Three principles the code enforces: every number originates in the engine; every number carries provenance; nothing is derived from a closed-box estimator’s content. This page states the method those principles serve.
Method
The calculation chain
Each step is a pure function. The trace object shown to the user is the record of these calls, in order, with inputs and outputs at each step.
Sizing parameterUser input per item — area, volume, power, duty. A warning if it lies outside the correlation’s validity range.
Purchased costFrom the library correlation at its base year and base CEPCI, or a vendor quote that bypasses the correlation and is tagged as such.
Material and pressure factorsFrom the library record; multiplicative on purchased cost.
EscalationCEPCI from the correlation base year to the estimate cost date. The index table is user-editable and the rows used are snapshotted into the estimate file.
Location factorThe library ships the US base location; any other location is a company row or a user override with a required note.
CurrencyUser-set FX rate and date.
Installed costSelectable: Lang, Hand, Guthrie bare-module, or user-defined percentages.
Indirects, EPCM, owner’s costPercentages, defaults by class, editable.
ContingencyPercentage, default by class from a cited source, editable.
Total and Monte Carlo bandTriangular or PERT per line or group, seeded; the seed is stored.
Class 3 method
AACE 18R-97 describes Class 3 as semi-detailed unit costs with assembly-level line items, prepared from preliminary quantity take-offs. The tool implements exactly that and nothing more. A bulk line carries a discipline, a quantity and a unit; quantities arrive by import of a line list or MTO, by manual entry, or by geometry from user-entered coordinates and rack polylines. Unit rates and labour norms come from a versioned company set, every row sourced, or from a public seed flagged as such. Labour is hours times craft rate times a productivity factor that the estimator states with a note.
Class 3 is claimed only when the user-stated definition level is within the Class 3 band and at least one line is quantity-priced. There is no library threshold on the quantity-priced share of cost; the share is reported so a reviewer can judge it.
Classes
Classes 5, 4 and 3 — what each one needs, what it is for
One reference plant. Three classes. One engine.
Move the slider: the app’s own capacity-scaling function runs on one checked library record and returns the figure. Switch the class: the drawing shows what a Class 5, 4 or 3 estimate has defined, and the band bracket redraws from the library’s class table. The page holds no arithmetic.
Illustrative reference plant — not an estimate. Tags are placeholders; each label names the parameter or quantity that class defines, never a value.
figure
The figure is the Class 5 screening result; Class 4 and 3 replace it with priced equipment and bulks. The slider’s ends are the exponent’s published validity window, as the engine reports it.
Record
Source
Reference point
Trace step, as the engine returns it
An estimator’s brief, class by class
Class 5 — concept screening
Class 5 is the screening estimate: project definition at its lowest, the widest expected accuracy range in the practice, and an end usage of concept screening — is this worth studying at all? In this tool it needs a capacity and a published process type; the engine scales a cited unit-capacity cost by a cited exponent. Report the class, its band, and a P-value from the seeded Monte Carlo run the tool stores. Contingency is the library’s default for the class, editable, cited. Concept select, nothing firmer.
Class 4 — study or feasibility
Class 4 is the study or feasibility estimate for pre-FEED: a preferred option exists, project definition has moved on, and the expected accuracy range narrows. It needs a sized equipment list with material and design pressure. Each item is priced from a cited correlation, factored, escalated, located and installed by Lang, Hand or bare-module; a size outside a correlation’s range is flagged, never hidden. State the class and quote the P-value from the stored run. Contingency is the library’s default for the class, editable, cited.
Class 3 — budget authorisation or control
Class 3 is the budget authorisation or control estimate that carries FEED to a final investment decision: project definition well advanced, the tightest expected accuracy range this tool will ever claim. It needs the Class 4 equipment list plus quantified bulks — a line list or MTO, manual lines, or geometry from coordinates you entered — priced at unit rates and labour norms from a company set you can cite, with a stated productivity factor. Fail the gate and the badge says Class 4. State the class, the P-value, and the library’s contingency default, editable, cited.
AACE class bands, as stored
Rendered from the library’s class table, the same record the app reads and the bracket above draws from. The class is assigned from the definition level the estimator states. No class better than 3 is ever presented or implied.
AACE 18R-97 classes as stored in the methods library (classes.json)
Class
Project definition
Expected accuracy
Class 5
0–2 %
-50 % / +100 %
Class 4
1–15 %
-30 % / +50 %
Class 3
10–40 %
-20 % / +30 %
The definition levels were checked against a printed reprint of the classification table; the accuracy percentages rest on the recommended practice as stated in the project brief and are noted in the record as not yet verified against a printed page of the RP itself.
Sources
Published sources
Equations and coefficients are implemented as methods, presented in the library’s own structure, and cited by edition, table and page. Publishers' typeset tables and figures are not reproduced. A record may carry an open item while a value is untranscribed; the release gate fails on any. The registry below is rendered from the library.
Plant Design and Economics for Chemical EngineersPeters, Timmerhaus & West · 5th ed. · 2003 · McGraw-Hillptw-5e
Analysis, Synthesis and Design of Chemical ProcessesTurton, Bailie, Whiting & Shaeiwitz · 3rd ed. · 2009 · Prentice Hallturton-3e
Analysis, Synthesis, and Design of Chemical ProcessesTurton, Bailie, Whiting, Shaeiwitz & Bhattacharyya · 5th ed. · 2018 · Pearsonturton-5e
Cost Estimate Classification System — As Applied in Engineering, Procurement, and Construction for the Process IndustriesAACE International · RP 18R-97, rev. 2020 ed. · 2020 · AACE Internationalaace-18r-97
Publication 946, How To Depreciate PropertyUS Internal Revenue Service · 2023 ed. · 2023 · IRSirs-p946-2023
The International System of Units (SI)BIPM · 9th ed. · 2019 · Bureau International des Poids et Mesuresbipm-si-9e
Guide for the Use of the International System of Units (SI)NIST · Special Publication 811, 2008 ed. · 2008 · National Institute of Standards and Technologynist-sp811-2008
Occupational Employment and Wage Statistics — May 2025 national estimatesUS Bureau of Labor Statistics · May 2025 ed. · 2026 · US BLSbls-oews-2025
OpenPyTEA2026openpytea-2026
Comparison of Commercial, State-of-the-Art, Fossil-Based Hydrogen Production TechnologiesLewis, McNaul, Jamieson, Henriksen, Matthews, White, Walsh, Grove, Shultz, Skone & Stevens (NETL) · DOE/NETL-2022/3241, April 12, 2022 ed. · 2022 · US Department of Energy, National Energy Technology Laboratorynetl-2022-3241
Engineering Costs (Cost Engineers' Notebook, Paper A7, Index 4.410)H. Carl Bauman (American Cyanamid) · Dec. 1964; presented at the 1964 AACE Convention, New York; 7 sheets ed. · 1964 · AACE International (then the American Association of Cost Engineers)bauman-1964
Factored Estimating for Process IndustriesKul B. Uppal, PE · conference paper, 14 pages ed. · 2007 · AACE International (venue and paper number not stated on the copy)uppal-2007
Skills and Knowledge of Cost EngineeringAACE International; Dr. Makarand Hastak, PE CCP, editor · 6th ed. · 2015 · AACE Internationalaace-sk6
Quality Guidelines for Energy System Studies: Cost Estimation Methodology for NETL Assessments of Power Plant PerformanceUS DOE National Energy Technology Laboratory · NETL-PUB-22580, February 2021 ed. · 2021 · US Department of Energy, NETLnetl-qgess-2021
Quality Guidelines for Energy System Studies: Cost Estimation Methodology for NETL Assessments of Power Plant PerformanceUS DOE National Energy Technology Laboratory · DOE/NETL-2011/1455, April 2011 ed. · 2011 · US Department of Energy, NETLnetl-qgess-2011
A pragmatic approach to using resource loading, production and learning curves on construction projects (web version: Applying Resource Loading, Production & Learning Curves to Construction)R. Max Wideman · Canadian Journal of Civil Engineering 21(6), Dec. 1994; quotes Allen (1979) ed. · 1994 · maxwideman.comwideman-1994
Planning, Estimating, and Control of Chemical Construction ProjectsPablo F. Navarrete and William C. Cole · 2nd ed. · 2001 · Marcel Dekkernavarrete-cole-2001
Contingency and Accuracy Range worksheets, Hackney model and RAND model (illustrative appendix to AACE RPs 42R-08 and 43R-08)AACE International; RAND model information used with permission from the RAND Corporation · 3-page workbook printout, undated ed. · 2008 · AACE Internationalaace-43r-08-models
Understanding Cost Growth and Performance Shortfalls in Pioneer Process Plants (R-2569-DOE)E.W. Merrow, K.E. Phillips and C.W. Myers · RAND publication R-2569-DOE ed. · 1981 · RAND Corporation for the US Department of Energyrand-r2569
Control and Management of Capital ProjectsJohn W. Hackney · 2nd ed. · 1992 · AACE Internationalhackney-1992
A new edition, an annual index update or a standards revision enters the library by superseding, never by overwriting: an issued estimate keeps computing on the records it names. Second reads and open items are listed in the library changelog.
Scope
What it deliberately does not do
No mechanical design (ASME or TEMA) and no closed-box volumetric bulk synthesis. Class 3 bulks are supplied quantities or pure geometry from coordinates the user entered.
No bundled commercial cost data. Import templates only, so a licensed rate set stays the licensee’s.
No content derived from any closed-box estimator: no proprietary component codes, cost basis values, factor sets, report layouts or terminology beyond generic industry terms.
No drawing or P&ID vision ingest, no document ingest by a language model, and no machine-learned residual correction on the estimate.
No class better than 3. The band on every output is the band for the class the definition supports.
AI assist never contributes a number. It proposes record keys, extracted text fields and prose; the user confirms; the engine computes identically with it off.
Questions
Write to hello@infertic.com. The product is built by Vantis Project Intelligence Ltd.