3.4
Defects per Million (6σ)
5
DMAIC Phases
99.99966%
6σ Yield Rate
1986
Created at Motorola

What Is Six Sigma?

Six Sigma is a data-driven methodology for eliminating defects and reducing variation in any process. Developed at Motorola in 1986 and popularized by General Electric under Jack Welch, it uses statistical tools to measure performance, identify root causes, and implement lasting solutions.

The name comes from statistics: a process operating at "six sigma" produces only 3.4 defects per million opportunities. Most companies operate between 3 and 4 sigma.

💡 Key Insight

Six Sigma isn't just about quality — it's about using data to make better decisions. If you can measure it, you can improve it.

The Sigma Levels

The sigma level tells you how capable your process is. Higher sigma = fewer defects = happier customers.

Sigma LevelDefects / MillionYieldWhat It Feels Like
2σ308,53769.15%Nearly 1 in 3 items has a defect
3σ66,80793.32%Typical unmanaged process
4σ6,21099.38%Pretty good — but still ~6K defects/million
5σ23399.977%World-class operations
6σ3.499.99966%Near-perfect quality
📋 What Does 99% vs 99.99% Mean? Perspective

At 99% quality (3.8σ): 20,000 lost mail articles per hour. 5,000 botched surgeries per week. 2 short or long landings at major airports daily.

At 99.99966% quality (6σ): 7 lost mail articles per hour. 1.7 botched surgeries per week. 1 bad landing every 5 years.

The difference between "good" and "excellent" quality can be life-changing — literally.

The steps get harder as you climb Each level costs more than the last: 2σ to 3σ is a 5× cut, 5σ to 6σ is a 69× one. The five rows of the table span five orders of magnitude, not five equal steps.

The page's own sigma table plotted on a logarithmic axis of defects per million opportunities: 2σ 308,537, 3σ 66,807, 4σ 6,210, 5σ 233, 6σ 3.4. Distance on this axis is ratio, so the widening gaps are real: the steps between adjacent levels are 5×, then 11×, then 27×, then 69×, each one harder than the last. The shaded band is where the page says most companies operate, between three and four sigma. Note where that sits: closer to 2σ than to 6σ, and about three orders of magnitude short of it. The page puts the same distance in units a room recognises: at 99 per cent quality, 20,000 lost mail articles an hour, 5,000 botched surgeries a week and two bad landings a day at major airports; at 99.99966 per cent, 7 lost articles an hour, 1.7 surgeries a week and one bad landing every five years.

DMAIC: The Five Phases

DMAIC (Define, Measure, Analyze, Improve, Control) is the structured problem-solving framework at the heart of Six Sigma.

The DMAIC Framework
Define
→
Measure
→
Analyze
→
Improve
→
Control

Define

What's the problem? Who's the customer? What does "good" look like? Create a clear problem statement, define scope, and identify the customer's critical-to-quality (CTQ) requirements. A good problem statement is specific and measurable: "Order accuracy dropped from 99.2% to 97.1% in Q3."

Measure

Collect baseline data. How bad is the problem right now? Establish your measurement system and validate it's accurate. Map the current process, measure cycle times, defect rates, and variation. You can't improve what you don't measure.

Analyze

Find the root cause — not symptoms. Use Pareto charts, fishbone diagrams, regression analysis, and hypothesis testing to identify the vital few factors driving your defects. SymplProcess's Pareto Analysis tool automates this.

Improve

Design and test solutions. Use pilot runs to validate improvements before full-scale rollout. Focus on solutions that address root causes, not workarounds for symptoms.

Control

Lock in the gains. Create control charts, standard operating procedures, and monitoring dashboards so improvements stick. Without control, processes drift back to their old state within weeks.

Common Pitfalls

✅ Signs You're Doing It Right

  • Decisions backed by data, not opinions
  • Root causes validated before solutions tried
  • Control plans in place before project closes
  • Team members from the process are involved

❌ Common Mistakes

  • Jumping to solutions in the Define phase
  • Analysis paralysis — 6 months of data collection
  • No control phase = improvements evaporate
  • Using Six Sigma for simple problems that just need a fix
⚡
Not everything needs Six Sigma If the fix is obvious (broken machine, missing training), just fix it. Six Sigma is for chronic problems where the root cause is unclear and data is needed to find it.

Interactive Demo

Explore how sigma level and process mean shift affect defect rates. Adjust the sliders to see the normal distribution change in real time.

⚡
Try It Yourself
Sigma Level Explorer
▼
Adjust the sigma level and mean shift to see how they affect defect rates. The 1.5σ shift represents long-term process drift — this is why 6σ short-term equals 4.5σ long-term (3.4 DPMO).
3σ
1σ6σ
1.5σ
0σ3σ
Reference (with 1.5σ shift)
1σ2σ3σ4σ5σ6σ
691,462308,53866,8076,2102333.4
133,614
DPMO
86.639%
Yield
1.5σ
Effective Sigma
13.3614%
Defect Rate
LSLUSLμ=1.5σTarget
Ready for the full knowledge check? Test your understanding with guided scenarios and data export.
PROTake the Pro Knowledge Check →
Free forever · Every feature included

Stop reading, start modeling

Model your process flow, run simulations, optimize staffing with TOC math, and test your knowledge with 107 interactive checks — all in one platform.

Open Workbench → Pro Knowledge Check

Take this to a room

The running order

For a team scoping an improvement project. They should leave able to write a measurable problem statement and to say why Control is the phase that decides whether any of it lasts.

7 beats · 9 min
  1. 1

    What the name means

    A process at six sigma produces 3.4 defects per million opportunities. Most companies run between three and four.

    • Motorola, 1986. Popularised by GE.
    • It is a statistical target, not a slogan.
    • The gap between three sigma and six is not a rounding difference.
  2. 2

    Why 99 per cent is not good

    This is the comparison that makes the point faster than any chart.

    • At 99 per cent: 20,000 lost mail items an hour, 5,000 botched surgeries a week, two bad landings a day at major airports.
    • At six sigma: 7 lost items an hour, 1.7 a week, one bad landing every five years.
    • Nobody accepts 99 per cent once it is stated in those units.

    Ask the room What would 99 per cent look like on our line, in units?

  3. 3

    Define is where projects die

    A vague problem statement guarantees a vague result, and it is the cheapest thing to fix.

    • Who is the customer and what does good look like to them?
    • Specific and measurable: order accuracy is 94 per cent against a target of 99.5.
    • Not: improve quality. That is a wish, not a scope.
  4. 4

    Measure before you theorise

    Baseline the problem, and validate that the measurement system itself is trustworthy first.

    • How bad is it right now, in numbers?
    • Is the gauge repeatable? A bad measurement system invents variation.
    • Map the current process, time it, count the defects.
  5. 5

    Analyse means the vital few

    Find the factors that actually drive the defects, not the ones that are easiest to blame.

    • Pareto to prioritise, fishbone to broaden, data analysis to confirm.
    • Test the cause: does it explain every instance?
    • A cause that explains half the cases is a contributing cause.
  6. 6

    Improve on a pilot

    Design and test, then roll out. Solutions go to root causes, not to symptoms.

    • Pilot before full-scale, always.
    • Prefer a process change or an error-proofing device over a reminder.
    • If the fix is training people to be more careful, keep going.
  7. 7

    Control is what makes it real

    Without control, processes drift back to their old state within weeks - and the gains vanish quietly.

    • Control charts, updated SOPs, a monitoring dashboard.
    • Someone owns the metric after the project closes.
    • A project without a control phase is a temporary result.

    Ask the room Name a past improvement here that quietly went backwards. What was missing?