How to use the sigma level calculator
Pick one of the three modes:
- Defects → DPMO & sigma. Enter the number of defects found, the number of units inspected and the number of defect opportunities per unit. You get DPU, DPO, DPMO, yield, first-time yield, the long-term Z and the sigma level, with a 95% confidence interval.
- Convert. Enter any one of DPMO (PPM), yield, sigma level, long-term Z, Cpk or Ppk and read all the others.
- Rolled throughput yield. List the yield of each process step (or the defects and units per step) to get RTY, normalized yield and total DPU.
The formulas
| Metric | Formula | Meaning |
|---|---|---|
| DPU | D / U | Average defects per unit |
| DPO | D / (U × O) | Defects per opportunity |
| DPMO | DPO × 1,000,000 | Defects per million opportunities |
| Yield | 1 − DPO | Fraction of opportunities without a defect |
| First-time yield | e−DPU | Probability a unit has zero defects (Poisson) |
| Long-term Z | Φ⁻¹(1 − DPO) | Normal quantile matching the observed defect rate |
| Sigma level | Long-term Z + 1.5 | Short-term capability by the Six Sigma convention |
| RTY | Y₁ × Y₂ × … × Yk | Probability a unit passes every step first time |
| Normalized yield | RTY1/k | Geometric mean yield per step |
| Total DPU | −ln(RTY) | Defects per unit over the whole process |
Φ⁻¹ is the inverse of the standard normal distribution. The calculator evaluates it in double precision, so the far tail (a few defects per million or per billion) is accurate rather than interpolated from a printed table.
The 1.5σ shift, stated plainly
Defect data collected over weeks or months are long-term data: they already contain tool wear, shift changes and batch-to-batch movement. The Six Sigma convention, introduced at Motorola, assumes that such drift moves the process mean by about 1.5 standard deviations, and therefore credits the process with a short-term capability 1.5σ higher than the long-term Z. That is why “six sigma” corresponds to 3.40 DPMO — the tail beyond 4.5σ — and not to the 0.001 DPMO beyond a true 6σ.
The shift is a convention, not a law of nature. This tool always shows both numbers — the long-term Z with no shift and the sigma level with the shift — and lets you change the shift (enter 0 to switch it off). If you have measurement data, estimate the real difference between short-term and long-term variation with the process capability calculator (Cpk versus Ppk) instead of assuming 1.5.
Sigma level to DPMO table
| Sigma level | DPMO (1.5σ shift) | Yield (1.5σ shift) | DPMO, no shift, one tail | PPM, no shift, both tails | Short-term Cpk |
|---|---|---|---|---|---|
| 1σ | 691,462 | 30.85% | 158,655 | 317,311 | 0.33 |
| 2σ | 308,538 | 69.15% | 22,750 | 45,500 | 0.67 |
| 3σ | 66,807 | 93.32% | 1,350 | 2,700 | 1.00 |
| 4σ | 6,210 | 99.379% | 31.7 | 63.3 | 1.33 |
| 5σ | 233 | 99.9767% | 0.2867 | 0.5733 | 1.67 |
| 6σ | 3.40 | 99.999660% | 0.001 | 0.002 | 2.00 |
“Both tails” is a centred process with specification limits at ± the sigma level; the familiar 2,700 PPM for ±3σ is in that column. A sigma level corresponds to a short-term Cpk of sigma level ÷ 3 against the nearer limit, and the long-term Z to Ppk × 3.
Worked examples
Defects to sigma level. An audit of 1,500 invoices, each with 8 fields that can be wrong, finds 34 errors.
Opportunities = 1,500 × 8 = 12,000. DPU = 34 / 1,500 = 0.0227. DPO = 34 / 12,000 = 0.002833, so DPMO = 2,833 and the yield is 99.717%.
Long-term Z = Φ⁻¹(1 − 0.002833) = 2.77; sigma level = 2.77 + 1.5 = 4.27. First-time yield per invoice = e−0.0227 = 97.76%.
Rolled throughput yield. Five steps have first-pass yields of 98.5%, 96.2%, 99.1%, 97.4% and 99.6%.
RTY = 0.985 × 0.962 × 0.991 × 0.974 × 0.996 = 91.10%. Although no step is below 96%, about one unit in 11 is reworked or scrapped somewhere; to finish 100 right-first-time units you must start 109.8.
Normalized yield = RTY1/5 = 98.15% per step, total DPU = −ln(RTY) = 0.0932, and the average step runs at a sigma level of 3.59.
Practical notes
- Count opportunities honestly. DPMO falls if you inflate the number of opportunities. Count only characteristics that are actually checked and that matter to the customer, and keep the definition fixed over time.
- Small counts are uncertain. Zero defects in 500 opportunities does not prove a six-sigma process; the calculator reports the 95% upper bound (about 3/n) so you can see what the data really support.
- DPMO versus PPM. PPM defective counts defective units; DPMO counts defects per opportunity. They coincide only when each unit has one opportunity.
- Track it over time. A single sigma level hides trends. Plot the defect rate on an attribute control chart and rank the causes with a Pareto chart.
Frequently asked questions
How do I calculate sigma level from DPMO?
Convert DPMO to a defect probability p = DPMO / 1,000,000, find the standard normal value Z that leaves p in the upper tail (Z = inverse normal of 1 − p), and add the 1.5 sigma shift: sigma level = Z + 1.5. For example 6,210 DPMO gives Z = 2.5 and a sigma level of 4.0.
Why is Six Sigma 3.4 defects per million and not 0.002?
A specification limit six standard deviations from a perfectly centred mean would give about 0.001 defects per million in one tail (0.002 for both). The Six Sigma convention assumes the mean drifts by 1.5 standard deviations in the long term, leaving the limit 4.5 standard deviations away, and the normal tail beyond 4.5 sigma is 3.4 per million.
What is the difference between DPMO and PPM?
PPM defective is the number of defective units per million units. DPMO is the number of defects per million opportunities, where each unit can have several opportunities for a defect. A unit with three defects counts once in PPM but three times in DPMO. They are equal only when there is exactly one opportunity per unit.
What is rolled throughput yield (RTY)?
Rolled throughput yield is the probability that a unit passes through every process step without a defect or rework. It is the product of the first-pass yields of all steps. Five steps at 98 percent each give an RTY of about 90 percent, which exposes the hidden factory of rework that a final-inspection yield hides.
How do I convert Cpk to sigma level and PPM?
For the nearer specification limit, short-term sigma level = 3 × Cpk, so Cpk 1.33 is 4 sigma and Cpk 2.0 is 6 sigma. The PPM beyond that limit with no shift is the normal tail beyond 3 × Cpk; with the conventional 1.5 sigma long-term shift it is the tail beyond 3 × Cpk − 1.5. The convert mode of this calculator shows both.
What yield corresponds to each sigma level?
With the 1.5 sigma shift: 2 sigma is about 69.1 percent, 3 sigma 93.3 percent, 4 sigma 99.38 percent, 5 sigma 99.977 percent and 6 sigma 99.99966 percent. The reference table on this page is computed directly from the normal distribution and updates if you change the shift.