CSSC MBB common mistakes
CSSC MBB · Understand the traps and how to avoid them.
Treating Cp and Cpk as interchangeable, reporting only Cp when the process mean is visibly off-centre from the specification midpoint.
Candidates memorise Cp = (USL−LSL)/6σ as 'the capability index' and stop there, forgetting that Cp assumes the process is centred. On exam items where the mean is shifted, Cp overstates capability and Cpk is the correct answer; the distractor set is built precisely around this confusion.
What to do instead
Before computing any index, check whether the process mean equals the specification midpoint. If it does not, compute Cpk = min[(USL−μ)/3σ, (μ−LSL)/3σ] and report that as the capability measure, using Cp only to quantify potential capability if the mean were centred. Practise deriving both from a given μ, σ, USL and LSL until the centring condition is automatic.
Choosing the wrong control chart type - e.g. an X-bar/R chart for attribute count data, or a p chart where a c chart is required - because the decision is made from the metric's name rather than its distribution.
The chart-selection decision tree is memorised as a list rather than as a set of questions about data type (variable vs attribute), subgroup size, and whether the attribute is a defect count per unit or a proportion. Under time pressure candidates pattern-match on words like 'proportion' or 'sample' and skip the underlying question.
What to do instead
Build a two-question habit: (1) Is the data variable (measured) or attribute (counted)? (2) If attribute, is it a proportion (p, np - needs a constant or varying sample size) or a count of defects per unit (c, u)? If variable, is subgroup size small (X-bar/R) or large (X-bar/S), or is it individual data (I-MR)? Apply the two questions to every chart item before reading the options.
Confusing special-cause with common-cause variation when interpreting a control chart, and reacting to a single point inside the limits as if it were a signal or dismissing a run of points as normal noise.
Candidates learn the three-sigma limits but not the supplementary run rules (e.g. eight consecutive points on one side of the centreline, six trending points, two of three beyond two sigma). They also conflate 'within limits' with 'in control' and 'out of limits' with 'process is bad', which is the opposite of how a stable process is judged.
What to do instead
Memorise the standard Western Electric / Nelson run rules and apply them as a checklist on every chart item: a single point beyond a limit, a run of eight on one side, six in a trend, two of three beyond two sigma, and fifteen points within one sigma. State explicitly that a process is in statistical control when only common-cause variation is present, regardless of whether it meets specification.
Computing DPMO or sigma level using the wrong defect opportunity count, or forgetting to convert between DPMO and Z (sigma level) via the normal table.
DPMO = (defects / (units × opportunities)) × 1,000,000 is simple arithmetic, but exam items deliberately vary whether the denominator is units, opportunities, or units × opportunities. Candidates also forget that a sigma level is the Z-value corresponding to the yield, and that the 1.5-sigma shift convention must be applied when the question asks for 'sigma level' in the Six Sigma sense.
What to do instead
Write the DPMO formula out with labelled terms before substituting: defects (count), units (count), opportunities per unit (count). Then convert yield to Z using the standard normal table, and add 1.5 only when the item explicitly uses the Six Sigma long-term shift convention. Do three worked examples with different opportunity counts until the units cancel correctly.
Mixing up the DMAIC phase deliverables - placing a control plan in Improve, a measurement system analysis in Analyse, or a Pareto in Define - when answering 'which phase does this tool belong to' items.
The tools are taught as a toolkit rather than as a phase-gated sequence, so candidates recall the tool but not its gate. Exam items exploit this by offering a plausible tool that belongs one phase earlier or later, and the candidate selects it because the tool itself is correct for the project.
What to do instead
Build a one-page phase map with the gate question for each phase: Define (what problem, what scope, what CTQ?), Measure (is the measurement system valid and what is baseline capability?), Analyse (what are the verified root causes?), Improve (what solution is validated and what is the new capability?), Control (what standard, owner and monitoring sustains the gain?). Attach each tool to the gate question it answers, then test yourself by naming the phase for twenty tools.
Treating a statistically significant hypothesis test as proof of practical significance, or failing to check the test's assumptions (normality, equal variance, independence) before selecting a parametric test.
Candidates focus on p-values and alpha/beta definitions but under-weight the assumption checks and effect size. Exam items often present a large-sample result with a tiny p-value and a trivial mean difference, or a non-normal small sample, and the correct answer is either 'not practically significant' or 'use a non-parametric alternative'.
What to do instead
For every hypothesis-test item, run a three-step check: (1) state H0 and Ha and the test's assumptions; (2) if assumptions fail, choose the non-parametric counterpart (Mann-Whitney for two independent groups, Wilcoxon signed-rank for paired, Kruskal-Wallis for k groups); (3) interpret the p-value alongside the effect size and confidence interval, not in isolation. Practise stating the conclusion in plain language that includes both statistical and practical significance.
Confusing the roles of the Design of Experiments stages - screening, characterisation, and optimisation - and misreading main effects versus interactions when interpreting an ANOVA output.
DOE is often learned as 'factorial designs' without the staged logic. Candidates cannot tell when a fractional factorial is appropriate (screening many factors) versus a response surface design (optimising a few), and they read a significant interaction as if it were a main effect, which changes the correct conclusion.
What to do instead
Learn the three stages with their purposes: screening (many factors, fractional factorial, resolution III/IV), characterisation (full factorial, resolution V), optimisation (response surface, central composite). When reading output, always check interaction terms first; if an interaction is significant, interpret the main effects only within the levels of the interacting factor. Work through one 2^3 factorial by hand and plot the interaction.
Misapplying FMEA severity, occurrence and detection ratings - especially reversing detection so that a high score means easy to detect - and computing RPN without considering criticality.
The three scales are memorised as numbers without their direction: severity 10 is worst, occurrence 10 is most frequent, and detection 10 is hardest to detect. Candidates reverse the detection scale and then rank the wrong failure mode as highest priority. They also treat RPN as the sole prioritisation rule, ignoring that a high-severity item with a moderate RPN can still demand action.
What to do instead
Write the three scales with their endpoints before rating: severity 1 (no effect) to 10 (hazard without warning); occurrence 1 (unlikely) to 10 (inevitable); detection 1 (certain to detect) to 10 (no detection control). Compute RPN = S × O × D, but rank by severity first, then RPN, and flag any item with severity 9–10 for mandatory action regardless of RPN.
Confusing the lean tools and their targets - applying 5S where poka-yoke is needed, or treating kanban as a scheduling system rather than a pull signal - when answering waste-elimination items.
Lean is learned as a vocabulary list (5S, kanban, kaizen, poka-yoke, SMED, value stream mapping) without the underlying question each tool answers. Exam items describe a specific waste (overproduction, waiting, defects, inventory) and the candidate picks a familiar tool rather than the one matched to that waste.
What to do instead
Map each tool to the waste it addresses: 5S to workplace organisation and motion waste; poka-yoke to defect prevention; SMED to changeover and waiting; kanban to overproduction and inventory; value stream mapping to identifying the seven wastes across the flow. For each item, first name the waste, then select the tool that eliminates it, and justify the match in one sentence.
Answering team and change-management items with generic 'communicate more' responses instead of the specific team-stage or stakeholder behaviour the item is testing.
The soft-skills section of the Body of Knowledge is skimmed because it feels like common sense, so candidates default to broad answers. Items that describe a storming-stage team, a resistant stakeholder, or a sponsor who is not engaged require the named concept (Tuckman stage, stakeholder analysis, RACI, force-field analysis), and the generic option is the distractor.
What to do instead
Learn the named models explicitly: Tuckman's forming–storming–norming–performing–adjourning, force-field analysis for driving versus restraining forces, stakeholder power/interest grid, and RACI for role clarity. For each scenario item, identify which model the situation invokes, then choose the action that model prescribes rather than the most diplomatic-sounding option.