Meaning
Statistical metrics measure the inherent variation of a stable process relative to specified engineering tolerances. A capability index calculates the ratio between the allowable design width and the actual natural spread of the process output under controlled conditions. This ratio quantifies whether a manufacturing step consistently produces parts within defined upper and lower specifications during steady state execution.
The evaluation requires a statistically stable process with normally distributed output, distinguishing inherent process capability from active machine capacity or shift throughput.
Process Yield
Calculating long-term yield projections from short-term pilot data introduces substantial risk when tool wear or thermal drift expands the natural process spread over full production runs. Initial pilot runs often exhibit minimal variance due to fresh tooling and tight environmental controls, yielding a capability index that overstates serial performance. When volume production begins, tool degradation and raw material lot variations reduce this statistical cushion.
Relying on early estimates causes unexpected scrap spikes and expensive secondary sorting operations.
Tolerance Limit
Specification boundaries established by engineering design define the absolute limits within which a manufactured component functions correctly. A capability index compares this total allowable tolerance band against six standard deviations of process output. If the process mean shifts away from the nominal midpoint, the one-sided capability index accounts for the reduced margin toward the nearest tolerance limit.
Maintaining an adequate buffer prevents non-conforming parts from entering downstream assembly lines.
Variance Spread
Process variation consists of short-term common cause noise and long-term special cause shifts. The capability index isolates natural variation by analyzing subgroups collected over continuous operating intervals. Comparing this index against observed defect rates identifies whether quality failures stem from inadequate machine precision or uncorrected mean shifts.
Controlling the natural spread ensures predictable performance across extended manufacturing campaigns.