IASSC Black Belt Free Sample Questions: The 20 Free Sample Questions, Worked

These are the program’s 20 free sample questions — the same set the free sample quiz serves, presented here in readable form with the correct answer and a full explanation for each. The set spans all five DMAIC phases and is tiered the way the program’s samples are built: five easy, seven medium, five hard, and three stretch questions.

The 20 free sample questions, worked

1. Why does the hidden factory resist ordinary improvement prioritization in many plants?

A. Operators hide rework deliberately to protect overtime earnings, and only surveillance reveals the true scale of the activity

B. Its work is normalized inside standard routings and labor standards, so no report ever isolates it as waste available for elimination

C. Hidden-factory output is sold as seconds at a discount, making its economics look attractive in contribution reports, if the budget holder approves

D. Accounting standards forbid separating rework labor, so managers must estimate the hidden factory from scrap value alone

Correct: B. Rework embedded in standard times and routings becomes invisible by design; deliberate concealment, seconds sales, and accounting prohibitions misdescribe a structural visibility problem.

2. A centered process has limits exactly four overall standard deviations on each side of the mean. What is its Pp?

A. Pp is 2.67, because the full width of eight sigma divided by three gives the performance index for this process

B. Pp is 4.00, because the index reports the sigma distance from the mean to each limit without further adjustment

C. Pp is 0.75, because performance indices invert the ratio of tolerance width to the natural process spread used

D. Pp is 1.33, because the tolerance width of eight sigma divided by six sigma gives the performance index value

Correct: D. Pp = (USL-LSL)/(6 x overall sigma) = 8 sigma / 6 sigma = 1.33. Centering makes Ppk equal to Pp in this symmetric case.

3. Which phrasing correctly states what a Kruskal-Wallis rejection supports, absent any shape assumptions?

A. The group with the largest sample has the largest median, because rank sums grow automatically with the size of each group

B. Every pair of groups differs significantly, because omnibus rejection is defined as all pairwise differences holding at once

C. All group means are different from one another in a specified direction that the statistic prints alongside its p-value here

D. At least one group's distribution is stochastically different, with values from it tending to sit higher or lower than others

Correct: D. The defensible omnibus claim is distributional: at least one population tends to produce larger (or smaller) values than another. Pairwise and median-specific claims each demand further work or further assumptions.

4. A grower tests five two-level inputs on tomato yield in a single replicate of a full factorial. How large is the run list?

A. Thirty-two runs, from 2 multiplied by itself five times, covering each of the five factors' level pairings

B. Ten runs are enough, since five factors with two levels apiece contribute two runs per factor to the total

C. Sixty-four runs are required, which is the count belonging to six two-level factors rather than five here

D. Twenty runs result, because five factors at two levels produce four runs for every factor in the design

Correct: A. With k=5, runs = 2^5 = 32, the full set of level combinations for five two-level factors.

5. Which operators need control-plan training first when a plan launches across three shifts with limited trainer capacity?

A. The day shift only, because night and weekend shifts can copy day-shift practice from the posted charts

B. Supervisors exclusively, since floor operators absorb plan requirements by imitating their supervisor, leaving direct operator instruction for a later training round

C. The newest hires, because experienced operators can be assumed competent on any new plan by default

D. Those covering the checks and reactions for critical characteristics on every shift, so no shift ever runs a critical control with untrained cover here

Correct: D. Coverage logic governs sequencing: every shift needs certified capability on the highest-risk controls from the first day the plan is live.

6. A Black Belt presents savings using straight wages while finance insists on loaded rates including benefits and payroll taxes. Whose convention governs the charter?

A. Finance's loaded-rate convention, agreed and recorded at charter time, because benefit figures must be comparable across every project the portfolio reports

B. The Black Belt's wage figure, because improvement teams best understand what an hour in their process is actually worth

C. Whichever rate produces the larger benefit, since the charter's purpose is to win approval for worthwhile projects, because unwinding it later is somebody else's problem

D. A midpoint between the two rates, splitting the difference so neither party's convention dominates the calculation

Correct: A. Comparability across the portfolio requires one agreed rate convention, set with finance before work starts; local knowledge, approval-seeking, and midpoint splits sacrifice that consistency.

7. A Black Belt needs a quick check of a new fixture and runs a short study: 2 operators, 5 parts, 2 trials. The results look poor. What is the right weight to give this outcome?

A. Treat it as a screening signal: a poor result warrants a fuller study, but the small design's estimates carry wide uncertainty

B. Treat it as definitive, because gage estimates stabilize completely once total measurements reach twenty in a study

C. Ignore it, because studies below ten parts are incapable of detecting even a completely unusable measurement system

D. Treat the poor result as proof the fixture is worse than the old one, since small studies bias estimates upward

Correct: A. Small designs give imprecise variance estimates, so a poor screening result is a warning that justifies a proper study rather than a final verdict — but it is a real signal, not noise to be ignored.

8. A team tests a claim that a process mean exceeds a regulatory floor. The analyst writes the alternative as 'mean equals the floor' by mistake, computes a tiny p-value, and claims victory. What went wrong conceptually?

A. The boundary claim sat in the wrong seat: p-values flow from the default's distribution, so the scrambled setup voids the meaning

B. Nothing went wrong, because p-values are symmetric and can be computed from whichever statement the analyst finds convenient

C. The only slip is labeling, which a footnote can repair while preserving the victory claim and its supporting arithmetic

D. The floor should have been replaced by the sample mean in both statements, which would have made the victory claim rigorous

Correct: A. Every p-value is anchored to the default position's sampling distribution; swapping the statements changes the anchor, so the tiny value no longer testifies about the claim being advanced.

9. A predictor's variance inflation factor is 10, computed from an auxiliary R-squared of 0.90 when that predictor is regressed on the others. Verify and interpret.

A. VIF = 1/(1 – 0.90) = 10 confirms it: the predictor is heavily explained by its companions, inflating its coefficient's variance

B. VIF = 1 – 0.90 = 0.10, so the predictor is nearly independent and its standard error is comfortably small, and the line crews would work to it

C. VIF = 1/(1 – 0.90) squared = 100, meaning the coefficient is ten times less precise than reported

D. VIF = 0.90/0.10 = 9, close to the threshold but computed by dividing the wrong pair of quantities here

Correct: A. 1/(1 – 0.90) = 10 exactly; a VIF at that level says the coefficient's variance is inflated tenfold by overlap with other predictors.

10. A study of solder defects uses each printed board twice: once as its own experimental unit and again after rework as a second unit. Why does a reviewer object?

A. The objection is misplaced, because defect counts from the same board are actually more comparable than across boards

B. The reworked board is not an independent unit, so treating it as one understates error and overstates the evidence

C. Boards may be reused freely provided the run order is shuffled, which restores independence between the two passes

D. Rework is objectionable only because it doubles the study cost without adding any new factor combinations

Correct: B. Experimental units must be independent; a reworked board carries its history with it, so its second result is correlated with the first.

11. Phase I data for a new chart included a weekend when a substitute resin was used. Limits computed from all the data are wide, and phase II rarely signals. What went wrong?

A. Phase I data must contain no special causes whatsoever, so the chart has to be discarded and the process requalified from scratch after that resin weekend.

B. Wide limits are correct here because phase II limits are required to widen every time phase I grows longer

C. The substitute resin weekend should have been charted as its own subgroup, which alone would have fixed the limits

D. The contaminated phase I sample inflated the variation estimate, so the trial limits are too generous and real departures hide inside them

Correct: D. Phase I exists to estimate the stable process; leaving a known disturbed period in the estimate bakes its extra variation into every future limit.

12. A Black Belt reviewing a supplier's CuSum finds the target was set to the average of the last month, which included a known bad week. What is the core defect?

A. The defect is that targets must be specification midpoints, and using an observed monthly average is prohibited in CuSum work even when that week is excluded.

B. The target absorbed the bad week's bias, so the chart now treats a shifted level as normal and accumulates evidence against the wrong centre

C. A one-month baseline is too short by rule; CuSum targets require a minimum of twelve months of history before use

D. Including the bad week only affects the first plotted point, after which the statistic corrects itself automatically

Correct: B. A CuSum measures departure from its target; a contaminated target shifts the zero point and the chart will defend the wrong level as if it were the aim.

13. A fishbone session for invoice errors produces 60 candidate causes. The sponsor asks which three matter. The Black Belt's next step should be:

A. Ask each participant to defend their own suggestions, then keep the three argued most persuasively in the room

B. Keep all sixty in scope until Improve, because narrowing causes early risks discarding the true root cause unknowingly, since missing it costs far more

C. Pareto the candidates against occurrence data — error logs, stratification by type and source — to shortlist causes for verification

D. Select the three causes cheapest to fix, since early wins matter more than causal importance at this stage of DMAIC

Correct: C. Brainstormed volume must be reduced with evidence: error records stratified by type, shift, or source show where occurrences concentrate, pointing verification at plausible major causes.

14. To estimate defect rates across three plants with very different volumes, a team samples proportionally within each plant. Which method and benefit?

A. Cluster sampling, chosen because sampling whole plants is cheaper than sampling within every plant in the study

B. Stratified sampling, ensuring each plant is represented in proportion and plant-level comparisons remain possible

C. Simple random sampling across the pooled output, which automatically guarantees the proportional plant mix desired

D. Systematic sampling from the combined shipping stream, which represents plants in proportion to their shipment timing

Correct: B. Dividing the population into plant strata and sampling within each is stratified sampling; it guarantees representation and supports per-stratum estimates.

15. Two candidate fits for the same data differ only in that one was estimated on 18 points and the other on 60 points spanning the same x range. Why will the larger study's slope interval usually be tighter?

A. More observations shrink the slope's standard error, since residual noise averages down while the x spread stays comparable

B. Larger samples raise the critical t value sharply, and that increase is what narrows intervals in practice

C. Sixty points guarantee a higher R-squared, and higher explained variation is what mechanically tightens slope intervals

D. The interval narrows because the slope estimate itself must grow with sample size, dwarfing its standard error

Correct: A. Slope standard error falls roughly with the square root of n; more points over the same range pin the line's tilt down harder.

16. A fractional screen estimates a contrast labelled A at 14 units, where A is aliased with BC. Later work shows the BC interaction is actually 10 units. What follows?

A. A's effect must be 24, because aliased effects always add constructively in fractional designs

B. The later finding invalidates the screen retroactively, so all its other contrasts must also be discarded

C. A remains 14 regardless, because aliasing affects only the interpretation of inactive effects

D. Much of the 14 may belong to BC, so A's own contribution could be as small as roughly 4 units

Correct: D. The contrast estimated A + BC = 14; if BC = 10, the portion attributable to A is about 14 – 10 = 4.

17. A Black Belt sees a histogram of bolt torque with a sharp cutoff at the lower specification and a normal-looking body above it. What likely happened?

A. The process naturally produces truncated data, because torque distributions terminate at specification limits physically

B. Parts below the limit were screened out before the histogram, so the data describe inspected output, not the process itself

C. The gauge cannot read below the limit, which proves the measurement system needs a wider range instrument installed

D. The cutoff proves the process is incapable, because capable processes never show abrupt edges in their histograms

Correct: B. A hard edge at the spec is the signature of prior sorting: the process distribution was amputated by inspection, and capability read from it would be flattered.

18. A call center's handle-time data fail every normality test, with p-values below 0.01 in samples of 30, 80, and 200 calls drawn weeks apart. A manager suggests collecting a much larger sample so the data themselves will become normal. Why is that reasoning wrong?

A. Larger samples do become normal, but only after the count passes one thousand calls in a single uninterrupted sample

B. The reasoning is right for handle times because service data always converge to a bell shape as volume grows

C. Larger samples reveal the true skewed shape more clearly; the central limit effect applies to sample means, not to the raw call times

D. The tests fail only because of the small samples, and at 200 calls a p-value below 0.01 carries no information at all, across all shifts

Correct: C. Collecting more calls sharpens the picture of a skewed population rather than changing it; the central limit theorem describes averages of samples, so raw handle times keep their skew no matter how many are gathered.

19. A Black Belt sizes a two-sample t study to detect a 0.4-standard-deviation mean difference with power 0.80 at alpha 0.05. A planning table suggests about 100 observations per group. The client offers 40 per group 'to start.' What consequence should be spelled out?

A. Power stays at 0.80 because the planning table's figures embed a safety factor of more than two that real studies never actually consume

B. At 40 per group, power for the targeted difference sinks toward roughly 0.40, so a null outcome leaves the question genuinely open

C. The smaller study becomes a pilot whose significant results are trustworthy but whose null results are even more decisive than the full study's

D. Reducing group size raises power in two-sample designs, because fewer observations tighten the sampling distribution of the difference

Correct: B. Power curves are unforgiving at small standardized effects: 0.4 sigma needs roughly 100 per group for 80% power, and 40 per group buys little better than a coin flip weighted toward missing.

20. A molding plant wants its andon escalation to be economically rational. Pulls currently stop the whole line for any fault, including faults one operator can clear in seconds. What refinement preserves the system?

A. Abolish line stops entirely and log every fault for weekly review, since stop costs always exceed defect costs in high-volume molding operations

B. Tier the response: operator-clearable faults get a brief signal and local fix, while faults beyond a defined time or scope escalate to a full stop

C. Charge the cost of each stop to the pulling operator's department budget, creating a financial incentive to reserve pulls for severe faults only

D. Require supervisor pre-approval before any pull, ensuring only economically justified stops interrupt the line during scheduled production time

Correct: B. Tiered escalation keeps fast local response cheap and reserves expensive full stops for faults that need them. Pre-approval or abolition would destroy the signal's immediacy.

A note on format. The real IASSC exam mixes multiple-choice with true/false questions, while PassProfessor’s bank trains you with four-option multiple-choice questions throughout — the same knowledge, in a format that forces you to discriminate between close alternatives. No worked set on this page reproduces the real exam’s item-type mix.

These questions are drawn from the same bank as the full program. Start with the 20 free sample questions (also presented readably on the free sample questions page), read the IASSC Black Belt exam guide, then see the full IASSC Lean Six Sigma Black Belt Exam Prep program — 1,500 original questions built at the official syllabus weights (Define 13% · Measure 21% · Analyze 27% · Improve 17% · Control 22%), with an explanation for every answer, for $59 with 6 months of access.

Disclaimer

PassProfessor is an independent exam-preparation provider. This program is exam preparation only: it is not the real exam, it provides no education credit, and completing it does not by itself qualify you for any license or certification. The real exam is run by its own exam body — you register for it, pay its fees, and sit it through that body, not through PassProfessor. All practice questions in this program are original; they are not actual exam questions.