Owners often hear “Class 3” and assume it means a more careful version of the Class 5 they already have. It does not. The classes describe how mature the project definition is, not how hard the estimator worked. A Class 3 estimate built on Class 5 information is still a Class 5 estimate, with false precision added.
The distinction matters because each class exists to support a different decision. A Class 5 tells an owner whether an idea is worth the money it takes to study it. A Class 3 asks a board to commit real capital. When the number that reaches the board was built on screening-level information, the project starts life with a budget it never had a basis for, and every trend report afterwards becomes an argument about whose fault that is.
Where the classes come from
AACE International Recommended Practice 17R-97, Cost Estimate Classification System (RP 17R-97), establishes the five classes, and RP 18R-97 applies them to the process industries. The principle in both is the same: the class is assigned from the maturity of the project definition deliverables in hand, not from the time spent, the software used, or the confidence of the person presenting it. Percent design complete is the primary driver. The widely published headline accuracy ranges run from roughly minus 50 to plus 100 percent at Class 5 and tighten as definition matures; the RPs give the full context, and the ranges vary by industry and by project. The full matrix, including the maturity requirements per class, is in RP 17R-97 and its industry variants, available to AACE members at web.aacei.org.
The practical reading is simple. Look at what the design team can hand you. If the deliverables are a block flow diagram and a capacity figure, you can produce a Class 5, whatever the deadline says. If they include P&IDs, plot plans and an equipment list with data sheets, a Class 3 becomes possible. The inputs set the ceiling. Nothing the estimator does can raise it.
What moves between Class 5 and Class 3
Between the two classes, the share of the estimate priced from measured quantities climbs from almost nothing to the majority of the direct cost, and each major account changes its nature on the way.
- Equipment moves from capacity factoring against analogous plants to budget quotations on the major items, with scope splits confirmed in writing.
- Bulks move from ratios against equipment cost to material take-offs from P&IDs and layout drawings.
- Labour moves from gross hours per unit of capacity to norms applied commodity by commodity, adjusted for site productivity.
- Indirects move from a flat percentage to a staffed organisation chart priced against a real schedule.
- Contingency moves from judgement to a quantified risk analysis with a register behind it.
None of these transitions is free. Each one needs design information that does not exist at concept stage, which is the whole point: the classes are a map of what the design can support, not a menu the schedule picks from.
The same scope, two honest numbers
Take a mid-size process unit, and call the eventual cost sixty million dollars so the arithmetic is easy to follow. At concept stage, the Class 5 is capacity factored from two analogous plants, adjusted for location and pricing year, and cross-checked against a published benchmark. It lands as a range, say forty to ninety-five million, with a short list of the assumptions that drive it: the capacity basis, the location factor, the exclusions. That range is wide, and it is supposed to be. Its job is to answer one question: is this a sixty million dollar idea or a two hundred million dollar idea?
Eighteen months later, at the end of FEED, the Class 3 for the same scope prices twelve major equipment items from budget quotations, takes bulks off the issued-for-design drawings, builds labour from norms against those quantities, and prices indirects from a twenty-month schedule with a staffed organisation chart. The risk register carries twenty-seven items, and the contingency that comes out of the analysis is nine percent, stated separately. The range narrows to roughly fifty-six to seventy million.
Notice what happened. The midpoint barely moved. What changed is the honesty of the range and the traceability behind it. Both numbers were right for their stage. The damage only starts when someone presents the first one with the second one’s confidence.
What should not move
The basis of estimate. The scope statement, the battery limits, the pricing date and the exclusions are recorded in a Basis of Estimate, the discipline set out in RP 34R-05. If those change between classes without being written down, the two numbers cannot be compared and the trend between them has no meaning. Every reconciliation we produce starts by lining up the two bases before a single dollar is discussed.
A Class 3 built on Class 5 information is a Class 5 with a nicer cover page.
Where this goes wrong on real projects
The label outruns the inputs. The stage-gate schedule says a Class 3 is due, the design deliverables are still at screening maturity, and the estimate gets the label anyway because the gate will not wait. The board approves a number with a plus or minus it does not actually have, and the first honest estimate of the project is delivered by the market, at tender, in public.
The middle class gets manufactured. A “Class 4” appears that is the old Class 5 with a tighter contingency and no new quantities. Nothing about the definition matured; only the presentation did. Ask what new deliverables arrived since the last estimate. If the answer is none, the class did not change.
The range gets forgotten. A Class 5 presented as a range becomes a single midpoint the moment it lands in a slide, and that midpoint gets quoted back at the team for the next four years. Present the range as the answer, name which end is more likely and why, and put the midpoint in the supporting text rather than the headline.
What to ask for
Two questions expose the real class of any estimate, whatever the cover page claims. First, what percentage of the direct cost is priced from measured quantities? Second, how were the indirects built, from a percentage or from a schedule? The answers tell you more than the label, because they cannot be adjusted without doing the work.
The class system only works when everyone reads it the same way. The estimate is not a prediction that gets graded later. It is a statement of what the definition supported on the day it was priced, and the class is the honest label on that statement.
Emerald Group prepares and independently reviews estimates from Class 5 through Class 3 for owners and engineering firms. If a number is about to carry a gate decision and you want it checked against the basis it claims, get in touch.