CSSC Green Belt cheat sheet

SIPOC (Suppliers, Inputs, Process, Outputs, Customers) is built before the charter to bound scope and shows only 4-7 high-level process steps.

A complete problem statement states what, where, when, and how much, plus the gap to target, and never names a cause or a solution.

Goal statement format: verb + metric + baseline + target + deadline, e.g. 'Reduce billing errors from 8% to 3% by 30 June'.

VOC is translated into CTQs with a CTQ tree: customer need, then quality driver, then measurable requirement.

COPQ = internal failure + external failure + appraisal + prevention costs; the first three are the cost of poor quality.

DMAIC improves an existing process; DMADV (Define-Measure-Analyze-Design-Verify) designs a new process or product when none exists.

Team development stages occur in order: Forming, Storming, Norming, Performing, Adjourning.

Roles: Champion/Sponsor funds the project and removes barriers; Process Owner sustains gains after handoff; MBB coaches BBs and GBs.

Kano categories are must-be (dissatisfiers), performance (more is better), and delighter (unexpected); a missing must-be requirement destroys satisfaction.

Gage R&R acceptance per AIAG: %GRR of 10% or less is acceptable, 10-30% is marginal and judged on cost and risk, above 30% is unacceptable.

DPMO = (defects / (units x opportunities per unit)) x 1,000,000; a 6-sigma short-term process corresponds to 3.4 DPMO.

Cp = (USL - LSL) / 6 sigma measures potential capability and ignores where the process mean sits.

Cpk = min[(USL - mean), (mean - LSL)] / 3 sigma; Cpk equals Cp only when the process is perfectly centered.

RTY = Y1 x Y2 x ... x Yn using first-pass yields; the normalized yield per step = RTY^(1/n).

Normal distribution coverage: about 68.27% within +/-1 sigma, 95.45% within +/-2 sigma, 99.73% within +/-3 sigma.

Central Limit Theorem: the distribution of sample means is approximately normal for n of 30 or more regardless of population shape.

Capability indices are valid only when the process is in statistical control and the data are normal or have been transformed.

Sample size for estimating a mean: n = (Z x sigma / E)^2, where E is the desired margin of error.

Hypothesis test logic: assume H0 is true; if p-value < alpha, reject H0, otherwise fail to reject H0 (never 'accept H0').

Alpha = P(Type I error) = producer's risk; beta = P(Type II error) = consumer's risk; power = 1 - beta.

Test selection: 1-sample t for mean vs target, 2-sample t for two means, paired t for before/after on the same units, ANOVA for 3+ means, chi-square for counts, F-test or Levene for variances.

Confidence interval for a mean: x-bar +/- t(alpha/2, n-1) x s / sqrt(n); a target outside the interval means the difference is significant at alpha.

ANOVA splits total sum of squares into between-group (factor) and within-group (error); p < alpha means at least one group mean differs.

Regression review: overall p-value, each predictor's p-value, R-squared and adjusted R-squared, plus residual plots for normality, equal variance, and independence.

FMEA: RPN = Severity x Occurrence x Detection, each scored 1 to 10; the highest RPNs get action and are re-scored afterward.

Root cause tools include 5 Whys, the fishbone with 6M (Man, Machine, Material, Method, Measurement, Mother Nature), and Pareto of causes.

A Multi-Vari chart separates positional, cyclical, and temporal variation to isolate the family of variation that dominates.

DOE fundamentals are randomization, replication, and blocking; randomization protects the experiment against nuisance factors and drift.

A full 2^k factorial estimates all main effects and interactions in 2^k runs; a fractional factorial screens many factors in fewer runs at the cost of confounding.

OFAT (one-factor-at-a-time) experimentation cannot detect interactions and needs more runs than a designed experiment.

Takt time = available work time / customer demand; it sets the required pace, which is compared against actual cycle time.

OEE = Availability x Performance x Quality; world-class OEE is roughly 85%.

The seven wastes (TIMWOOD) are Transport, Inventory, Motion, Waiting, Overproduction, Overprocessing, and Defects.

SMED cuts changeover time by converting internal setup (machine stopped) into external setup (machine running).

Poka-yoke prevents an error from being made or makes it immediately obvious; 5S is Sort, Set in order, Shine, Standardize, Sustain.

A Pugh or weighted solution matrix scores options against criteria, and the chosen solution is piloted before full rollout.

Control limits are +/-3 sigma from the process mean and are computed from the data; specification limits come from the customer and are never plotted as control limits.

Chart selection: continuous subgroups use X-bar-R (n up to about 9) or X-bar-S (n of 10 or more); individuals use I-MR; proportions use p or np; defect counts use c or u.

Out-of-control signals include a point beyond 3 sigma, 8 consecutive points on one side of the center line, 6 consecutive trending points, and 2 of 3 points beyond 2 sigma.

Common cause variation is inherent and stable; special cause variation is sporadic and assignable and is handled with an out-of-control action plan (OCAP).

Cpk uses within-subgroup (short-term) sigma while Ppk uses overall (long-term) sigma, so Ppk is less than or equal to Cpk for the same data.

A control plan names the metric, monitoring method and frequency, the owner, and the reaction plan for each critical X.

Gains are standardized through updated SOPs and work instructions, operator training, and mistake-proofing of the critical inputs.

Capability and the measurement system are re-verified after improvement; a drifting process means the project is not ready to close.

Project closure hands off to the process owner with documented results, lessons learned, and an audit schedule.