Chance the final cost comes in at or below each value. Hover to read it.
Cost build-up to the budget
Spend profile
Monthly spend
Cumulative spend
Design-stage hold points from the lessons library
Project details
Stage and cost model
Estimate class
Stage crosswalk: RIBA · FEL · IA · PAF · UK business case · AACE · NRM1
Pick the project type first. It sets the suggested reference class on the Reference Class tab and decides how directly the rail station lessons library applies.
Outside view — reference class forecasting
This takes the bottom-up estimate and applies the empirical uplift for the chosen reference class at the confidence level you want to budget to. The uplift comes from what comparable completed projects actually cost. It does not come from this project's risk register.
Uplift applied
Uplift curve for the selected class
Reference class register
Commandment 5, Validate With External Data. Every uplift carries a source and a verification status. Values marked verify are reproduced from the published tables as best known and should be checked against the primary document before they go into a client report. If you have your own completed-project outturn data, add it as a custom class. That is the strongest reference class you can use.
Baseline gate readiness
Ten gate tests from the lessons library, each broken into evidence criteria. Status is worked out from the evidence, not chosen by hand. Must criteria are hold points. Where a criterion has a number (open clashes, % of interfaces with owners), you enter the number and it sets the result. Cost and schedule checks read live from this estimate. A Critical gate at Fail means the estimate can still be reported, but it should not be approved as a budget until the gap is closed or priced.
Click a gate to open its criteria.
RIBA 2–3 Rail & Station Design Lessons Library
Driver register — mark what applies to this project
Lessons
Sources
Method, limitations and data notes
Project documents — load the engine
Drop in drawings, specifications, existing cost plans, BoQs, WBS and CBS spreadsheets, benchmark rates or completed-project outturn data. Spreadsheets import directly after you map their columns. Claude reads drawings and specifications and proposes quantities, scope and risks, and nothing reaches the estimate until you accept it.
DROP FILES HERE
or click to choose · PDF (drawings, specs) · XLSX / XLS / CSV · DOCX · PNG / JPG · TXT
CAD and BIM files (DWG, IFC, RVT) can't be read in a browser. Export the sheets to PDF first.
Document register
For ticked documents:
Original files stay in this browser on this computer. Use "Save + docs" to carry them to another machine. When documents are sent to Claude they go directly to api.anthropic.com. Check the client's confidentiality terms before sending their drawings. Anthropic currently limits PDFs to about 100 pages and 32 MB per request, so for bigger sets enter a page range; those pages are sent as images.
Work and cost breakdown structures
Every line can carry a WBS code (where the work sits) and a CBS code (what kind of cost it is). Totals roll up through dotted codes, and the outside-view budget is spread across them in proportion.
Budget by CBS (top level)
Budget by WBS (top level)
Benchmarking
AI indicative rate ranges by discipline
Claude proposes P10–P90 unit-rate ranges from the scope. These are a starting point for the quantity surveyor. They are not reference-class data, and every row stays tagged AI indicative until you overwrite the rate or mark it checked.
Estimate line items
Budget by discipline
Widest line-item ranges (top 10)
Probabilistic line-item roll-up (Monte Carlo)
Percentiles do not add. Summing every line's P80 overstates the project P80 unless every item moves together perfectly. This roll-up samples each line from a lognormal fitted to its P10 and P90 and links them through one shared market factor. The correlation setting is a judgement: 0 means independent lines, 1 means they all move together.
Oncosts and adjustments
Cost build-up chart
Cost build-up to outside view
Whole-life cost — option test, not capex only
WLC = capital + energy + planned maintenance + expected reactive maintenance + renewals + possession/disruption + end of life − residual value. The library treats this as a decision framework, not a benchmark. Populate it with the client's asset data, operating hours, failure rates, possession costs and discount policy.
Presets: Infrastructure Australia has used 7% real as its central rate with 4% and 10% sensitivities. The UK Green Book uses 3.5% for years 0–30, stepping down after. Confirm the client's current policy before relying on either.
Recurring cost families (annual, real AUD)
Renewals and major replacements
Present value
Discount rate sensitivity
Effectus AI advisor — RCF, PCEM and design-stage lessons
The advisor sees your project, estimate, reference class, gate results and the driver register. Try one of these:
Cost estimate report
Anthropic API connection
Needed only for the AI features. Everything else works offline. The key goes straight from this browser to api.anthropic.com and nowhere else.
About this build
Effectus Cost Intelligence — RCF standalone
Eight cost model formats: TMR PCEM, UK TAG, US FTA, Canada PSPC, AACE, NRM, Australian Government and the Platform 19 legacy structure Stage frameworks: RIBA, FEL, OnQ, IA, PAF, UK business case and more, with a crosswalk Effectus Gate Criteria v1.0 (48 evidence criteria)
Reference class forecasting after Flyvbjerg's outside view, with UK DfT and HM Treasury uplift tables
RIBA 2–3 Rail & Station Design Lessons Library (36 lessons, 12 drivers, 10 gate tests, 14 sources)
Whole-life cost framework and PV calculator
Monte Carlo line-item roll-up (lognormal, one-factor correlation)
Ten Commandments tagging: Effectus mapping, proposed
Data Projects save to a .json file (Save / Open, top right). The browser also keeps a working copy where browser storage is available.