Estimating framework

Estimate Classes & Australian Estimate Stages

How mature is the estimate, how accurate can it be, and which contingency level will the funding authority expect? This guide lines up AACE estimate classes 1 to 5 with the Commonwealth, Queensland, NSW and Victorian estimate stages, and sets out the maturity checks Cenex applies to every risk-adjusted estimate.

AACE 18R-97 DITRDCA GN2 · GN3A · GN4 TMR PCEM 9th ed. INSW CCF DTF RTCC · HVHR
Foundation

An estimate is only as good as the definition behind it

Estimate classes tie the expected accuracy of an estimate to how much of the project is actually defined, not to the effort, software or time spent on it. A class 5 estimate prepared with great care is still a class 5 estimate.

CLASS

How mature is it?

The share of engineering complete, and the estimating method that maturity allows.

STAGE

What decision is it for?

Strategic assessment, business case, pre-tender or award: each authority names its own stages.

P-LEVEL

How much confidence is funded?

P50 or P90, and who holds the difference, varies by authority and project tier.

AACE International

AACE estimate classes 1 to 5

AACE Recommended Practice 18R-97 is the generic classification matrix. Accuracy is shown as a pair of ranges: the best and worst of what is typical for that class, on the low side and on the high side.

Class Maturity
(project definition)		 Typical end use Methodology Expected accuracy: low Expected accuracy: high
Class 50% to 2%Concept screeningCapacity-factored or parametric models, judgement or analogy−20% to −50%+30% to +100%
Class 41% to 15%Study or feasibilityEquipment-factored or parametric models−15% to −30%+20% to +50%
Class 310% to 40%Budget authorisation or controlSemi-detailed unit costs with assembly-level line items−10% to −20%+10% to +30%
Class 230% to 75%Control, or bid and tenderDetailed unit costs with forced detailed take-off−5% to −15%+5% to +20%
Class 165% to 100%Check estimate, or bid and tenderDetailed unit costs with detailed take-off−3% to −10%+3% to +15%
 AACE RP 18R-97 generic matrix. The ranges describe the spread around the point estimate (including contingency) at about 80% confidence, roughly P10 to P90. AACE 98R-18 covers road and rail infrastructure.
Ranges overlap on purpose

Maturity bands overlap because industries reach each class at different points of design. Where an estimate sits on a boundary, Cenex classifies it to the less mature class.

Typical, not guaranteed

These are typical ranges, not targets. A project's own range comes from its QRA; the class ranges are the benchmark that QRA is checked against.

Each class is covered in depth in our AACE Guide: Class 5 · Class 4 · Class 3 · Class 2 · Class 1 · Accuracy range calculator.

Australian authorities

Estimate stages and the class to expect

Australian funding authorities describe estimates by project stage rather than by AACE class. The table below is the starting crosswalk Cenex uses; the class actually claimed is set by design maturity and data quality, as described in the maturity checks.

Generic stageTypical AACE classTypical estimating basis
Strategic assessmentClass 5Parametric and benchmark rates ($/km, $/m²), analogous projects
Preliminary business caseClass 4Concept design; quantities factored or measured from concept drawings
Final business caseClass 3Reference design; measured quantities with semi-detailed rates
Pre-tender (detailed design)Class 2Detailed design; first-principles build-ups
Tender / contract awardClass 1Issued-for-tender design; check estimate against bids
 Indicative: a starting point, not a rule. Projects can be ahead of or behind their stage. Authority-specific stage names are below.
COMMONWEALTH

DITRDCA Cost Estimation Guidance

Applies to projects seeking funding under the Infrastructure Investment Program.

  • GN2: the base estimate excludes contingency and escalation, and is broken down in the Project Cost Breakdown (PCB) categories.
  • GN3A (v2, Nov 2023): probabilistic P50 and P90 outturn for projects over $25M total project cost; risk factors are the preferred method; 5,000–10,000 iterations; correlation tested at 0 and 1.
  • GN4 (v3, July 2026): escalation sits outside contingency; outturn is P50 and P90 by financial year.

Stage names follow the department's Notes on Administration; confirm the current names for each submission. See also our DITRDCA guide.

QUEENSLAND

TMR Project Cost Estimating Manual (9th ed., Nov 2025)

Estimates progress through the Queensland project lifecycle phases, from strategic planning through concept and development to implementation and finalisation.

  • Planned (line-item) and unplanned (register) risk, reported at P50 and P90; the QTRIP budget is set at P90.
  • An estimate category check against Table 3.3.7's expected contingency ranges.
  • Annexure L (Form M4755) summary; a P5–P95 table at 5% steps, unescalated (§4.1.13.3); contingency monitoring (§10.1.21).

More in our PCEM phase estimates guide and AACE vs PCEM.

NEW SOUTH WALES

INSW Cost Control Framework (Mar 2025)

With NSW Treasury TPG24-29 and TPG25-06.

  • Projects are tiered. Monte Carlo is the minimum method for Tier 1 (high profile, high risk).
  • Contingency is provisioned at P90 for Tier 1 and P50 for Tier 2.
  • Schedule QRA is required; the mean is the central case for cost–benefit analysis.

Gate numbering (0–3) and agency escalation rates are confirmed for each project.

VICTORIA

DTF Risk, Time, Cost & Contingency; HVHR (2024)

The method required depends on whether the project is High Value High Risk (HVHR) and on its stage:

  • non-HVHR: deterministic;
  • HVHR business case: parametric, with probabilistic simulation of risk events;
  • HVHR procurement: probabilistic simulation;
  • HVHR delivery: integrated cost and schedule risk analysis.

The required percentile at each stage is confirmed with DTF for each project.

Contingency

How contingency P-levels differ

Every authority asks for P50 and P90, but they fund, hold and escalate them differently. The submission has to report the level, the split and the escalation basis the authority expects.

AuthorityFunded levelWho holds whatEscalation
Commonwealth (DITRDCA)P50 outturn approvedP90 held against the projectOutside contingency (GN4); outturn by financial year
Queensland (TMR)P90 (funding, QTRIP budget)Funded at P90; P50, P75 and P80 also reportedOutside contingency
NSW (INSW)P90 Tier 1 · P50 Tier 2Project Director holds P50 contingency; agency senior executive manages P90 − P50Applied by element; sensitivity and deferral shown
Victoria (DTF)TEI, usually set at P90 (confirm with DTF)Excess risk split between project, department and DTFInside the base risk estimate (the P50 is outturn)
 From the authorities' published documents as reviewed in October 2026.

For the full framework comparison, see Australian Risk Frameworks Compared. For how we build the QRA itself, see Risk Modelling & P90 Contingency.

Quality assurance

Estimate maturity and data-quality checks

Is the risk-adjusted estimate consistent with how mature the estimate is? Every Cenex QRA model carries an Estimate Maturity check that classifies the estimate, rates the data behind it, and tests the simulated result against outside benchmarks.

Classify

Record the stage and design maturity. The maturity suggests a class using the AACE 18R-97 bands; we confirm or overwrite it.

Rate the data

Score eight elements from 1 to 5 against written descriptors. The weighted score caps the class the estimate can claim.

Run

Simulate the model as normal.

Check

Compare the result with the class's accuracy range, the contingency band and the reference class. Amber is a prompt to explain or revisit, not a failure.

The eight data-quality elements

ElementScore 1 looks likeScore 3 looks likeScore 5 looks like
Scope definitionNeed and options only; no defined solutionPreferred option with a concept layoutDetailed design substantially complete; scope frozen
QuantitiesFactored from similar projectsMeasured from concept drawings or a 3D modelMeasured from detailed design and checked
Rates and pricing basisBenchmark $/km or $/m²Recent tender rates adjusted to this projectFirst-principles build-up and current supplier quotes
Programme and stagingDuration assumedSummary programme with key stagingResourced construction programme; staging agreed
Site investigationDesktop onlySome boreholes and service records; contamination study plannedGeotechnical, services (pot-holing) and contamination investigations complete
Risk identificationNo workshop; generic allowancesRisk workshop held; register with ownersSeveral workshops; quantified register reviewed by stakeholders
Approvals and stakeholdersApprovals pathway unknownPathway known; key approvals in progressApprovals obtained
Market knowledgeNo market informationRecent comparable tendersMarket sounding done; bidder interest confirmed
 Weighted score of 4 or more (High) supports up to Class 2; 3 or more (Medium) Class 3; 2 or more (Low) Class 4; below 2 (Very low) Class 5.

The checks against the result

CheckCompared withWhat a flag means
Stage and maturity agreeClass by maturity vs class typical for the stageMore than one class apart: the stage or maturity is mis-stated, or the estimate is ahead of or behind its stage.
Data quality supports the classClass claimed vs class the data supportsThe class claimed is more mature than the data allows. Improve the weakest elements or use a less mature class.
Downside range (P10 below P50)The class's low accuracy rangeNarrower than expected: ranges are probably too tight or too independent.
Upside range (P90 above P50)The class's high accuracy rangeNarrower: high ranges, register or correlation understated. Wider: check for double counting.
P90 contingency as % of baseAn indicative band for the class, set by your organisation's or the agency's policyOutside the band: explain why (well-defined scope, high-risk site) or revisit the inputs.
Reference class (P50 and P80)Optimism-bias uplifts for the project typeWell below the outside view: be ready to explain why this project is better defined than its class.
Tail beyond P90Mean of the worst 10% of outcomes, above P90, as % of baseAbove 5%: a few rare, costly risks dominate; consider treating them, or a management reserve.

The outside view

The reference-class cross-check uses the optimism-bias uplifts from Flyvbjerg and COWI (2004) for the UK Department for Transport: the cost growth similar projects showed after approval, at 50% and 80% certainty.

Reference classUplift at P50Uplift at P80
Roads15%32%
Rail40%57%
Fixed links (bridges and tunnels)23%55%
Never add it on top

These uplifts apply to an approved estimate with no project-specific risk analysis. Stacking one on a full QRA counts the same risk twice, so we use the reference class only to test the QRA. Our Reference Class Forecasting guide explains the method.

Go deeper

Related guides

AACE GUIDECost Estimate Classification System

The five classes in depth, with a page for each.

Read → PCEM GUIDEPhase-Based Estimates

How TMR estimates evolve through the project lifecycle.

Read → RISK GUIDECost Risk & Contingency Hub

TMR, RES and DITRDCA frameworks, Monte Carlo and P50/P90.

Read → SERVICERisk Modelling & P90 Contingency

How Cenex builds and reports a QRA.

Read →

Not sure what class your estimate is?

We review estimates against the class and stage they are meant to support, and tell you what it would take to move up a class.

Talk to an estimator Cost estimating services