Ppk
Ppk, or process performance index, measures a process's actual long-term capability against upper and lower specification limits using overall variation, not just within-subgroup data. Calculated as the smaller of (USL - mean)/(3σ) or (mean - LSL)/(3σ), it reflects real-world drift, tool wear, and setup differences in CNC machining or millwork operations.
On the shop floor, Ppk is applied to critical dimensions like hole size, pocket width, or panel thickness where long-run drift from tool wear, thermal growth, coolant effects, or fixture repeatability can push parts toward a spec limit even if first-off parts look acceptable. Engineers collect measured data from pilot lots, PPAP runs, or production audits and compare the distribution to spec limits to assess whether the process is centered and stable enough to meet customer requirements. In millwork or edgebanding, cumulative variation from saw setup, feed speed, glue line thickness, substrate moisture, and trimming shifts the mean and widens overall spread; Ppk captures this real-world spread better than short-term metrics. It is especially useful when the process is not fully stabilized or when a customer-facing performance number reflecting actual shipped product is needed. Common targets are Ppk ≥ 1.67 for strong acceptance, 1.00 to 1.67 for improvement, and < 1.00 as not meeting requirements, though exact thresholds depend on part criticality and customer specifications.
- Using a short, stable startup sample to claim good performance: The first 10–30 parts may be centered, but later tool wear, thermal soak, or clamp relaxation widens total spread and collapses Ppk, creating false control.
- Confusing Ppk with Cpk: Cpk uses within-subgroup variation and assumes stability; Ppk uses total dataset. A cell can show acceptable Cpk while Ppk exposes real shipped risk from shift or fixture changes.
- Bad measurement discipline: Gage variation, inconsistent probing, or thermal growth during inspection distorts Ppk. In precision CNC or finish-critical millwork, poor metrology can make the process look worse or better than it is.
When should Ppk be used instead of Cpk?
Use Ppk when you need actual long-term performance including shift and drift; use Cpk when the process is statistically stable and you want short-term capability based on within-subgroup variation.
What does a low Ppk usually mean in machining?
The process mean is too close to a spec limit, total variation is too wide, or both. Common causes are tool wear, spindle thermal growth, fixture inconsistency, tool deflection, or uncontrolled setup differences between shifts.
Why is the smaller side of the Ppk formula used?
Because process performance is limited by the nearest spec boundary. If the mean is off-center, one side fails first, so the minimum of the upper and lower ratios defines the real risk.