<!-- argument-hint: [characteristic, 12-step process step, estimating method, risk/uncertainty, EVM term, audit, software/learning-curve/AOA, chapter number] -->
GAO Cost Estimating and Assessment Guide (GAO-20-195G)
Source: GAO (US Government work, public domain) | Chapters: 8
When to use
Use this skill to develop, manage, or evaluate a program cost estimate: framing the estimate's purpose and scope, structuring it with a WBS, choosing estimating methods, turning a point estimate into a risk-adjusted budget, documenting and updating it, auditing it against GAO's reliability criteria, and connecting it to earned value management for execution. This is the cost-estimating companion to the GAO gao-schedule and gao-tra guides and the management-side counterpart to a program's life cycle cost estimate.
Prerequisites: none — plain Markdown; no MCP server, API key, or licence tier needed at runtime.
How to Use This Skill
- Without arguments — load the core frameworks below: the four characteristics + 18 best practices, the 12-step process, the estimating methods, risk/uncertainty, EVM, and the specialized techniques.
- With a topic — ask about a characteristic ("credible"), a step ("ground rules and assumptions", "point estimate"), a method ("parametric", "learning curve"), a risk concept ("Monte Carlo", "confidence level", "contingency"), or an EVM term ("CPI", "EAC", "performance measurement baseline").
- With a chapter —
ch02/ch03/ch05 (the 12 steps), ch04 (risk/uncertainty), ch06 (auditing), ch07 (EVM), ch08 (specialized techniques).
Supporting files: glossary.md, patterns.md, cheatsheet.md.
One discipline, two uses. The Guide is both an authoring standard (how to build a reliable estimate) and an audit standard (how to grade one). The same four characteristics and 18 best practices serve both. EVM (ch07) is the execution-side bridge: the estimate says what a program could cost; EVM measures whether it is delivering against a baseline built from that estimate.
Core Frameworks & Mental Models
What a cost estimate is — and why reliability is a process, not luck
A cost estimate is the sum of individual cost elements, built from established methods and valid data, projecting what a program will cost in the future given what is known now. It is not a one-time deliverable — managing it means revising it as actuals arrive, adjusting it when the program changes, and analyzing the gaps. The Guide's central claim: reliability is the predictable output of a disciplined, repeatable process, not seniority or luck. A reliable estimate matters structurally for government — every dollar committed to one program is denied to another, so an unrealistic estimate can ripple cuts across a whole portfolio.
The Four Characteristics of a Reliable Cost Estimate
GAO's research identifies four characteristics that a reliable estimate exhibits. They are the assessment yardstick auditors apply, and the design targets an estimator engineers toward:
| Characteristic | What it demands |
|---|
| Comprehensive | full life cycle + complete scope (a complete WBS); all GR&As documented; nothing omitted or double-counted |
| Well documented | sources, methods, and calculations captured so an unfamiliar analyst could rebuild it; traced to source data |
| Accurate | unbiased (not optimistic or conservative); most-likely-cost basis; inflation-adjusted; reconciled to actuals |
| Credible | limitations exposed via sensitivity and risk/uncertainty analysis; major drivers cross-checked; reconciled to an independent cost estimate (ICE) |
The 18 best practices are the actionable steps that produce these characteristics; each best practice maps to one characteristic. An estimate that fully reflects all four is deemed reliable.
The 12-Step Cost Estimating Process
The procedural backbone, grouped into four phases:
- Plan & define (Steps 1-4): (1) define the estimate's purpose — intended use fixes scope and detail; (2) develop the estimating plan — team, study plan, schedule; (3) define the program via a cost-free technical baseline description; (4) determine the estimating structure — the work breakdown structure (WBS), the product-oriented decomposition that ties cost, schedule, EVM, and risk together.
- Build the number (Steps 5-7): (5) identify ground rules and assumptions (global + element-specific); (6) obtain the data — collect, validate, normalize (the most difficult, time-consuming step; data quality is the biggest lever); (7) develop the point estimate — select a method per WBS element, build the model, time-phase, sum, and cross-check against an ICE.
- Quantify confidence (Steps 8-9): (8) sensitivity analysis (vary one input at a time); (9) risk and uncertainty analysis (vary many inputs together via simulation → probability distribution + confidence level).
- Finalize & sustain (Steps 10-12): (10) document the estimate; (11) present it to management for formal approval (not finished until approved); (12) update it against actuals and program changes over the life cycle.
Each step's chapter closes with a "Survey of Step" — Process Tasks, Best Practices, and Likely Effects if criteria are not fully met — which is what makes each step auditable.
Estimating Methods
Choose by program maturity and available data; the method can differ per WBS element:
| Method | Use when | Trade-off |
|---|
| Analogy | early; a close similar system exists | fast, low data; weak without a good analog |
| Parametric | drivers + cost history, design still vague | repeatable via cost estimating relationships (CERs); needs valid data range |
| Engineering build-up | mature design | most accurate; slow and data-hungry |
Supplemented by expert opinion, extrapolation from actuals (averages, EACs), and learning curves (Y = AX^b) for recurring production.
Risk and Uncertainty → a risk-adjusted budget
A point estimate alone carries no confidence information and tends to under-fund. Sensitivity analysis isolates one input at a time (ranked on a tornado chart) to find what the estimate is most responsive to. Risk and uncertainty analysis lets inputs vary together — via the three-point method (min/most-likely/max) and/or the risk-driver method (probability × cost effect, from the risk register) — driven by Monte Carlo simulation, producing a total-cost probability distribution. Reading the cumulative distribution (S curve) gives a defensible confidence level; contingency is sized as the gap between the point estimate and that level.
Earned Value Management (EVM) — measuring execution
Spending data alone misleads (the railroad example: looks under-budget while badly behind). EVM adds the missing third number — the value of work performed. Three measures: BCWS (planned value), BCWP (earned value), ACWP (actual cost). Derived indexes: CPI = BCWP/ACWP (cost), SPI = BCWP/BCWS (schedule), TCPI (efficiency needed on remaining work). The EAC forecasts total cost from cumulative CPI. The estimate is converted into a time-phased performance measurement baseline (PMB); an integrated baseline review (IBR) confirms it is executable — the crucial link between cost estimating and EVM. EVM systems are validated against EIA-748's 32 guidelines. OMB requires EVM for major acquisitions with development effort.
Specialized Techniques & Internal Control
Beyond the 12 steps: software cost estimating (size first via SLOC/function points/COSMIC, convert to effort; classify code new/reused/adapted/auto-generated); learning curves (unit vs. cumulative-average theory; Anderlohr/retrograde for production breaks); Analysis of Alternatives (AOA) (22 best practices in 5 phases; NPV where benefits monetize); WBS templates (MIL-STD-881D, NASA/DOE/PMI handbooks); and the Green Book (GAO-14-704G) internal-control framework (5 components, 17 principles) — because the cost estimating process is itself an internal control.
Chapter Index
| # | Section | Key content |
|---|
| ch01 | Fundamentals & the four characteristics | What a cost estimate is; reliability as process; the four characteristics; 18 best practices; 12-step process; LCCE; why government needs reliable estimates |
| ch02 | Steps 1-4: Planning & program definition | Purpose & scope; estimating plan/team; technical baseline description; the work breakdown structure (WBS) |
| ch03 | Steps 5-7: Ground rules, data, point estimate | Global vs. element-specific GR&As; data collection/validation/normalization; estimating methods (analogy/parametric/build-up); cross-checks and the ICE |
| ch04 | Steps 8-9: Sensitivity, risk & uncertainty | Sensitivity (one-at-a-time, tornado chart); three-point & risk-driver methods; Monte Carlo; S curve, confidence level, contingency |
| ch05 | Steps 10-12: Document, present, update | Documentation template & rationale; management approval; updating against actuals and program changes |
| ch06 | Auditing & validating | The characteristics/best practices as gradable criteria; mapping practices to characteristics; assessing agency guidance against the 12 steps; audit criteria & data collection |
| ch07 | Earned value management | EIA-748 (32 guidelines); 13 EVM activities; PMB/IBR/CPR; BCWS/BCWP/ACWP; CPI/SPI/TCPI; EAC; OTB/OTS; EVM system acceptance |
| ch08 | Specialized techniques & internal control | Software cost estimating; learning curves (Anderlohr/retrograde); WBS templates; TRLs; AOA (22 practices); the Green Book (5 components, 17 principles) |
Topic Index
- 18 best practices → ch01, ch06, cheatsheet
- Accurate (characteristic) → ch01, ch06
- Analogy method → ch03, cheatsheet
- Analysis of Alternatives (AOA) → ch08, ch01
- Anderlohr / retrograde (production break) → ch08
- Audit / validation of an estimate → ch06, patterns
- CPI / SPI / TCPI (performance indexes) → ch07, cheatsheet
- Comprehensive (characteristic) → ch01, ch06
- Confidence level / S curve → ch04, cheatsheet
- Contingency / management reserve → ch04, ch07
- Cost estimating relationship (CER) → ch03, ch04
- Credible (characteristic) → ch01, ch06
- Data collection / validation / normalization → ch03
- Documentation (Step 10) → ch05, patterns
- Earned value management (EVM) → ch07, ch01, cheatsheet
- EIA-748 / 32 guidelines → ch07
- Engineering build-up method → ch03, cheatsheet
- Estimate at completion (EAC) → ch07
- Four characteristics (reliable estimate) → ch01, ch06, cheatsheet
- Green Book / internal control → ch08
- Ground rules & assumptions (GR&A) → ch03
- Independent cost estimate (ICE) → ch03, ch05
- Integrated baseline review (IBR) → ch07, ch08
- Learning curve (Y = AX^b) → ch08, ch03
- Life cycle cost estimate (LCCE) → ch01, ch02
- Monte Carlo simulation → ch04
- Parametric method → ch03, cheatsheet
- Performance measurement baseline (PMB) → ch07
- Point estimate (Step 7) → ch03
- Present / management approval (Step 11) → ch05
- Risk & uncertainty analysis (Step 9) → ch04, cheatsheet
- Risk driver / risk register → ch04
- Sensitivity analysis (Step 8) → ch04, cheatsheet
- Software cost estimating → ch08
- Technical baseline description (Step 3) → ch02
- Technology readiness level (TRL) → ch08
- Three-point estimate → ch04
- Tornado chart → ch04
- 12-step process → ch02, ch03, ch04, ch05, cheatsheet
- Update the estimate (Step 12) → ch05
- Well documented (characteristic) → ch01, ch05, ch06
- Work breakdown structure (WBS) → ch02, ch08
- WBS templates (MIL-STD-881D, NASA/DOE/PMI) → ch08
Supporting Files
- glossary.md — key cost-estimating, risk, and EVM terms, alphabetical, with chapter references
- patterns.md — reusable patterns (run the 12 steps; engineer toward the four characteristics; pick a method by maturity; convert the point estimate to a risk-adjusted budget; document/update; audit; pair with EVM; specialized techniques) with When/How/Trade-offs
- cheatsheet.md — the 12 steps, four-characteristics table, method-selection table, risk toolkit, EVM index reference, and tells & smells
Scope & Limits
Covers: program cost estimating per the GAO Cost Estimating and Assessment Guide (GAO-20-195G, March 2020) — the four characteristics of a reliable estimate and the 18 best practices; the full 12-step cost estimating process; estimating methods (analogy, parametric, engineering build-up, learning curves); sensitivity, risk, and uncertainty analysis (Monte Carlo, confidence levels, contingency); documenting, presenting, and updating; auditing/validating an estimate against the characteristics; earned value management (EIA-748, the EVM measures and indexes, PMB, IBR, EAC forecasting, system acceptance); and specialized techniques (software cost estimating, Analysis of Alternatives, WBS templates, technology readiness levels, the Green Book internal-control framework).
Thin on / does not cover in depth: schedule estimating and schedule risk analysis — the Guide deliberately hands these to the GAO Schedule Assessment Guide (GAO-16-89G; see gao-schedule), and technology-maturity assessment to the Technology Readiness Assessment Guide (GAO-20-48G; see gao-tra). It does not reproduce the full text of referenced standards and frameworks — EIA-748, MIL-STD-881D, the Green Book (GAO-14-704G), ICEAA/AACEI/PMI standards — these are named and summarized, not copied. Embedded source figures and tables (discipline maps, WBS diagrams, worked method tables, EVM exhibits) are described/paraphrased here, not reproduced; redraw from the originals if a rendered exhibit is needed. The Guide is GAO best-practice guidance, oriented to US federal acquisitions and DoD/NASA examples; it is not a regulation, though several statutes (Weapon Systems Acquisition Reform Act, Nunn-McCurdy, FASA, Clinger-Cohen, the FAR/DFARS EVM provisions) make parts of it effectively mandatory for major acquisitions.
Jurisdiction: US Government public domain work. The guidance is voluntary best practice but is broadly adopted and partly mandated for federal major acquisitions; it is adaptable to non-federal programs.