SPC XL Sigma Capability Calculation Change
SPC XL 2000 Build 4 and SPC XL 2010 change the way Sigma Capability is changed from 3*Cp to NORMSINV ((1,000,000 – dpm)/1,000,000) + 1.5. The reason for this change is discussed by Mark Kiemele, the president of Air Academy Associates, below.
A major change in SPC XL Build 4 is the way in which Sigma Capability is computed. In previous builds of SPC XL, Sigma Capability was computed as 3*Cp or 3*Cpk (if only one-sided spec) for continuous data. In Build 4, Sigma Capability for continuous data will be computed the same way as attribute data Sigma Capabilities are calculated, namely, based on the dpm value as follows (in Excel terms): NORMSINV ((1,000,000 – dpm)/1,000,000) + 1.5.
Just a comment on why we are moving our software and materials to Sigma Capability being computed based on dpm (or 3.4 dpm = 6 Sigma capability). While using 3*Cp to calculate sigma capability conforms to the original Six Sigma Research Institute's ideas, we can no longer argue with more than a decade of inertia that has the Six Sigma community equating 6 sigma to 3.4 dpm, 5 sigma to 233 dpmo, etc. The 3*Cp calculation (or 3*Cpk for one-sided specs) is really a "potential" measure and is easily calculated, as long as Cp or Cpk is provided, which it still will be in SPC XL Build 4. Under this method of computing Sigma Capability, it is possible, for example, to have 500,000 dpm and still have a 6 Sigma capability because Cp does not use the center of the process in its calculation, and is instead dependent only on the value of sigma and the USL and LSL. This is an unexpected result for most practitioners, causing more confusion than it is worth and possibly providing a gross overestimate of the current capability of a process. Our moving to the 3.4 dpm = 6 Sigma capability definition is an attempt to become more standardized with most of the rest of the Six Sigma world, which uses dpm (after the 1.5 sigma shift) as the basis for the sigma capability or sigma rating which Six Sigma practitioners have learned to expect. This method also allows for benchmarking on both attribute and continuous data, because both types of data will generate sigma capabilities that are based on defect rates, namely dpm. The only difference that a SPC XL Build 4 user will detect from previous versions is the value printed out for Sigma Capability.
Mark J. Kiemele, Ph D
President, Air Academy Associates