DPMO and Sigma Level: Understanding Six Sigma Performance

• Defects versus defective units: A defect is an individual failure to meet a requirement. A defective unit is a product or transaction containing one or more defects. For example, one application with three errors represents three defects but only one defective application.
   

• Defect opportunities: An opportunity is a clearly defined possibility for a defect to occur. The number of opportunities should reflect meaningful requirements, not an arbitrary count designed to improve the reported sigma level. Consistency is essential when comparing performance over time.
   

• Short-term versus long-term sigma level: This is where Six Sigma terminology can become confusing. In traditional capability analysis, short-term performance generally describes variation observed under relatively consistent operating conditions. Long-term performance considers the additional changes that may occur over time.
  
  The traditional Six Sigma reporting convention assumes that a process may experience a shift equivalent to 1.5 standard deviations between short-term and long-term performance. Consequently: Conventionally reported sigma level = Sigma equivalent of observed defect rate + 1.5. For example, an observed defect rate equivalent to a Z-score of 3.0 is commonly reported as a 4.5 sigma process under this convention.
  
  IMPORTANT: Adding 1.5 does not improve the actual process or change the observed defect rate. It changes how performance is expressed. Nor does it prove that the process actually experienced a 1.5 sigma shift.

• Understanding the 1.5 sigma shift: The 1.5 sigma adjustment originated as a convention for relating idealized short-term process capability to anticipated long-term performance. It helps explain why the familiar Six Sigma benchmark is 3.4 defects per million opportunities rather than the much smaller theoretical defect rate associated with a perfectly centred, stable, normally distributed process having six standard deviations to its nearest specification limit. However, the 1.5 sigma adjustment is a convention, not a universal statistical law.

​

• Actual process shifts may be smaller, larger, or absent: This distinction is especially important for transactional and service processes, where performance is often measured using counts of errors rather than continuous measurements. In these situations, the sigma level calculated from DPMO is best understood as a standardized expression of observed defect performance—not proof of an underlying normal distribution or a measurable process shift.

• Reporting sigma level responsibly: Consider the statement, “Our process operates at 3.2 sigma.” That sounds precise, but it leaves important questions unanswered. Was the sigma level calculated directly from the observed defect rate? Was the conventional 1.5 sigma adjustment included? What time period was measured? How were defect opportunities defined?
  
  A more meaningful report might state, “Our process recorded 45,000 DPMO during the reporting period, equivalent to approximately 3.20 sigma using the conventional 1.5 sigma adjustment.” This communicates the observed performance, the calculation convention, and the reporting period. The most useful sigma level is not necessarily the highest number. It is the one whose meaning is clearly understood.