Cpk (within-subgroup capability) assesses potential capability based on short-term variation, while Ppk (overall capability) includes both short and long-term variation. A significantly lower Ppk indicates that the process variation over time (between subgroups) is larger than the variation within subgroups. This is a classic sign of a process that is not stable or in statistical control. The primary focus must be on identifying and eliminating the sources of this between-subgroup variation to achieve process stability before addressing pure capability (Cpk).

The fundamental purpose of an FMEA is prevention. The most robust and preferred action is to eliminate the cause of the failure mode or reduce its rate of occurrence (Occurrence). Improving detection only means you are better at finding the failure after it has already occurred, which is a less effective control. While RPN can guide priorities, the philosophy of prevention prioritizes reducing Occurrence over improving Detection. Severity reduction, when possible through design change, is also highly valuable but may not always be feasible.

Using a gage beyond its calibration due date invalidates the measurement results because the gage’s accuracy can no longer be assured. Therefore, any product acceptance decision based on that measurement lacks objective evidence. This is a serious failure of the quality management system, constituting a major nonconformance. The argument that the gage ‘seemed fine’ or that the product was within spec is irrelevant; the integrity of the decision-making process has been compromised. The system must ensure measurements are made with equipment whose accuracy is known and maintained.

Rolled Throughput Yield (RTY) is the product of the individual first-pass yields: 0.95 * 0.90 * 0.98 = 0.8379 or 83.8%. It represents the probability that a unit will pass through *all* process steps correctly the first time, without any scrap or rework. This is a powerful metric because it highlights the hidden factory of rework and repair. An 83.8% RTY means over 16% of units require extra, non-value-added processing, directly impacting cost, capacity, and cycle time, which is critical information for a manager.

The natural process spread is represented by the width of the control limits (11.2 – 8.8 = 2.4 mils for 3σ). Therefore, the 6σ process spread is 2 * 2.4 = 4.8 mils. The specification spread (tolerance) is 12 – 8 = 4 mils. Since 4.8 mils > 4 mils, the natural variation of the process is wider than the customer’s allowable variation. This means the process is inherently unable to consistently meet specifications, even though it is statistically in control. Capability (Cpk/Ppk) would be less than 1.0. Being in control is necessary but not sufficient for capability.

This is a fundamental quality management concept. A **correction** is an immediate fix to deal with a problem that has occurred (e.g., reworking a defective part, replacing a faulty unit). It addresses the symptom. A **corrective action**, however, goes deeper. It involves investigating to find the root cause of the nonconformity and implementing systemic changes to prevent the problem from happening again. Effective quality systems require moving beyond corrections to robust corrective actions to achieve continual improvement.

The main advantage of double (or multiple) sampling plans is reduced average sample number (ASN). For lots that are either very good (few defects) or very bad (many defects), a decision to accept or reject can often be made after inspecting the first, smaller sample. Only lots of intermediate quality require the full second sample to be inspected. This leads to less inspection effort on average compared to a single sampling plan, which always uses its full sample size, providing economic efficiency while offering the same protection (AQL, LQL) as the single plan.

When facing multiple potential factors (5 in this case), a screening design is needed to efficiently separate the significant factors from the noise. A full factorial (2^5=32 runs) may be overkill initially. A 2-level fractional factorial design is specifically intended for this purpose—it uses a carefully chosen subset of the full factorial runs (e.g., 8 or 16) to estimate main effects and some interactions. This makes it highly efficient for identifying the ‘vital few’ factors that have the greatest impact, which can then be studied in more detail. OFAT is inefficient and cannot detect interactions.

A raw count of complaints can be a misleading metric. If sales volume (the opportunity for complaints) has significantly decreased, the complaint *rate* (e.g., complaints per 1000 units shipped) may not have improved at all, or could even be worse. The apparent improvement could simply be an artifact of lower output. Before attributing a trend to process changes, one must normalize the data (use a rate or ratio) to account for changes in the underlying volume or scale of activity. This is a critical step in valid metric analysis.

Cpk (within-subgroup capability) assesses potential capability based on short-term variation, while Ppk (overall capability) includes both short and long-term variation. A significantly lower Ppk indicates that the process variation over time (between subgroups) is larger than the variation within subgroups. This is a classic sign of a process that is not stable or in statistical control. The primary focus must be on identifying and eliminating the sources of this between-subgroup variation to achieve process stability before addressing pure capability (Cpk).0