Skip to content

Learn

Six Sigma explained: sigma levels, DPMO and belts

What Six Sigma is, how a sigma level relates to defects per million opportunities and to Cpk, what belts and DMAIC are, and where the method is misunderstood.

Published · Updated · 6 min read

What Six Sigma is

Six Sigma is a set of methods for reducing variation and defects, developed at Motorola in the 1980s and later widely adopted. The name refers to the standard deviation, written σ: a process with six standard deviations between its mean and the nearest specification limit produces very few defects.

In practice it is a way of working as much as a statistic. Improvement is done as projects with a defined problem, a measured baseline, analysed causes, tested solutions and a plan to hold the gain. Most projects follow the DMAIC cycle: define, measure, analyse, improve, control.

Sigma levels and defects per million

Defects per million opportunities (DPMO) counts defects against all the chances there were to make one: DPMO = defects ÷ (units × opportunities per unit) × 1 000 000. The sigma level converts a DPMO into a single number.

Imagine 200 housings, each with three defined inspection opportunities. Six defects across four housings give 6 / 600 × 1,000,000 = 10,000 DPMO. Four of 200 housings had defects, so the defective-unit rate is 2%. Keep the opportunities per housing fixed when comparing periods.

Using the common convention that allows the process mean to shift by 1.5σ over the long term, the usual values are:

  • Two sigma: about 308 500 DPMO
  • Three sigma: about 66 800 DPMO
  • Four sigma: about 6 210 DPMO
  • Five sigma: about 233 DPMO
  • Six sigma: 3.4 DPMO

The 1.5 sigma convention

The 1.5σ shift is a convention, not a law of nature, and it is debated. Whatever convention you use, state it, and use the same one whenever you compare processes. Comparing a sigma level computed one way with another computed a different way is a common source of confusion.

How it relates to Cp and Cpk

Process capability indices and sigma level describe the same thing from different angles. For a centred, stable process with normally distributed data, the short-term sigma level is about three times Cpk. A Cpk of 1.33 therefore corresponds to roughly four sigma, and a Cpk of 2.0 to six sigma in the short term.

Multiplying Cpk by three gives the distance to the nearest limit in units of within-subgroup standard deviation. Converting that distance into a defect probability needs a distribution model. A gap between Cpk and Ppk shows that short-term and overall variation differ. Use the time series to investigate why.

Belts and roles

Six Sigma organisations use belt levels to describe training and responsibility. The names vary between companies, but the common pattern is:

  • Yellow Belt: understands the basics and supports projects
  • Green Belt: leads smaller projects alongside a normal job
  • Black Belt: leads complex projects and coaches others
  • Master Black Belt: guides the programme and trains Black Belts
  • Champion or sponsor: removes obstacles and owns the business result

DMAIC, DMADV and the customer

DMAIC is for improving an existing process. DMADV (define, measure, analyse, design, verify) is used when designing a new process or product so that it meets customer requirements from the start. Both begin with the customer: the characteristics that matter to them, called critical to quality (CTQ) characteristics, set what is measured and what counts as a defect.

The role that matters most is the sponsor. Projects without a sponsor who cares about the result tend to stall, however good the statistics are.

Where Six Sigma goes wrong

The common problems are well known:

  • Jumping to solutions before the process and the measurement system are understood
  • Treating the sigma level as the goal instead of the customer's requirement
  • Projects that end with a presentation and no control plan, so the gain fades
  • Using statistics where there is too little data for them to mean anything

Standards and further reading

Book a call about your factory

Tell us what you make and which machines you run, and we will walk through how Qynetic fits. If you would rather get news and early access, join the waitlist.

Learn