ASQ Certified Quality Engineer (CQE) — Advanced Study Guide

ASQ Certified Quality Engineer (CQE) — Advanced Study Guide

Issuing body: American Society for Quality (ASQ) Jurisdiction: International (US-based) Level: Professional Prerequisites: 8 years on-the-job experience in CQE Body of Knowledge areas, including 3 years in a decision-making position (full-time, paid). Education waivers: diploma 1 yr, associate's 2 yrs, bachelor's 4 yrs, master's/doctorate 5 yrs (one waiver only). Exam ANAB-accredited under ISO/IEC 17024. Renewal: Every 3 years — 18 recertification units (RUs) or retake the exam.

1. How the credential is earned

1. Verify eligibility against the 8-year rule (minus your one education waiver). ASQ audits applications — "decision-making" means you had authority to define, execute, or control projects/processes and owned the outcome.
2. Apply on asq.org; schedule at a Prometric center (computer-based; paper-and-pencil still offered at select events).
3. Sit the exam: 175 multiple-choice questions (160 scored + 15 unscored pretest), 5 hours 18 minutes, open book (bring bound references — no loose sheets). Scaled pass score 550/750.
4. Maintain: 18 RUs per 3-year cycle (employment, courses, presentations, published work) or re-examine.

Exam strategy: 5h18m for 175 questions ≈ 109 seconds each — generous by certification standards because it's open book and calculation-heavy. Domain VI (Quantitative Methods) is 36 of 160 scored questions (~23%), the single biggest block. Tab your references by domain; pre-work every formula in this guide so the exam is arithmetic, not derivation. The trap is over-looking-up: if you look up more than ~20 questions, you'll run out of time.

2. Knowledge domains

ASQ's CQE Body of Knowledge has seven domains. Question counts below are per widely published prep-provider summaries of the ASQ BoK (I: 18, II: 16, III: 23, IV: 25, V: 27, VI: 36, VII: 15 of 160 scored) — verify against the current ASQ BoK before booking, as ASQ rebalances periodically.

Domain I — Management and Leadership (~11%)

Quality philosophies (Deming's 14 points and system of profound knowledge; Juran's trilogy — plan/control/improve — and fitness for use; Crosby's zero defects; Taguchi's loss function; Ishikawa's fishbone), the ASQ Code of Ethics (the exam tests that public safety and honesty outrank employer loyalty), leadership vs. management, team dynamics and facilitation, cost of quality: Prevention + Appraisal + Internal Failure + External Failure (PAF model). The exam's CoQ trap: prevention spending reduces total CoQ — the optimal CoQ is not zero quality cost, it's the minimum of the total curve, and external failures cost roughly an order of magnitude more than catching the defect in-house.

Domain II — The Quality System (~10%)

QMS standards and models (ISO 9001 structure, the process approach, risk-based thinking), documentation hierarchy (policy → procedures → work instructions → records), quality planning, training and competence, auditing fundamentals (first/second/third-party audits; audit as evidence-based, not fault-finding), corrective vs. preventive action and the CAPA loop. Know what each ISO 9001 clause requires at the level of "which clause covers design controls" (8.3) and "which covers nonconforming outputs" (8.7).

Domain III — Product, Process, and Service Design (~14%)

Design inputs/outputs/reviews/verification/validation/changes (the ISO 9001 8.3 stage gates — the exam maps questions to these), QFD/House of Quality (customer whats → technical hows; the roof is the correlation matrix of hows — the single most-tested QFD fact), FMEA: Severity × Occurrence × Detection = RPN (each 1–10; RPN prioritizes but the exam tests that high Severity alone can mandate action regardless of RPN), design FMEA vs. process FMEA, tolerance design and Taguchi: quality loss L = k(y − T)² — loss grows with the square of deviation from target, which is why Taguchi targets nominal-the-best rather than just "within spec." Reliability basics (MTBF, bathtub curve) and maintainability appear here at the conceptual level.

Domain IV — Product and Process Control (~16%)

Acceptance sampling: AQL (acceptable quality level — producer's risk α point) vs. LTPD/LOT (lot tolerance — consumer's risk β point); the OC curve plots P(accept) vs. lot fraction defective p. Derivation: for a single-sampling plan (n, c), P(accept|p) = Σ_{x=0}^{c} C(n,x) p^x (1−p)^{n−x} (binomial; hypergeometric for small lots without replacement — the exam tests knowing which applies: binomial when n/N ≤ 0.1). Producer's risk α = 1 − P(accept|AQL); consumer's risk β = P(accept|LTPD). ANSI/ASQ Z1.4 (attributes) and Z1.9 (variables): know how to read the tables (code letter from lot size → sample size; AQL → accept/reject numbers), switching rules (normal → tightened after 2 of 5 lots rejected).

Statistical process control: control charts distinguish common-cause from special-cause variation. X-bar & R (subgroups, variables data): UCL/LCL = x̿ ± A₂R̄ — derivation: A₂ = 3/(d₂√n), embedding the ±3σ limits scaled by subgroup size. Individuals & moving range for n=1. p-chart (fraction defective): p̄ ± 3√(p̄(1−p̄)/n); c-chart (defects per unit): c̄ ± 3√c̄; u-chart for varying sample sizes. Western Electric rules (a point beyond 3σ; 2 of 3 beyond 2σ; 4 of 5 beyond 1σ; 8 on one side of center) — the exam tests pattern recognition, not just limit math.

Process capability: Cp = (USL − LSL) / 6σ — potential capability, assumes centered. Cpk = min[(USL − μ)/3σ, (μ − LSL)/3σ] — actual, penalizes off-center. The exam's core trap: Cp = 2.0 with Cpk = 0.8 means a capable process run off-center — fix centering, not variation. Pp/Ppk use overall σ (long-term) vs. Cp/Cpk's within-subgroup σ — Ppk ≤ Cpk always. Capability requires a stable (in-control) process first; computing Cpk on an out-of-control process is the classic error.

Measurement systems: Gage R&R — %GRR = 100 × σ_measurement / σ_total; <10% acceptable, 10–30% conditional, >30% unacceptable. Resolution: the ndc (number of distinct categories) ≥ 5. Calibration traceability to NIST/SI.

Domain V — Continuous Improvement (~17%)

The 7 QC tools (Pareto, fishbone, check sheet, histogram, scatter, flowchart, control chart), 7 management tools, PDCA/Deming cycle, DMAIC vs. PDCA (DMAIC is the Six Sigma project framework; know when each applies), root-cause tools (5 Whys — stops at process causes, not people; fault tree), lean wastes (TIMWOOD+talent), kaizen vs. breakthrough, benchmarking, and quality awards (Baldrige criteria — leadership, strategy, customers, measurement, workforce, operations, results).

Domain VI — Quantitative Methods and Tools (~23%: the exam's center of gravity)

Probability distributions: binomial (defectives in n), Poisson (defects per unit, λ = np for rare events), normal (the workhorse — 68/95/99.7 rule), exponential (time between failures, memoryless), Weibull (reliability — see CRE guide for depth), hypergeometric (sampling without replacement), t/chi-square/F (sampling distributions for inference).

Hypothesis testing — the full ritual the exam demands: state H₀/Hₐ → choose α → select test statistic → compute → compare to critical value (or p-value) → conclude in plain language ("there is insufficient evidence the means differ"). Type I error (α: reject a true H₀ — producer's risk) vs. Type II (β: fail to reject a false H₀ — consumer's risk); power = 1 − β. z-test (σ known), t-test (σ unknown), paired t (before/after — the exam tests recognizing pairing), chi-square (variance, goodness-of-fit, contingency tables), F-test (compare two variances), ANOVA (compare 3+ means: F = MS_between/MS_within). One-sided vs. two-sided — the exam punishes wrong-tail choices.

Regression and correlation: least squares minimizes Σ residuals²; r measures linear association (−1 to 1); r² is fraction of variation explained. The exam's trap: correlation ≠ causation, and extrapolation beyond the data range is unjustified.

DOE basics: 2^k factorials — main effects and interactions from high/low runs; interaction means the effect of A depends on B's level (non-parallel lines on the interaction plot). Randomization, replication, blocking — know which combats what (blocking isolates nuisance variation; replication estimates pure error).

Domain VII — Risk Management (~9%)

Risk = probability × severity (the FMEA RPN is one operationalization). Risk registers, risk matrices, mitigation hierarchy (avoid → reduce → transfer → accept), business continuity basics, and the link to QMS risk-based thinking (ISO 9001 6.1). The exam tests that risk management is proactive and documented, not reactive firefighting.

3. Key equations

4. Worked examples

Example 1 — Capability. USL = 10.5, LSL = 9.5, μ = 10.1, σ = 0.15. Cp = 1.0/0.9 = 1.11. Cpk = min[(10.5−10.1)/0.45, (10.1−9.5)/0.45] = min[0.89, 1.33] = 0.89. Verdict: marginal and off-center high — recenter the process (Cp says the spread fits; Cpk says it's misplaced). Expected nonconforming ≈ 2 × Φ(−3×0.89) ≈ 0.76%.

Example 2 — OC curve point. Plan n = 50, c = 1. At p = 0.02: P(accept) = (0.98)^50 + 50(0.02)(0.98)^49 = 0.364 + 0.372 = 0.736. So a 2%-defective lot passes 74% of the time — if that's the LTPD point, consumer's risk β = 0.74, far too high; tighten (raise n or lower c).

Example 3 — X-bar & R limits. 20 subgroups, n = 5, x̿ = 50.0, R̄ = 4.0. A₂ (n=5) = 0.577. UCL = 50 + 0.577×4 = 52.31; LCL = 47.69. A subgroup mean of 52.8 → out of control, investigate (special cause), do not "adjust the process into spec."

Example 4 — Two-sample t. Supplier A: n=25, x̄=100.2, s=1.5; B: n=25, x̄=99.4, s=1.6. Pooled s ≈ 1.55; SE = 1.55√(2/25) = 0.438. t = 0.8/0.438 = 1.83; df = 48; two-sided critical at α=0.05 ≈ 2.01 → fail to reject H₀. Conclusion: insufficient evidence of a mean difference (not "the suppliers are equal" — the exam punishes that phrasing).

Example 5 — Gage R&R. σ_repeatability = 0.8, σ_reproducibility = 0.6, σ_part = 5.0. σ_measurement = √(0.64+0.36) = 1.0; σ_total = √(1+25) = 5.099. %GRR = 100×1.0/5.099 = 19.6% → conditionally acceptable; ndc = 1.41×5.0/1.0 = 7.05 → ≥5, adequate resolution. Improve reproducibility (operator training) before trusting the gage for capability studies.

5. Standards/regulations cross-reference

• ISO 9001:2026 — Domain II's backbone; know the clause structure and the audit approach to each clause.
• IATF 16949:2016 — automotive overlay: adds APQP/PPAP timing, control plans, MSA, and the "shall"s that go beyond 9001. CQE holders in automotive live here.
• AS9100D — aerospace overlay: counterfeit-parts prevention, configuration management rigor, risk emphasis. Know how it differs from base 9001.
• AIAG core tools — APQP, PPAP, FMEA (AIAG-VDA edition), MSA, SPC, Control Plan. The exam-adjacent body of practice for Domains III–IV.
• ANSI/ASQ Z1.4 / Z1.9 — sampling standards for Domain IV table-lookup questions.
• FDA 21 CFR Part 820 (QMSR) and EU MDR 2017/745 — regulated-industry overlays: design controls, CAPA, complaint handling. A CQE in med devices must know these cold.

6. Pitfalls and trap patterns

• Cp vs. Cpk. High Cp + low Cpk = off-center, not incapable. This single distinction is worth several questions.
• Control vs. specification limits. Control limits come from the process (±3σ); spec limits come from the customer. Never compare a control chart to spec limits.
• Stability before capability. Cpk on an out-of-control process is meaningless — the exam will offer it as a distractor calculation.
• "Accept H₀" language. You fail to reject H₀; you never prove the null. The exam tests the phrasing.
• RPN worship. RPN ranks; it doesn't decide. High severity with low RPN can still require action — the exam's favorite FMEA trap.
• Binomial vs. hypergeometric. n/N > 0.1 (destructive testing, small lots) → hypergeometric; the exam gives you lot size for a reason.

7. Practice problems (answers at end)

1. USL=20, LSL=12, μ=15.5, σ=1.0. Compute Cp and Cpk; diagnose.
2. Sampling plan n=80, c=2. Estimate P(accept) at p=0.01 using Poisson approximation.
3. p-chart: 25 subgroups of n=100, total defectives 150. Compute center line and limits. A subgroup has 12 defectives — in control?
4. Before/after training scores (paired, n=12): mean difference d̄=4.0, s_d=5.0. Test at α=0.05 whether training helped (one-sided).
5. A process has Cp=1.67, Cpk=1.67. What does the equality tell you? What fraction nonconforming (approx)?
6. %GRR = 35%, ndc = 3. Acceptable? What two actions?
7. Explain in one sentence each: producer's risk, consumer's risk, and which OC-curve point defines each.

Answers: 1. Cp = 8/6 = 1.33; Cpk = min[(20−15.5)/3, (15.5−12)/3] = min[1.5, 1.17] = 1.17 — capable but slightly off-center low. 2. λ = 0.8; P = e^−0.8(1 + 0.8 + 0.32) = 0.449×2.12 = 0.952. 3. p̄ = 0.06; limits 0.06 ± 3√(0.06×0.94/100) = 0.06 ± 0.0712 → UCL 0.131, LCL 0 (truncate). 12/100 = 0.12 < 0.131 → in control. 4. t = 4/(5/√12) = 2.77; df=11, one-sided critical 1.796 → reject H₀; evidence training helped. 5. Perfectly centered; ≈ 2×Φ(−5) ≈ 0.00006% (≈0.6 ppm). 6. No — %GRR >30% and ndc <5. Actions: fix the measurement system (calibration, procedure, training) before any capability or SPC work; increase resolution. 7. Producer's risk α = P(reject | lot at AQL) — the OC point at p=AQL; consumer's risk β = P(accept | lot at LTPD) — the OC point at p=LTPD.

Related credentials in this discipline

• CRE → asq-certified-reliability-engineer-cre-study-guide.md
• CQA → asq-certified-quality-auditor-cqa-study-guide.md
• Six Sigma Green/Black Belt → shared: ../shared/six-sigma-green-black-belt-study-guide.md

Original ETS reference writing, verified against public sources on 2026-09-30. Thin sourcing is flagged inline. Not legal advice — confirm current editions with the issuing body.