Why Six Sigma? And Not Seven or Five Sigma?

The Six Sigma methodologies were formulated by Bill Smith at Motorola in order to arrest the quality downslide of the company in 1986. Six Sigma tools are developed by taking inspiration from traditional Quality Control, TQM, and TPM etc. Apart from Motorola other pioneer user of Six Sigma are Honeywell, GE etc.

• Normal distribution: Suppose you are inspecting an assembly of two metal plates bolted in 25 places. You have taken 100 such assemblies for inspection. The numbers of missing bolts in any assembly are the number of defects in that particular assembly. Now, if you plot a graph by taking number of defects in Y axis and number of assemblies in X axis you will get a bell shaped curve. In other words, the number of defects in the sample assemblies are normally distributed.

Any variable is the sum of many number of independent factors and will follow a normal distribution curve. Most of the practical processes follow normal distribution curve.

• Standard deviation (σ): It tells how closely the inspected data is clustered around the mean of the data set. The greater the value of standard deviation, the wider and flatter the bell shaped normal distribution curve will be .

• Process capability: Process capabilities are measurements which tells what percentage of process output falls outside process specifications. In other words it compares process spread to USL (Upper Specification Limit) and LSL (Lower Specification Limit). See the snapshot below (click on it to enlarge):

• Look at the snapshot above. A normally distributed process if mean plus minus 3σ range of output is between USL and LSL then around 99.997% (or almost all) of the output will be non-defective or the process is of 6σ level.
• Now, if the USL and LSL of a process are such that the difference between USL and LSL is less than 6σ (6 X standard deviation) value of the process, then it won’t be a Six Sigma process. To make this process a Six Sigma process you have to reduce the values of σ. The Six Sigma methodology broadly tells about how to reduce the value of σ and make it a six sigma process.

• DPMO (Defect Per Million Opportunities) is used for counting defects in statistical processes. A Six Sigma process will have 99.997% accuracy or 3.4 DPMO which is extremely accurate.

• We have already discussed that Six Sigma processes are fairly accurate (99.997% accurate, 3.4 DPMO). Achieving that level of accuracy itself is very difficult and lots of systematic approach is required apart from capital investment. But if you want to improve the process further to seven or eight sigma level it increases the cost of the process drastically to achieve little more accuracy. Even though, some of the aerospace company has now started talking about seven sigma levels but most of the organizations feels that up to Six Sigma is good enough.

• Running a process in five sigma levels is quite possible, in fact most of the traditional companies run in 3 to 4 sigma levels. But a process which is running in 5 sigma levels could be improved to six sigma levels by following some systematic approach and little more investment.

Six Sigma is a systematic and data driven approach to improve process capability. Accuracy level of a six sigma process is 3.4 DPMO or 99.997% which is fairly accurate. Most of the pioneer organizations of Six Sigma users found that up to the sigma level of six is right trade off between cost and quality however running a process in five sigma or seven sigma is individual organizations personal call on the basis of required accuracy and cost.