Six Sigma: Meaning, DMAIC & Formula for Factories
6 min read
By
TranZact Solutions Team
· Published


Six Sigma is a data-driven method for cutting defects and variation in a process. A process at Six Sigma level makes no more than 3.4 defects per million opportunities, which is 99.99966% defect-free. Teams improve an existing process in five steps called DMAIC: Define, Measure, Analyse, Improve and Control.
What does Six Sigma mean?
Sigma (σ) is the symbol for standard deviation, a measure of spread. Six Sigma means six standard deviations fit between the process average and the nearest specification limit. The tighter the spread, the fewer parts fall outside the limit.
Bill Smith introduced Six Sigma at Motorola in 1986. Jack Welch made it central at GE in 1995. The method is now set out in the standard ISO 13053:2011.
Six Sigma fixes the process so fewer bad parts get made, rather than catching them later. Our guide to QA vs QC explains that difference.
What 3.4 defects per million really means
3.4 defects per million opportunities equals a 99.99966% yield. The figure allows for processes drifting about 1.5 sigma over time. Not every process needs six sigma; each one needs a target that fits its cost and risk.
The Six Sigma formula (DPMO and sigma level)
The Six Sigma formula is DPMO, or defects per million opportunities:
DPMO = defects ÷ (units × opportunities per unit) × 1,000,000
A defect is anything that fails the customer's specification. An opportunity is each separate thing you check that could fail, such as one dimension on a drawing.
Sigma level table
Sigma level | DPMO | Yield |
|---|---|---|
1 | 691,462 | 30.9% |
2 | 308,538 | 69.1% |
3 | 66,807 | 93.3% |
4 | 6,210 | 99.38% |
5 | 233 | 99.977% |
6 | 3.4 | 99.99966% |
Like most Six Sigma tables, this one includes the 1.5 sigma shift.
Worked factory example
This example is illustrative. A machine shop makes 1,000 parts and checks 5 dimensions on each. That gives 5,000 opportunities. Inspection finds 25 defects.
DPMO = 25 ÷ (1,000 × 5) × 1,000,000 = 5,000. That sits between 4 sigma (6,210) and 5 sigma (233), just above 4 sigma. The yield is 99.5%.
The result is only as good as the counts behind it. Accurate counts start with consistent quality inspection.
DMAIC: the 5 stages of Six Sigma
DMAIC is the five-stage method Six Sigma teams use to improve an existing process. Here it runs on one example: rejections on a CNC-turned shaft at an Indian auto-parts unit.
Define. Write the problem in one line: shafts rejected for oversize diameter. Note the customer's specification, what rejections cost and who is on the team.
Measure. Count units made and defects per batch, by machine and shift. First check that the gauge gives the same reading twice. This is your baseline.
Analyse. Find the root cause. Compare defects by machine, operator, tool life and raw material batch. A fishbone diagram helps list every possible cause.
Improve. Test one fix on a small run, such as a shorter tool-change interval. Keep it only if defects drop against the baseline.
Control. Lock in the gain. Update the work instruction, put a control chart at the machine and review the numbers every month.
DMADV for new products
DMADV (Define, Measure, Analyse, Design, Verify) is also called Design for Six Sigma. Use DMAIC to fix a process that exists and DMADV to design a new one.
DMAIC on one CNC rejection problem
Define: one part, one defect
Measure: defects per batch and shift
Analyse: find the root cause
Improve: test one fix
Control: chart it, review monthly
Where TranZact fits: the data a Six Sigma project needs
First, the limit. TranZact is not a statistics or SPC tool. It doesn't draw control charts or calculate sigma levels; use Excel or a statistics package for that.
What a project needs most is clean operating data. TranZact's manufacturing AI software keeps sales, purchase, inventory, production and planning on one system. 10,000+ manufacturers use it.
DMAIC stage | Data you need | Where TranZact records it |
|---|---|---|
Define | Cost of rework and scrap | Production costing |
Measure | Units made, defects by batch | Work orders; batch-wise stock tracking (Scale+); Batch Tracking & QC (Dominate) |
Analyse | Steps, machines, supplier lots | Process routing; machine planning (Dominate); GRN for each delivery |
Improve | Changes to materials or steps | Multi-level BOM; process routing |
Control | Ongoing tracking | 65+ reports; approvals & alerts; real-time MIS (Dominate) |
Quality control for raw material and finished goods is on the Dominate plan; see pricing for each plan. TranZact integrates with Excel and Google Sheets, so you can run the statistics there.
Clean batch tracking is the base for most defect analysis. Comparing tools? Our quality control software buying guide lists what to check.
Six Sigma principles
Most guides list six principles. Each one has a plain meaning on the shop floor.
Focus on the customer. The customer's specification decides what counts as a defect.
Use data and facts. Decide from defect counts, not the loudest opinion in the room.
Improve the process. Fix the step that makes bad parts, not just the parts.
Act early. Watch the trend instead of reacting to each customer complaint.
Work across teams. Purchase, stores and production team members own the problem together.
Aim for perfection, accept small failures. Some fixes won't work, and that is part of the method.
Six Sigma belts and roles
White Belt: knows the basic ideas and terms.
Yellow Belt: joins projects and helps collect data.
Green Belt: leads small projects alongside a regular job.
Black Belt: leads larger projects full time and coaches Green Belts.
Master Black Belt: trains Black Belts and guides the whole programme.
Champion: a senior leader who picks projects and clears roadblocks.
India has many White Belt and Green Belt course providers. An owner-led factory can start with one person trained to Green Belt level.
Six Sigma vs Lean vs Kaizen
Six Sigma cuts variation. Lean cuts waste. Kaizen makes small improvements every day. Lean Six Sigma combines the first two, removing waste and variation in the same project.
Six Sigma | Lean | Kaizen | |
|---|---|---|---|
Goal | Fewer defects, less variation | Less waste and waiting | Small, steady improvement |
Method | DMAIC projects led by data | Map the flow, remove waste | Everyone suggests and tests small fixes |
Tools | SPC, FMEA, DOE, control charts | Value stream map, 5S, kanban | PDCA, 5S, suggestion schemes |
Size of change | Medium to large, one project at a time | Process-wide | Small and continuous |
Best for | Repeat defects you can measure | Long lead times, excess stock | Building a daily improvement habit |
For more, read our guides to kaizen and traditional vs lean manufacturing.
Common Six Sigma tools and honest limits
The common tools are statistical process control (SPC), control charts, process mapping, FMEA and DOE. Many factories also track OEE next to defect rates.
Six Sigma has critics. Joseph Juran called it "a basic version of quality improvement". Some statisticians call the 1.5 sigma shift arbitrary. In 2010 the Wall Street Journal reported on how often Six Sigma projects fail.
One lesson holds either way. Without clean data, a project has no baseline to prove a gain.
Ready for your first Six Sigma project?
One problem that keeps repeating
A written definition of the defect
Defect counts by part and batch
Units made in the same period
Cost of rework and scrap
One person trained to Green Belt
Start with one process
Pick the one rejection that costs you most. Count it, run DMAIC and prove the gain with numbers. TranZact goes live in 2–3 weeks, so your baseline data can start soon.
Don't be the bottleneck. Become the AI-Run Factory.




