Sigma Level, DPMO, and Cpk: Reconciling Two Ways of Reporting the Same Process
Ask for a process’s capability and you’ll get one of two answers depending on who’s asked: a Cpk value from a quality engineer, or a “sigma level” from anyone with a Six Sigma background. They’re describing the same underlying process performance, but they’re not directly interchangeable numbers, and the gap between them trips up more capability discussions than it should.
Two Metrics, One Underlying Idea
Cpk measures how many standard deviations fit between the process mean and the nearer spec limit, divided by three. It’s a ratio — spec distance over half the process spread — and it’s calculated straight from process data with no adjustment layered on top.
Sigma level and DPMO (defects per million opportunities) come out of the Six Sigma tradition and describe the same underlying relationship — process spread relative to spec — but expressed as an expected defect rate rather than a capability ratio. A process running at “six sigma” is meant to correspond to roughly 3.4 defects per million opportunities.
If both metrics are describing the same normal distribution against the same spec limits, they should convert cleanly back and forth using the standard normal distribution. And under short-term, perfectly centered conditions, they do: a Cpk of 1.5 and a sigma level of 4.5 (short-term) describe the same process.
Where the 1.5 Sigma Shift Comes From
The complication is that the commonly cited Six Sigma sigma-level table doesn’t describe short-term performance directly — it bakes in a long-term shift assumption. The reasoning: short-term process capability studies capture the process at its best, running consistently over a short sampling window, while real long-term performance drifts due to tool wear, material lot changes, seasonal effects, and all the sources of variation that a short capability study window won’t capture. Motorola’s original Six Sigma framework accounted for this by assuming a long-term shift of 1.5 sigma away from center, and the standard sigma-level-to-DPMO conversion table is built around that assumption.
This is why “six sigma” corresponds to 3.4 DPMO rather than the roughly 2 parts per billion a genuinely centered six-sigma process would produce without any shift assumption — the widely cited number already has the 1.5 sigma shift baked into it. A Cpk of 2.0, calculated directly from data with no shift assumption applied, is the short-term equivalent of “six sigma” in the pre-shift sense; comparing it directly against a DPMO table that already assumes long-term drift will make the process look worse than the Cpk alone suggests, because the DPMO figure is answering a different question — sustained long-term performance — than the raw Cpk is.
Why This Matters for Reporting
Mixing these up in a report produces numbers that look inconsistent even when they’re both technically correct, because they’re answering different questions. A team reporting “Cpk of 1.67, six sigma quality” is combining a short-term capability figure with a long-term-adjusted sigma level, and the two aren’t the same claim about the process even though they’re being presented side by side as if they were.
The cleaner approach separates the claims explicitly: report Cp/Cpk for short-term potential, Pp/Ppk for actual long-term performance over the sampled period, and treat sigma level and DPMO as a translation of whichever of those four numbers is actually being converted — stating clearly whether the 1.5 sigma shift assumption is included in that translation or not. A DPMO figure without a stated assumption about long-term shift is incomplete, because the same underlying Cpk can map to noticeably different DPMO figures depending on whether the shift is applied.
The Practical Takeaway
None of the four capability indices need a sigma-level translation to be useful on their own — Cp, Cpk, Pp, and Ppk are complete, well-defined numbers without any conversion applied. The translation to sigma level and DPMO is a communication layer for audiences more familiar with Six Sigma terminology, and it needs the shift assumption stated explicitly whenever it’s used, or it invites exactly the kind of apples-to-oranges comparison that makes two accurate numbers look contradictory side by side.
Calculating the Underlying Numbers Correctly
Getting Cp, Cpk, Pp, and Ppk right in the first place — with the correct within- versus overall-sigma calculation and a normality check ahead of it — is the foundation any sigma-level conversion depends on. SigmaDesk’s process capability calculator calculates all four indices side by side with a built-in normality check, free in the browser, part of the complete SigmaDesk SPC platform.
Report the capability index the data actually supports, and state plainly whether any shift assumption was applied before translating it into a sigma level. That single disclosure resolves most of the confusion the conversion otherwise creates.
