DPMO (defects per million opportunities) counts how many defects occur for every million chances to make one. Converting the defect rate to a z-value with the normal distribution and adding the conventional 1.5σ shift gives the familiar sigma level, which lets you compare processes of very different complexity.
Independent ways one unit can fail that you inspect for.
4.5 sigma by the usual convention (z = 3.0 without the 1.5σ shift).
About 1,333 defects for every million chances to make one.
Share of all opportunities that turned into a defect.
Share of opportunities that were right first time.
Long-term z-value from the normal distribution, no shift added.
DPO = D / (U × O) DPMO = DPO × 1,000,000 Yield = (1 − DPO) × 100 % z = Φ⁻¹(1 − DPO) Sigma level (short-term convention) = z + 1.5 D = defects, U = units, O = opportunities per unit; Φ⁻¹ = inverse standard normal CDF
A PCB assembly line checks every board for 3 defect opportunities. Last month: how good was it in sigma terms?
The line runs at about 1,333 DPMO, a 4.5 sigma level by the usual convention (3.0 sigma without the shift). Reporting both values avoids the classic confusion between the long-term z and the 'sigma level' figure.
Paste it into your AI assistant to be walked through the calculation with your own data.
Act as a quality engineer and help me calculate DPMO and the sigma level for my process. Use exactly these formulas: DPO = D / (U × O) DPMO = DPO × 1,000,000 Yield = (1 − DPO) × 100 % z = Φ⁻¹(1 − DPO), where Φ⁻¹ is the inverse standard normal CDF Sigma level (short-term convention) = z + 1.5 Rules: 1. Work step by step and show every intermediate value. 2. Repeat my counts first and check the decimal and thousands separators. 3. If D, U or O is missing or unclear, ask me — never assume a number of opportunities. 4. Report the sigma level both with and without the 1.5σ shift and say which is which. 5. If D = 0, do not give a sigma level; tell me to report the sample size and period instead. My data: Defects D = Units U = Opportunities per unit O = What counts as an opportunity =
The buttons copy the prompt first, then open the assistant in a new tab.
Next step · DOCAMIN
Draft an improvement project charter from this result in DOCAMIN.
Open DOCAMIN →Learn the method · Academy
Turn a sigma level into an improvement project with the PDCA cycle in our free course.
Start the free course →PPM counts defective units per million units; DPMO counts defects per million opportunities. A unit with several possible defects has several opportunities, so DPMO normalises for complexity, while PPM tells you how many units a customer would reject.
It is a convention: long-term data are assumed to drift by about 1.5 standard deviations compared with short-term performance, so the long-term z is raised by 1.5 to state a 'short-term' sigma level. It is an assumption, not a law — always say whether your figure includes it.
DPMO is an attribute measure: it needs counts of defects. With measured data, use the Cp/Cpk calculator instead; its expected PPM can be converted into a sigma level with the same inverse normal function.
Count only independent, inspectable ways the unit can fail that matter to the customer, define them in writing and keep them stable over time. Splitting one failure mode into many 'opportunities' inflates the denominator and makes the process look better than it is.
Last tested: October 2026 · Tested with: automated math tests (3+ cases) · Reviewed by: Quality for Everyone editorial team
