated range is simply copied from a table — on either system — is claiming a confidence it has not earned.

Why this matters on a real program

On the rail, data-center and healthcare programs I work on, both systems turn up in the same room. The EPC or process side speaks AACE; the cost-consulting side speaks CIQS; the owner's board speaks neither fluently and assumes the numbers and letters are interchangeable. The task for anyone bridging the two is to translate the maturity and the contingency philosophy — what deliverables exist, what method they support, and how much of the answer is still risk — rather than the label.

Do that, and the systems reinforce each other: AACE's discipline about inputs and methodology and the CIQS's discipline about deliverables and elemental structure are complementary strengths. Translate the label mechanically, and you hand a board a control budget that reads tighter than it is.