Anatomy of a Class 3 Cost Estimate: Basis, Contingency, and What Sanction Reviewers Check

Introduction

Every project reaches the meeting where someone asks: "how much, and how sure are we?" The estimate that answers has a class, and the class is a claim — about the maturity of the engineering underneath it, the methods used to price it, and the range the final number should be trusted within.

AACE International's 18R-97 gives the classification for the process industries: Class 5 at concept screening down to Class 1 at full definition. The workhorse of project sanction is Class 3 — typically built on 10–40% engineering definition, carrying an expected accuracy in the region of −10/−20% to +10/+30%, and intended to support the final investment decision and project budget. This post is about what a genuine Class 3 contains, where the numbers come from, and what an experienced reviewer will probe before the sanction paper is signed.

The Basis of Estimate Comes First

The single document that separates an estimate from a number is the Basis of Estimate (AACE 34R-05 gives the recommended practice). It states, in writing:

• What is being estimated — the scope, by system and battery limit, aligned with the design basis and layout at a stated revision.
• What deliverables the estimate stands on — which P&IDs, MTOs, equipment lists, and datasheets, at which revisions. This is where the class claim is made honest: a Class 3 estimate quantified from Class 5 deliverables is a Class 5 estimate with ambitions.
• How each part was priced — budget quotations, database rates, factors — with the pricing date, currency, and exchange-rate basis.
• Allowances — design growth per discipline, material take-off allowances, cutting and waste — stated as percentages, not buried in unit rates.
• What is excluded, and every assumption that replaces missing information.
• The escalation basis — from the pricing date to the expenditure profile, with the index used.

A reviewer who cannot find these in one document has already learned the most important thing about the estimate.

Building the Number

For a typical upstream facility estimate at Class 3, the structure runs:

Direct costs. Major equipment priced from budget quotations — at Class 3, quotes should cover the cost-dominant items rather than everything, with the balance from recent database pricing. Bulks (piping, steel, E&I, civils) quantified from material take-offs where drawings exist and factored where they do not. Construction labour from norms and productivity factors appropriate to the location — the factor that varies most between geographies and the one most worth challenging.

Indirect costs. Construction management, temporary facilities, camps and logistics, vendor representatives, commissioning support.

Services. Engineering, procurement, and management — estimated bottom-up from the schedule and team plan at this class, not as a blind percentage.

Owner's costs, escalation, and contingency — each as a separate, visible line. Merging them is the classic way to make an estimate unreviewable.

A simplified worked shape, for a modular facility:

Major equipment (budget quotes)             10.0
Bulks (MTO + factors, 0.9 × equipment)       9.0
Construction labour (0.6 × equipment)        6.0
Directs                                     25.0
Indirects (20% of directs)                   5.0
EPCM services (15% of directs + indirects)   4.5
Base estimate                               34.5
Contingency (risk-based, P50 ≈ 12%)          4.1
Total (excl. owner's costs & escalation)    38.6
Stated accuracy −15% / +25%  →  32.8 to 48.3

The factors above are illustrative — the point is the visible structure. Every line traces to a method, and every method is declared in the basis.

Contingency Is Not a Feeling

Contingency covers the cost of what the estimate basis does not yet know — quantity growth as engineering completes, pricing variability, execution risks within the defined scope. It does not cover scope changes, which belong to management reserve and change control, and it does not cover escalation, which has its own line.

At Class 3 the defensible route is risk-based (AACE 41R-08 range estimating with 44R-08 expected-value treatment of discrete risks): range the significant cost elements, capture discrete risks from the project risk register, run the Monte Carlo, and select the funding level as policy — commonly P50 for the estimate with the gap to P70/P80 visible to the owner. A flat percentage inherited from the last project is a placeholder, not an analysis — and reviewers can tell, because a risk-based contingency comes with a distribution and a driver list, and a guessed one comes alone.

What Sanction Reviewers Actually Check

Having sat on both sides of this table, the review converges on five questions:

1. Does the deliverable maturity support the class claim? The reviewer maps the basis-of-estimate deliverable list against 18R-97's maturity matrix. This is the first and most common failure.
2. Do the benchmarks hold? Cost per tonne of topsides, bulk-to-equipment ratios, labour hours per unit — compared against the reviewer's own history. Outliers are not necessarily wrong, but unexplained outliers are.
3. Where did the prices come from, and when? Quote coverage of the cost-dominant equipment, the age of database rates, and the escalation bridging pricing date to spend profile.
4. Is the contingency an analysis or a number? Distribution, drivers, and the funding-level decision — plus a drawdown expectation the project can be held to.
5. Is growth history acknowledged? Design-growth allowances compared with what this owner's last three projects actually experienced — the most uncomfortable and most useful comparison in the file.

An estimate that survives these five is not necessarily right — Class 3 still carries a real range, and the concept-stage decisions upstream of it dominate the outcome. But it is honest about what it knows, which is what a sanction decision actually requires.

Common Failures

• Class inflation — calling it Class 3 because the schedule needed one, not because the definition supports one.
• Invisible allowances — growth hidden inside unit rates, making the estimate impossible to reconcile against the next revision.
• Contingency doing three jobs — absorbing scope risk and escalation because those conversations were harder to have separately.
• A single currency-and-date blur — quotes from three years and four currencies combined without a stated basis.
• The estimate and the tender never reconciled — the pricing structure you mandate in RFQs is the calibration data for the next estimate; discard it and every project starts from zero.

Conclusion

A Class 3 estimate is an engineering deliverable with the same standards of traceability as a P&ID: a stated basis, methods matched to maturity, allowances and contingency that are visible and defensible, and an accuracy claim the underlying definition can support.

Built that way, the estimate does more than pass sanction. It becomes the project's cost model — the reference every trend, change, and tender result is measured against. Built any other way, it is a number with a class label, and the project finds out which during execution.