Overview
Measure establishes a trustworthy baseline. Process characteristics covers flow metrics (lead time, cycle time, takt time, work in process, throughput, process cycle efficiency) and analysis tools such as value stream maps, process maps and SIPOC. Data collection covers data types and measurement scales, sampling methods and data collection plans with operational definitions. Measurement systems covers measurement system analysis (MSA): gage repeatability and reproducibility, bias, linearity, stability and attribute agreement. Basic statistics covers descriptive measures, the central limit theorem and graphical methods. Probability covers basic rules and the common distributions: normal, binomial, Poisson, hypergeometric, exponential, chi-square, t and F. Process capability covers Cp, Cpk, Pp, Ppk and Cpm, short-term versus long-term variation, non-normal data, and attribute measures such as DPMO and sigma level.
Candidates must compute and interpret these numbers and know which one fits the data. Measure validates Domain 4's CTQs and feeds Domain 6's hypothesis tests and models.
How it's tested
Most items are calculations or interpretation of outputs: compute DPMO, rolled throughput yield, Cp or Cpk, a probability, or a standard error; read a gage R&R table; or choose a distribution or sampling method. Scenario items test sequence, such as validating the measurement system before assessing capability.
-
Compute the metric
Data are given and you calculate DPMO, RTY, takt time, Cpk, a z-score probability, or a binomial or Poisson probability.
-
Interpret the output
A gage R&R, capability or descriptive statistics output is shown and you judge whether the measurement system or process is acceptable.
-
Choose the right method
A data situation is described and you pick the measurement scale, sampling plan, distribution or capability index that fits.
A process shows Cp of 1.8 and Cpk of 0.9, and the options include reducing variation, centering the process, or tightening specifications. Cp compares spread to tolerance and ignores location, while Cpk includes location. A large gap between them means the spread would fit but the mean sits near one limit, so the best action is to center the process, not reduce variation.
Subsections 6
Official subsections from the ASQ outline, in outline order.
5.1 Process characteristicsProcess flow metrics and process analysis tools.
- What it means
- This covers measures of flow, such as lead time, cycle time, takt time, work in process (WIP), throughput, touch time and process cycle efficiency, and tools that document a process, such as SIPOC, process maps, swim-lane maps, value stream maps and spaghetti diagrams.
- What to understand
- Little's law links the flow metrics: WIP = throughput × lead time, so cutting WIP cuts lead time at constant throughput. Takt time is available time divided by customer demand and sets the pace the process must match. Process cycle efficiency is value-added time divided by total lead time.
- How it shows up
- Flow data are given and you calculate takt time, lead time or efficiency, or you choose the mapping tool that reveals a specific problem such as excessive handoffs.
- Common confusion
- Treating cycle time and takt time as the same; cycle time is how long the process takes per unit, while takt is how often the customer requires a unit.
5.2 Data collectionData types, measurement scales, sampling methods and collection plans.
- What it means
- This covers continuous (variables) versus discrete (attribute) data, the nominal, ordinal, interval and ratio scales, sampling methods (simple random, stratified, systematic, cluster, rational subgrouping), and data collection plans built on operational definitions and check sheets.
- What to understand
- Continuous data carry more information per observation, so they need smaller samples than attribute data. The scale determines valid statistics: medians suit ordinal data, while means need interval or ratio data. An operational definition states exactly what to measure and how, so different collectors get the same result.
- How it shows up
- A data situation is described and you identify its scale, choose a sampling method that represents subgroups, or fix a collection plan that lacks an operational definition.
- Common confusion
- Calling survey ratings on a 1 to 5 scale interval data and averaging them without caution; they are ordinal.
5.3 Measurement systemsMeasurement system analysis: gage R&R, bias, linearity, stability and attribute agreement.
- What it means
- Measurement system analysis (MSA) quantifies how much observed variation comes from the measurement process. Precision components are repeatability (equipment variation) and reproducibility (appraiser variation); accuracy components are bias, linearity and stability. Attribute agreement analysis does the same for pass/fail judgments.
- What to understand
- Common guidance (AIAG) treats %GRR under 10% of study variation or tolerance as acceptable, 10% to 30% as possibly acceptable depending on the application, and over 30% as unacceptable, with number of distinct categories (ndc) of at least 5. High repeatability points to the gage; high reproducibility points to appraiser method or training.
- How it shows up
- A gage R&R output is shown and you decide whether the system is acceptable and what to fix, or a scenario asks what must happen before a capability study.
- Common confusion
- Confusing accuracy (bias, closeness to the true value) with precision (repeatability and reproducibility, closeness of repeated readings to each other).
5.4 Basic statisticsDescriptive statistics, the central limit theorem and graphical methods.
- What it means
- This covers measures of central tendency (mean, median, mode) and dispersion (range, variance, standard deviation), the central limit theorem, and graphical tools such as histograms, box plots, scatter diagrams, run charts, stem-and-leaf plots and normal probability plots.
- What to understand
- The central limit theorem says sample means are approximately normal for large enough samples regardless of the population's shape, with standard error σ/√n. Skewed data pull the mean toward the tail, so the median better represents the center. Box plots show median, quartiles and outliers at a glance.
- How it shows up
- A data set or plot is described and you compute a statistic, interpret skew or outliers, or calculate a standard error.
- Common confusion
- Believing the central limit theorem makes individual observations normal; it applies to the distribution of sample means.
5.5 ProbabilityProbability rules and common distributions.
- What it means
- This covers basic concepts (independence, mutually exclusive events, addition and multiplication rules, conditional probability, permutations and combinations) and distributions: normal, binomial, Poisson, hypergeometric, exponential, lognormal, Weibull, chi-square, t and F.
- What to understand
- Choose the distribution by the data: binomial for a count of defectives in n trials with constant p, Poisson for a count of defects per unit of area or time, hypergeometric for sampling without replacement from a small finite lot, exponential for time between failures at a constant rate. Chi-square, t and F are sampling distributions used in tests.
- How it shows up
- A scenario gives counts or rates and you choose the distribution and compute a probability, or apply the addition or multiplication rule.
- Common confusion
- Using binomial for defects per unit, where one unit can have several defects; that situation is Poisson.
5.6 Process capabilityCapability and performance indices, short- and long-term variation, non-normal and attribute capability.
- What it means
- Process capability compares the voice of the process with the voice of the customer. Capability indices Cp and Cpk use within-subgroup (short-term) variation; performance indices Pp and Ppk use overall (long-term) variation; Cpm penalizes distance from target. Attribute capability uses DPMO, yield and sigma level.
- What to understand
- Capability is meaningful only for a stable process, so confirm statistical control first. Cp = (USL − LSL) / 6σ_within; Cpk = min[(USL − μ) / 3σ_within, (μ − LSL) / 3σ_within]. Non-normal data need a transformation (Box-Cox, Johnson) or a fitted non-normal distribution before indices are valid.
- How it shows up
- Specification limits, mean and standard deviation are given and you compute an index, or a pair of indices is given and you diagnose centering versus spread.
- Common confusion
- Reporting capability for an out-of-control process, or mixing up Cpk (within) with Ppk (overall).
High-yield concepts
The terms, rules and formulas this domain turns into questions most often.
-
DPU, DPO and DPMO
DPU = defects / units; DPO = defects / (units × opportunities per unit); DPMO = DPO × 1,000,000. For 25 defects in 500 units with 4 opportunities each, DPMO = 25 / 2,000 × 10⁶ = 12,500.
- Remember
- Opportunities must be defined consistently, or DPMO can be inflated by counting trivial opportunities.
- Tested as
- A standard calculation item, often followed by converting to sigma level.
- Trap
- Dividing by units only, which gives DPU, not DPO.
-
Sigma level
Sigma level is the z-value for the process yield, conventionally reported short term as Z_ST = Z_LT + 1.5. Reference points: 3σ ≈ 66,807 DPMO, 4σ ≈ 6,210, 5σ ≈ 233, 6σ ≈ 3.4.
- Remember
- 12,500 DPMO gives Z_LT ≈ 2.24 and a short-term sigma level of about 3.74.
- Tested as
- Conversion items between DPMO, yield and sigma level.
- Trap
- Forgetting whether the table or question includes the 1.5 shift.
-
First-pass yield and rolled throughput yield
First-pass yield (FPY) is the fraction of units passing a step without rework or scrap; rolled throughput yield RTY = FPY₁ × FPY₂ × … × FPYₙ. Steps of 0.95, 0.90 and 0.98 give RTY ≈ 0.838.
- Remember
- RTY can be estimated from defects per unit as RTY ≈ e^(−DPU).
- Tested as
- Calculation items show a multi-step process and ask for overall yield.
- Trap
- Using final yield after rework, which hides the hidden factory that RTY exposes.
-
Cp and Cpk
Cp = (USL − LSL) / 6σ measures potential capability if centered; Cpk = min[(USL − μ) / 3σ, (μ − LSL) / 3σ] measures actual capability including location, both using within-subgroup σ.
- Remember
- Cpk ≤ Cp always; equality means the process is centered. A common minimum target is Cpk ≥ 1.33.
- Tested as
- The most frequent Measure calculation and the basis for centering-versus-spread diagnosis.
- Trap
- Using the distance to the nearer limit divided by 6σ instead of 3σ.
-
Pp, Ppk and Cpm
Pp and Ppk use the same formulas as Cp and Cpk but with overall (long-term) standard deviation; Cpm = (USL − LSL) / [6√(σ² + (μ − T)²)] penalizes deviation from target T.
- Remember
- Ppk below Cpk signals shifts and drifts between subgroups over time.
- Tested as
- Interpretation items compare within and overall indices.
- Trap
- Treating Ppk and Cpk as interchangeable.
-
Little's law, takt time and process cycle efficiency
Lead time = WIP / throughput; takt time = available production time / customer demand; process cycle efficiency (PCE) = value-added time / total lead time.
- Remember
- With 450 available minutes and demand of 90 units, takt is 5 minutes per unit.
- Tested as
- Flow calculations in process characteristics questions.
- Trap
- Calculating takt from the process's capacity rather than customer demand.
-
Gage R&R acceptance
%GRR compares measurement variation with study variation or tolerance; under 10% is generally acceptable, 10% to 30% may be acceptable depending on use and cost, and above 30% is unacceptable. ndc should be at least 5.
- Remember
- Repeatability is equipment variation; reproducibility is appraiser variation.
- Tested as
- Output-interpretation items ask whether to trust the gage and what to fix.
- Trap
- Retraining appraisers when repeatability, not reproducibility, dominates.
-
Accuracy terms: bias, linearity, stability
Bias is the difference between the average measurement and a reference value; linearity is how bias changes across the measurement range; stability is how bias changes over time.
- Remember
- These are accuracy properties, separate from precision (R&R).
- Tested as
- Definition and scenario items about calibration and drift.
- Trap
- Calling a gage that drifts over weeks a linearity problem.
-
Measurement scales
Nominal (categories), ordinal (ranked categories), interval (equal spacing, no true zero, such as °C), ratio (equal spacing with true zero, such as weight).
- Remember
- Only ratio data support statements like twice as much.
- Tested as
- Classification items and questions on valid statistics for a scale.
- Trap
- Classifying temperature in Celsius as ratio data.
-
Central limit theorem and standard error
Sample means are approximately normally distributed for sufficiently large n, with mean μ and standard error σ/√n.
- Remember
- Quadrupling the sample size halves the standard error.
- Tested as
- Supports control limits for averages and confidence intervals; appears in calculations.
- Trap
- Applying it to individual values instead of sample means.
-
Normal z-score
z = (x − μ) / σ converts a value to standard units for table lookup; about 68.27%, 95.45% and 99.73% of values fall within 1, 2 and 3 standard deviations.
- Remember
- One-tail area beyond z = 3 is about 0.135%.
- Tested as
- Probability and percent-out-of-specification calculations.
- Trap
- Using the two-tail table value when only one specification limit applies.
-
Binomial and Poisson
Binomial P(x) = C(n, x) pˣ(1 − p)ⁿ⁻ˣ for x defectives in n trials; Poisson P(x) = e^(−λ) λˣ / x! for x defects when the average is λ.
- Remember
- Poisson approximates binomial when n is large and p is small, with λ = np.
- Tested as
- Distribution selection and probability calculations.
- Trap
- Using binomial for defects per unit of area or time.
-
Probability rules
Addition: P(A or B) = P(A) + P(B) − P(A and B). Multiplication for independent events: P(A and B) = P(A) × P(B). Conditional: P(A|B) = P(A and B) / P(B).
- Remember
- Mutually exclusive events have P(A and B) = 0; they are not independent.
- Tested as
- Short probability calculations and definition items.
- Trap
- Treating mutually exclusive and independent as the same idea.
-
Sampling methods
Simple random gives every unit equal chance; stratified samples within known subgroups; systematic takes every kth unit; cluster samples whole groups; rational subgrouping groups units made under similar conditions.
- Remember
- Systematic sampling can be biased if the process has a cycle that matches the interval.
- Tested as
- Method-selection questions for representative data.
- Trap
- Using simple random sampling when known strata such as shifts must be represented.
Side by side
Concepts candidates mix up, compared on the point that decides the answer.
Cp vs. Cpk vs. Pp vs. Ppk
| Concept | Meaning | Key difference | Common trap |
|---|---|---|---|
| Cp | Tolerance width divided by 6 within-subgroup σ | Ignores where the mean sits | Reporting a high Cp for an off-center process as capable |
| Cpk | Distance from mean to nearer limit divided by 3 within σ | Includes centering; short-term | Dividing by 6σ instead of 3σ |
| Pp | Tolerance width divided by 6 overall σ | Uses long-term variation including shifts | Confusing with Cp when subgroups drift |
| Ppk | Distance to nearer limit divided by 3 overall σ | Actual long-term performance including location | Expecting Ppk to exceed Cpk |
Repeatability vs. reproducibility vs. bias vs. linearity vs. stability
| Concept | Meaning | Key difference | Common trap |
|---|---|---|---|
| Repeatability | Variation when one appraiser measures the same part repeatedly | Equipment variation (precision) | Fixed by training appraisers |
| Reproducibility | Variation between appraisers measuring the same parts | Appraiser variation (precision) | Fixed by replacing the gage |
| Bias | Average measurement minus reference value | Accuracy at one point | Confused with precision |
| Linearity | Change in bias across the operating range | Accuracy across size | Confused with drift over time |
| Stability | Change in bias over time | Accuracy across time | Confused with linearity |
Binomial vs. Poisson vs. hypergeometric vs. normal
| Concept | Meaning | Key difference | Common trap |
|---|---|---|---|
| Binomial | Number of defectives in n independent trials, constant p | Each unit is pass or fail | Used for several defects on one unit |
| Poisson | Number of defects per unit, area or time | Count with no fixed upper limit | Used for defectives with known n |
| Hypergeometric | Defectives in a sample drawn without replacement from a small lot | p changes as units are drawn | Replaced by binomial when the lot is small |
| Normal | Continuous symmetric distribution of measurements | Described by μ and σ | Assumed for skewed data such as cycle times |
Exam traps
Each trap is an answer that looks right. Learn the clue that should change your mind.
-
Capability before control
- What candidates pick
- Calculate Cpk immediately from baseline data.
- Why it looks right
- Capability is the metric the customer asked about.
- Why it's wrong
- Capability indices predict future performance only for a stable process; an out-of-control process has no single predictable distribution.
Clue to watch for The scenario mentions points beyond control limits or shifts in the baseline.
-
Trusting the gage
- What candidates pick
- Start root-cause analysis on baseline data without MSA.
- Why it looks right
- Data already exist and the project is on a deadline.
- Why it's wrong
- If measurement variation is large, apparent process differences may be measurement noise; MSA comes first.
Clue to watch for Operators disagree on readings or the gage has not been studied.
-
Wrong fix for R&R
- What candidates pick
- Retrain inspectors when repeatability is 80% of total GRR.
- Why it looks right
- People are the usual suspect in measurement problems.
- Why it's wrong
- Repeatability is equipment variation; the fix is the gage, fixture or method, not appraiser training.
Clue to watch for The GRR breakdown shows repeatability much larger than reproducibility.
-
Cp tells the whole story
- What candidates pick
- Declare the process capable because Cp is 2.0.
- Why it looks right
- Cp above 1.33 looks excellent.
- Why it's wrong
- Cp ignores location; if Cpk is low, much output may be outside one limit.
Clue to watch for The mean is noticeably closer to one specification limit, or Cpk is given separately.
-
Averaging ordinal ratings
- What candidates pick
- Report the mean of 1 to 5 satisfaction ratings as a precise measure.
- Why it looks right
- Numbers invite averaging.
- Why it's wrong
- Ordinal scales have unequal intervals; the median and frequency distribution are the appropriate summaries.
Clue to watch for Data are rankings or Likert-type ratings.
-
Final yield hides rework
- What candidates pick
- Report 99% yield for a process where many units are reworked before shipping.
- Why it looks right
- Almost everything eventually ships.
- Why it's wrong
- Rolled throughput yield counts first-pass success at each step and reveals the hidden factory of rework.
Clue to watch for The scenario mentions rework loops or repair stations.
Answer framework
Use this sequence for any Measure-phase calculation or scenario.
- Classify the data as continuous or attribute and identify its measurement scale before choosing any statistic or distribution.
- Confirm the measurement system is adequate with gage R&R or attribute agreement before trusting baseline data.
- Check process stability with a control chart, since capability indices and yield predictions are only meaningful for a stable, predictable process.
- Choose the right formula, within σ for Cp and Cpk, overall σ for Pp and Ppk, opportunities for DPMO, and multiply step yields for RTY.
- Interpret the result against the question: compare Cp with Cpk to separate centering from spread, and convert DPMO to sigma level only when asked.
Practice question
Want more like this? The CSSBB practice test has questions for every domain.
Checklist
0 of 14
Tick each one you can do without notes. Your ticks stay in this browser.
Study strategy
Because Measure carries 25 scored questions, plan the most study time here. Start with the logical sequence the exam tests: define the data, validate the measurement system, confirm stability, then assess capability. Keep that order in mind for every scenario.
Next, drill calculations in short daily sets: DPMO and sigma level with the 1.5 shift, RTY, takt time and Little's law, then Cp, Cpk, Pp and Ppk. Work each formula by hand and with your calculator until you can do it in under two minutes. Practice z-table lookups for one-sided and two-sided areas, and binomial and Poisson probabilities for small counts.
Then study MSA outputs. Learn to read a gage R&R table, decide acceptability with %GRR and ndc, and diagnose whether repeatability or reproducibility dominates. This area is commonly underestimated because it looks like a table-reading exercise, yet the follow-up action is what gets tested.
Build a formula page and tab it in your open-book reference, including a normal table and a DPMO-to-sigma table. Afterwards, practice mixed Measure sets under time pressure, then move to Analyze, where the probability distributions here become hypothesis tests and confidence intervals.
FAQ
How many questions come from Measure?
ASQ assigns 25 of the 150 scored questions to Measure, about 16.7% of the exam, the largest single section.
What is the difference between Cpk and Ppk?
Both use the distance from the mean to the nearer specification limit divided by three standard deviations. Cpk uses within-subgroup (short-term) variation; Ppk uses overall (long-term) variation, so Ppk is usually lower.
Why must MSA come before capability analysis?
Observed variation includes measurement variation. If the gage contributes a large share, capability indices understate the process and root-cause work may chase noise.
How do I convert DPMO to sigma level?
Convert DPMO to a yield, find the corresponding z-value from the normal table (long-term), then add 1.5 for the conventional short-term sigma level. Many references include a ready-made conversion table.
When should I use the Poisson distribution?
When counting defects per unit, area or time interval, where a unit can have more than one defect and there is no fixed number of trials.
What if my data are not normal?
Check for special causes first. If the process is stable but skewed, transform the data (for example Box-Cox) or fit an appropriate distribution such as Weibull before computing capability.
Which calculator should I bring?
ASQ restricts calculators to approved models. Check the current ASQ calculator policy before exam day and practice with the exact model you will use.
Other Quality Management certifications
Where the same subject shows up in other exams. Cards marked "Same subject" open the matching domain.
Test your Domain 5 knowledge
You have covered flow metrics, data collection, measurement system analysis, statistics, probability and capability. Test whether you can run these calculations accurately at exam pace.