Six Sigma

Six Sigma is a data-driven methodology for reducing process variation and defects, originally developed at Motorola in the 1980s and popularized by General Electric. The name comes from the statistical goal of placing the specification limits six standard deviations from the process mean, so that defects become extremely rare. In the long term, because processes drift, Six Sigma assumes a 1.5 sigma shift in the mean, which corresponds to 3.4 defects per million opportunities (DPMO). Without this shift, six sigma from the mean would correspond to about 2 defects per billion. The methodology is applied through the DMAIC framework for improving existing processes: Define (state the problem, scope, customer requirements, and project goals), Measure (quantify current performance, validate the measurement system with Gage R&R, and establish baseline capability), Analyze (identify root causes using tools such as fishbone diagrams, Pareto charts, hypothesis tests, and regression), Improve (develop, pilot, and implement solutions using design of experiments and lean tools), and Control (sustain the gains with control charts, standard work, and response plans). A variant called DMADV (Define, Measure, Analyze, Design, Verify) is used for designing new processes or products. Key metrics include DPMO, process capability indices Cp and Cpk, and sigma level. Cpk measures how well the process fits within its specification limits while accounting for centering: Cpk = min[(USL - μ)/(3σ), (μ - LSL)/(3σ)]. A process with Cpk = 2.0 and a centered mean is roughly at Six Sigma capability before the 1.5 sigma shift. Roles in a Six Sigma organization include Champions (senior sponsors), Master Black Belts (program leaders and trainers), Black Belts (full-time project leaders), Green Belts (part-time project leaders), and team members. Lean Six Sigma combines Six Sigma's focus on variation and defects with Lean's focus on eliminating waste and improving flow. Common failure modes of Six Sigma programs include selecting projects with no real business impact, applying statistical tools without understanding the process, and failing to sustain improvements after the project closes.