How Transport and Main Roads requires risk to be identified, modelled and costed on Queensland transport projects — base estimate plus probabilistically modelled contingency, expressed at P90, with a P50 reserve held by the project manager and a P75 value locked in at tender award.
The TMR Project Risk Management and Contingency Development Process Manual (1st Edition, March 2023) is the document that governs how risk is identified, modelled and costed on Queensland transport infrastructure projects. It sits beneath TMR's corporate Risk Management Framework and ISO 31000:2018, and is read alongside the Project Cost Estimating Manual (PCEM), the OnQ project-management framework, Engineering Policy 153 (Risk Context Profiles, July 2020) and the Australian Government's cost-estimation guidance notes.
It applies to the pre-construction phases — planning, concept and development — of QTRIP (Queensland Transport and Roads Investment Program) projects, and treats risk and contingency as a living process re-assessed at every milestone. Its legislative basis is the Financial Accountability Act 2009 (Qld) and the Financial and Performance Management Standard 2019.
The estimator's best prediction of quantities and current rates — before any inherent or contingent risk, or escalation, is added.
A financial reserve for identified, accepted risks. Developed probabilistically and expressed at P90 beyond the business-case milestone.
Anticipated cost increase over time (inflation, market, supply). Distinct from contingency — modelled and reported separately.
Why a client should care: if a project is QTRIP-funded, the contingency in its estimate must be developed the TMR way — base estimate plus probabilistically modelled contingency, expressed at P90, with a P50 reserve held by the project manager. An estimate that uses a flat-percentage contingency will not satisfy TMR for anything but the smallest, lowest-risk jobs.
The Project Assurance Framework (PAF) applies to capital expenditure ≥ $100M or significant risk/complexity; OnQ Type 1 gating applies to projects of $50M–<$100M. The risk process streamlines ISO 31000 into six activities: establish scope, context & criteria → identify → analyse → evaluate → treat → monitor, review, communicate & consult.
This is the manual's central conceptual split (Section 8). TMR's formal pairing is inherent vs contingent — not "inherent vs residual". The two kinds of risk are modelled in completely different ways.
Variability in the measured quantities and rates already in the base estimate. These are always present — you cannot remove them, only quantify them. They are modelled as ranges on existing line items.
Example: concrete priced at $200–$250/m³ over a quantity of 75–100 m³.
Modelled as: three-point ranges — lowest likely / most likely (= base estimate) / highest likely — using continuous distributions (Triangular, PERT).
Unmeasured items that may or may not occur — discrete risk events not in the base estimate. They are two-dimensional: each has a severity (cost impact) and a likelihood of occurring at all.
Example: a contaminated-soil discovery that may add $1.5M, but only has a 30% chance of occurring.
Modelled as: discrete events drawn from the dollarised risk register, using discrete functions (Bernoulli, Binomial, Discrete) for the probability×impact structure.
Inherent risk lives inside existing line items as a min/most-likely/max range. Contingent risk lives outside the base estimate as discrete events, each scaled by its probability of occurring.
Modelled directly on base-estimate line items as lowest likely / most likely / highest likely ranges (Triangular, PERT). The cost is incurred either way — only its magnitude is uncertain.
Modelled as separate two-dimensional events (Bernoulli, Binomial, Discrete) drawn from the dollarised risk register. Each may not occur at all — severity is scaled by likelihood.
Inherent ranges and contingent events are combined in one Monte Carlo run, with correlation modelled, to produce the S-curve from which P50, P75 and P90 are read.
The manual sets out ten categories of unplanned, contingent risk to scan for at every milestone. Where given, its historical-data benchmarks are shown below as the manual's illustrations — not fixed Cenex rates.
"Scope creep" as design matures and refinements emerge. Historically benchmarked at 3%–8% of construction cost per design stage.
Revised technical standards, government policy, planning schemes or regulatory requirements emerging during development or delivery.
Utility and service relocations, rail interfaces, council requirements and external approvals controlled by parties outside the project.
Scope change that alters the project's benefits — new service, performance or operational requirements as the project develops.
Costs borne directly by the principal (PM, consultants, approvals, internal staff). Historically ~30% of the unspent portion in early stages.
Schedule extension and the escalation/overheads it triggers. The manual cites Flyvbjerg (2004): projects on average delayed 20% versus the business-case date.
Variations and changes after award. Average cost increase ~11% of the EFCT (estimate for comparison with tenders) at award.
Valuation uncertainty, negotiation and tribunal outcomes, easements, and business relocation/disturbance costs.
Allowances for scope not yet quantified: 1–3% (S2D), 3–5% (business case), 5–7% (strategic estimate) of total construction cost.
Risks unique to the individual project that do not fall under the nine generic categories — captured directly from the risk workshop.
Engineering Policy 153 prescribes Risk Context Profiles across 10 profiling categories (geotechnical; environment / weather / cultural heritage & native title; vulnerable road users; stakeholders; procurement; project management; safety & wellbeing; contract administration; construction; finalisation). TMR's Risk Assessment & Ratings Matrix uses 8 risk dimensions (WHS; time/schedule; assets/operations/services; performance & capability; historical & Indigenous heritage; environmental/climate; media & reputation; financial).
The manual sets out two "distinctly different" methods — and is unambiguous about which it prefers.
A flat percentage on the base estimate (item- or factor-based). The manual states it "is not recommended by the department or projects funded by the Australian Government" and "can potentially deliver very inaccurate contingency provisions."
Acceptable only: as an approximation to P50/P90 where risk-adjusted out-turn is < $25M. Annexure A's factor-based tool rates six factors (scope, risk identification, constructability, key dates, site information, interfaces).
Illustrative only: in the manual's worked example these sum to 35% total contingency for P90, with the P50 contingency set at 40% of the P90 contingency (= 14%).
Quantitative risk analysis using Monte Carlo simulation on a structure of base estimate + contingency (+ escalation). Inherent risks are modelled on line items; contingent risks as discrete events.
Tooling: @RISK is TMR's named/recommended tool (Crystal Ball also referenced). External estimators must supply @RISK-compatible model files.
Settings: Uniform / Triangular / PERT / Discrete distributions; rule-of-thumb 10,000 iterations (more to smooth the S-curve); correlation must be modelled, or total risk is underestimated.
A deterministic flat percentage is a single guessed number bolted onto the base estimate. A probabilistic contingency is read off a simulated distribution that reflects the project's actual risks — defensible, and tied to a confidence level.
The Australian Government mandates Monte Carlo for all federally funded projects with total out-turn (P90) > $25M. TMR mandates it for high-risk/high-value work — state projects > $10M and federal projects > $25M.
It carries no confidence level and no link to the project's specific risks. The manual permits it only below the $25M out-turn threshold, as a rough stand-in for P50/P90.
All three confidence levels are read off the same S-curve (the cumulative probability distribution from the Monte Carlo run). Each one is owned by a different party, and the gap between them drives the savings the program reinvests.
The most likely estimate; 50% likelihood of not being exceeded. Used for budget-setting. The project manager holds a contingency reserve up to P50.
"Unlikely to be exceeded" at 90% confidence. TMR's estimating policy mandates P90 for all estimates beyond the business-case milestone. P90−P50 is held at portfolio level.
At tender award an Approved Project Delivery Value (APDV) is set at P75 — the EFCT run through the model. Applies to projects > $25M.
The cumulative S-curve maps cost ($) against the probability of not being exceeded. The PM holds to P50; portfolio holds P90−P50; at award the APDV is set at P75. Where the tender-award P75 lands below the business-case P90, the difference is project savings returned to the program for reinvestment.
The PM is given a contingency reserve up to P50, the budget-setting level. Day-to-day risk drawdown is managed against this reserve.
At tender award the EFCT is run through the Monte Carlo model and the Approved Project Delivery Value is set at P75. APDV applies to projects > $25M.
Project savings = business-case P90 budget − tender-award P75 (APDV), returned for reinvestment. Manual worked example: P90 $2,301,268 − P75 $1,995,444 = $305,824. The savings policy applies to projects ≥ $10M total budget (with exclusions such as local-government grants, corridor preservation and the natural-disaster program).
Project (to P50) → portfolio (P90−P50) → consolidated via APDV at award (P75). The total project cost / "Estimated cost" = base estimate + contingency + escalation, expressed at P90 (the total out-turn cost).
The probabilistic model is only as good as the data behind it. Two instruments supply that data — both run continuously across the project life cycle.
Described as "the most essential document". It captures the nature and level of each risk, the owner, mitigation measures, and likelihood and consequence — with consequence expressed in dollars.
"A vital component", held at every milestone, run in three phases:
TMR's Risk Assessment and Ratings Matrix (likelihood × consequence heatmap) is suitable only for Type 3 / low-risk projects and as the early-lifecycle screen feeding the quantitative analysis. It does not replace the Monte Carlo model for high-value or high-risk work.
Where projects are large or federally funded, the manual prescribes both the governance route and the exact contents of the @RISK output pack that must accompany the submission.
Federal submissions for projects > $25M must include the workshop details, SMEs and top risks with mitigations; the underlying risk register; and the full @RISK output set with re-runnable input files:
| Required output | What it demonstrates |
|---|---|
| Histogram | The shape and spread of simulated cost outcomes |
| S-curve | Cumulative distribution from which P50, P75 and P90 are read |
| Simulation summary | Sampling type, iteration count and random-number generator (RNG) |
| Tornado / regression-rank table | Which inputs drive the most cost variance (sensitivity) |
| Summary statistics | At 5% intervals from 5% to 95% |
| Re-runnable input files | So the reviewer can reproduce and stress-test the model |
Ongoing reporting (§11) covers contingency spent-to-date plus forecast versus baseline; remaining/unallocated contingency; opportunity value offsetting demand; ineffective treatments; and an overall statement on contingency health. Control-effectiveness ratings: Effective (<3% failure likelihood), Largely effective (<10%), Partially effective, and so on.
Independent, first-principles, framework-aligned. Cenex builds the contingency that satisfies the manual — and challenges it, rather than inflating it.
First-principles base estimate; inherent risk modelled on line items; contingent risk modelled from a dollarised risk register; @RISK at 10,000 iterations with correlation modelled; S-curve reported at P50/P90 with P75 for award.
Cenex facilitates the milestone risk workshops and produces the EP153 Risk Context Profiles, the dollarised risk register, and the federal-submission-ready @RISK output pack — histogram, S-curve, tornado, and summary stats at 5% intervals.
No downstream delivery interest. Contingency is challenged, not inflated — sized to genuine residual risk so the P90 holds and savings are realised at the P75 APDV, returned to the program.
The TMR manual rarely sits alone. Cenex aligns the same model to the Commonwealth (DITRDCA) P50/P90 out-turn requirements and the RES Contingency Guideline best-practice tests in one consistent run.
The TMR manual is one of three named frameworks Cenex builds contingency against. Explore the rest of the cost-risk hub.
The cross-jurisdiction best-practice authority on contingency development (3rd Edition, 2025).
The federal funding gate — P50 approval basis, P90 held and released on demonstrated need.
TMR vs RES vs Commonwealth, side by side — thresholds, P-values and contingency ownership.
Cenex's QRA and Monte Carlo service — methodology, contingency, sensitivity and when you need it.
See also the related PCEM chapter, Managing Risk & Developing Contingencies, and the cost-risk hub overview.
Cenex delivers independent, first-principles, @RISK-based contingency that satisfies the TMR manual at P50/P90 with P75 for award — and the federal-submission output pack to match.