Key takeaways
Short answer: Process capability and process performance are two families of indices that look almost identical but use different variation. Capability (Cp, Cpk) uses short-term, within-subgroup variation — the process's potential if only common-cause variation were present. Performance (Pp, Ppk) uses long-term, overall variation — what the process actually delivered over time, including the shift and drift between subgroups. Performance indices are usually lower, and the gap reveals instability the capability view hides. For the stability question beneath, see in control vs in spec.
Process capability — Cp and Cpk — measures what a process is potentially capable of, using short-term variation. Specifically, capability indices are calculated from the within-subgroup variation: the variation seen within small, rationally-grouped samples taken close together in time, which captures the inherent, common-cause variation of the process at its best, excluding the drift and shifts that happen between subgroups over longer periods. Cp compares the tolerance to this short-term spread (ignoring centering); Cpk adds centering, exactly as in Cp vs Cpk. Because it uses the tight short-term variation, capability represents the process's potential — what it could achieve if it were perfectly stable, with only that inherent variation present. Capability answers: how good could this process be, based on its inherent short-term variation? It is the best-case, potential view.
Process performance — Pp and Ppk — measures what a process actually delivered, using long-term variation. Performance indices are calculated from the overall variation: the total variation across all the data over a longer period, which includes not just the within-subgroup common-cause variation but also the shifts and drift between subgroups — the way the process mean wanders and the spread changes over time. Pp is the performance analogue of Cp (spread versus tolerance, ignoring centering); Ppk the analogue of Cpk (adding centering). Because it uses the larger long-term variation, performance represents the actual, realized result — what the process really achieved over time, including all its instability. Performance answers: how good was this process actually, over the real period, including all its drift and shifts? It is the actual, realized view, not the potential.
The clean distinction is the variation each uses: capability (Cp/Cpk) uses short-term within-subgroup variation, performance (Pp/Ppk) uses long-term overall variation. This makes capability a measure of potential (what the process could do based on inherent variation) and performance a measure of actual results (what it really did over time, including drift). The formulas are otherwise identical — the only difference is which standard deviation goes in: the short-term within-subgroup estimate for capability, the long-term overall estimate for performance. This is why the indices look so similar and are so easily confused, yet mean different things. The relationship is informative: because long-term overall variation includes the between-subgroup drift that short-term within-subgroup variation excludes, performance indices (Pp/Ppk) are usually lower than capability indices (Cp/Cpk). The gap between them measures how much the process drifted and shifted over time — instability that the short-term capability view does not see.
The most useful insight comes from comparing the two. If Cp/Cpk (short-term capability) is high but Pp/Ppk (long-term performance) is noticeably lower, the gap reveals that the process drifts or shifts over time: its inherent short-term variation is fine, but the mean wanders or the spread changes between subgroups, so the actual long-term result is worse than the potential. This points at instability — special-cause variation entering over time, the world of common versus special cause and being in control. Conversely, if capability and performance are close, the process is stable — its short-term and long-term variation are similar, meaning little drift. So the two indices together diagnose more than either alone: capability shows the potential, performance shows the reality, and the gap between them quantifies the instability. A large gap is a signal to find and remove the drift, not to reduce the inherent variation.
A process is measured two ways. Using short-term within-subgroup variation, its Cpk comes out at 1.5 — based on its inherent variation, it looks highly capable. But using the long-term overall variation across months of data, its Ppk is only 1.0 — the actual realized result is much worse. The gap from 1.5 to 1.0 is the tell: the process's inherent short-term variation is tight (high capability), but over time the mean has drifted and shifted between subgroups, so the long-term performance is substantially lower than the short-term potential. The diagnosis is instability, not excessive inherent variation — the fix is to find and eliminate the drift (a special cause entering over time, a slowly-wandering setting, a temperature cycle) rather than to reduce the within-subgroup spread, which is already good. Capability said the process could be excellent; performance said it actually was not, because it would not hold still.
These indices sit beneath the quality factor of OEE, like Cp and Cpk generally. Process performance (Pp/Ppk), being the long-term actual result, corresponds most closely to what the OEE quality factor experiences in practice — the real, realized defect rate over time, including drift, not the short-term potential. A process with high short-term capability but poor long-term performance will produce a worse yield and scrap picture than its capability suggests, dragging the quality factor down through the drift the capability view misses. The connection to stability ties directly to in control vs in spec and common versus special cause variation: the capability-performance gap is, in effect, a measure of the instability that makes the OEE quality factor worse than the process's potential.
Fabrico trends the live quality outcome over time — the long-term, realized result that process performance (Pp/Ppk) captures and that the OEE quality factor actually experiences. By showing whether the quality factor holds steady or drifts over weeks and shifts, it surfaces exactly the instability that opens a gap between short-term capability and long-term performance, flagging a process that is capable in principle but drifting in practice. Book a demo to see your long-term process performance, not just its potential.
Process capability (Cp, Cpk) uses short-term, within-subgroup variation and measures the process's potential. Process performance (Pp, Ppk) uses long-term, overall variation and measures the actual realized result over time, including drift. Performance is usually lower than capability.
Because Ppk uses long-term overall variation, which includes the drift and shifts between subgroups that the short-term within-subgroup variation behind Cpk excludes. The extra long-term variation makes the performance index lower than the capability index.
It means the process drifts or shifts over time. The short-term inherent variation is tight (high Cpk), but the mean wanders or spread changes between subgroups, so the long-term actual result (Ppk) is worse. The gap signals instability to be removed, not inherent variation to reduce.
Use capability (Cp/Cpk) to assess the process's potential from its short-term variation, ideally once the process is stable. Use performance (Pp/Ppk) to assess what the process actually delivered over time, including drift. Reading both together reveals instability the capability view misses.
Process performance (Pp/Ppk), the long-term actual result, corresponds most closely to what the OEE quality factor experiences — the real defect rate over time, including drift. A high-capability but low-performance process drags the quality factor down through the drift capability misses.
Connexion
Blog
