A bimodal distribution in a histogram often signals that data from two distinct sources, methods, shifts, or operators are being combined. The correct approach is to stratify the data (e.g., by shift, operator, machine) to identify and understand the different processes causing the two peaks. Calculating a single mean or assuming normality would mask the root cause.

When the data is attribute data (defective/not defective) and you want to compare the proportion (or percentage) of defects from two independent samples (before vs. after), the test for equality of two proportions (e.g., a 2-proportion test) is the correct choice. T-tests and ANOVA are used for continuous data.

A core principle of root cause analysis is to move from potential causes to verified root causes using data. Brainstorming generates hypotheses, but these must be tested and confirmed with evidence before solutions are developed. Jumping to a solution based on an unverified assumption is a common pitfall.

The first response to a single point outside control limits should be to investigate for a special, assignable cause. This includes checking for measurement error, material defect, or a temporary glitch. One should not automatically adjust the process, as this could over-control a stable process if the point was due to a unique, non-recurring event.

In the Define phase, a SIPOC (Suppliers, Inputs, Process, Outputs, Customers) diagram is used to gain agreement on the scope of the process under review. It helps the team and stakeholders understand the process at a high level, identify key start and end points, and clarify who provides what and who receives the output, ensuring everyone is aligned before deep analysis begins.

A fractional factorial design is specifically intended for screening experiments. It allows you to study many factors (and some of their interactions) with fewer experimental runs than a full factorial design. This efficiency makes it ideal for identifying the ‘vital few’ significant factors from a larger list of potential causes early in an improvement project.

The heart of a control plan is the reaction plan. It clearly states the actions to be taken (e.g., who to notify, what to check, when to stop production) if control charts, audits, or other monitoring tools signal that the process is moving toward an out-of-control condition. This empowers the process owner to sustain the improvement.

Cp measures potential capability (spread relative to specs), while Cpk measures actual capability (accounting for centering). A Cp (1.5) that is significantly higher than Cpk (0.8) indicates that the process spread is capable (good variability), but the process mean is not centered between the specification limits. The primary opportunity is to shift the process mean toward the target.

The standard categories (often the 6 Ms: Man, Machine, Material, Method, Measurement, Mother Nature) are merely common starting points. The power of the tool comes from the team brainstorming causes relevant to their specific situation. If ‘Customers’ is a logical major category for their problem (e.g., customer-provided data is incorrect), it should absolutely be included. The diagram should be tailored to the context.

A bimodal distribution in a histogram often signals that data from two distinct sources, methods, shifts, or operators are being combined. The correct approach is to stratify the data (e.g., by shift, operator, machine) to identify and understand the different processes causing the two peaks. Calculating a single mean or assuming normality would mask the root cause.0