**Introduction:**

In today's competitive market place, there is a need for business organisations to ensure continual improvement. Manufacturing companies experience growing pressure to improve quality, increase productivity, and reduce cost with limited resources. Service organisations need to reduce response time, eliminate errors, and improve customer satisfaction.

Though system certifications such as ISO9001:2008 could bring in some degree of discipline and quality improvement in organisations, it is not sufficient to address the real challenges.

We need to take a closer look at the manufacturing and service processes and deploy suitable techniques to enhance process capability. While 'percentage defects' is a thing of the past, achieving PPM (parts per million) defect levels is the challenge before today's managers.

Now comes the question of HOW ?

It is true that there are many approaches suggested by various quality experts. Corporations all over the world have been experimenting with one approach after the other, with little or no success. In this context, choosing a feasible path has become very important.

Many top management personnel are not aware that relatively simple techniques like SPC and EPC can be put to use to achieve quantum jumps in quality improvement and cost reduction.

Our experience shows that a step-by-step approach, based on statistical data analysis (SPC / EPC / Six Sigma) and managerial / technical corrective actions can deliver the desired results. Let us try to understand what is SPC, EPC, and Six Sigma all about.

**Top**

**Statistical Process Control (SPC):**

SPC is a time-tested and effective control scheme used for process capability analysis and process monitoring. SPC techniques consists mainly of Pareto Analysis, Scatter Diagram and Regression Analysis, and Statistical Control Charts.

**1. Pareto Analysis**

It is perhaps the most useful tool in the early stages of quality improvement initiatives. It can be deployed to identify the vital few and screen out trivial many.

Let us look at the following data on defect counts, taken from the inspection log of a garment manufacturing unit:

**Date** |
**Day** |
**Production (Pcs)** |
**Improper Stitch** |
**Missing Button** |
**Length Mismatch** |
**Reverse Buckle** |
**Total** |

05/07 |
Mon |
200 |
12 |
5 |
2 |
1 |
20 |

06/07 |
Tue |
250 |
8 |
10 |
5 |
0 |
23 |

07/07 |
Wed |
200 |
7 |
9 |
3 |
2 |
21 |

08/07 |
Thu |
150 |
4 |
6 |
2 |
1 |
13 |

09/07 |
Fri |
350 |
12 |
15 |
5 |
2 |
34 |

10/07 |
Sat |
250 |
10 |
12 |
4 |
0 |
26 |

What is the general conclusion ?

As the production increases, proportionately more defects are reported on those days.

Is this justification sufficient if you are looking for defect reduction ?

Let us summarise the same data in a different way, as shown below (Pareto Table):

**Defect** |
**Weekly Total** |
**% Total** |
**Cum. %** |

Missing Button |
57 |
41.61 |
41.61 |

Improper Stitch |
53 |
38.69 |
80.30 |

Length Mismatch |
21 |
15.33 |
95.63 |

Reverse Buckle |
6 |
4.37 |
100.00 |

**TOTAL** |
137 |
100.00 |
- |

Now, what is the conclusion ?

Missing buttons and Improper stitches contribute 80% of total defects. If the corrective and preventive actions can be focussed on elimination of the root causes of these two dominant defects from the process, we can easily achieve a significant reduction in overall defect tally.

Pareto Analysis can be effectively utilized for ...

- Machine down time analysis
- Dominant fault analysis
- Floor rejection analysis
- Customer complaint analysis etc.

**2. Scatter Diagram and Regression Analysis**

They are very useful in the study of inter-relationship between a key process output variable (KPOV) and a key process input variable (KPIV). If there is a significant relation between the two, the process output can be controlled effectively by controlling the process input.

There are many practical situations where measurement of product quality is not easy. For example, in case of mechanical properties of heat-treated steel, by the time the product is cooled, sample is taken and tested, a lot of production could have already happened. In case the test piece fails, you have already generated huge pile of scrap.

In such situations, it is worthwhile to explore whether the product quality (Y) can be controlled by controlling one/more process parameters (X's).

Consider the following data:

**Sl. No.** |
**Str. Rate (X)** |
**Imp. % (Y)** |
**.** |
**Sl. No.** |
**Str. Rate (X)** |
**Imp. % (Y)** |

1 |
16 |
7.1 |
. |
11 |
36 |
16.4 |

2 |
18 |
8.0 |
. |
12 |
38 |
15.5 |

3 |
20 |
8.4 |
. |
13 |
40 |
18.9 |

4 |
22 |
9.5 |
. |
14 |
42 |
18.5 |

5 |
24 |
11.8 |
. |
15 |
44 |
20.6 |

6 |
26 |
10.4 |
. |
16 |
46 |
19.8 |

7 |
28 |
13.3 |
. |
17 |
48 |
21.7 |

8 |
30 |
14.8 |
. |
18 |
50 |
22.8 |

9 |
32 |
13.2 |
. |
19 |
52 |
23.6 |

10 |
34 |
14.7 |
. |
20 |
54 |
25.4 |

The scatter diagram and regression line for this set of data shall be as below:

Please click on the picture to enlarge.

We can predict the value of Product Characteristic (Y) for various values of Process Characteristic (X) using the following equation:

Y = -0.5011 + 0.4635 X

**3. Control Charts**

Statistical Process Control charts (or simply, SPC charts) are used for monitoring the process performance and process variations. These charts may be constructed for monitoring of process parameter or product characteristic.

A control chart differs from an ordinary chart in the following aspects:

- Control chart has a centre line depicting the average process performance.
- It has two control lines, namely, Lower Control Limit (LCL) and Upper Control Limit (UCL). The control limits are calculated on the basis of natural (short- term) variations in the process.
- When a plotted point falls within the control limits, no action needs to be taken. But, any point falling outside the control limits requires further investigation / process adjustment.

Control charts can be constructed for both the variable (say, diameter) and attribute (say, surface defects) data.

The most commonly used variable control charts are the X-Moving Range chart, Xbar-Range chart, and Xbar-Sigma chart.

Widely used attribute control charts are the p-chart, np-chart, c-chart and u-chart.

Let us consider the following data on weight of tablet, taken from a pharmaceutical company:

Upper Specification Limit (USL) = 1.1 gram

Lower Specification Limit (LSL) = 0.9 gram

Target Value (T) = 1.0 gram

**Sl. No.** |
**Time** |
**Weight (Grams)** |
**.** |
**Sl. No.** |
**Time** |
**Weight (Grams)** |

1 |
06:00 |
1.05 |
. |
11 |
11:00 |
1.06 |

2 |
06:30 |
1.02 |
. |
12 |
11:30 |
1.09 |

3 |
07:00 |
1.06 |
. |
13 |
12:00 |
1.01 |

4 |
07:30 |
1.09 |
. |
14 |
12:30 |
1.00 |

5 |
08:00 |
1.05 |
. |
15 |
13:00 |
0.99 |

6 |
08:30 |
1.01 |
. |
16 |
13:30 |
0.96 |

7 |
09:00 |
1.08 |
. |
17 |
14:00 |
1.00 |

8 |
09:30 |
1.10 |
. |
18 |
14:30 |
0.99 |

9 |
10:00 |
1.06 |
. |
19 |
15:00 |
1.02 |

10 |
10:30 |
1.02 |
. |
20 |
15:30 |
1.04 |

For this data, X-Moving Range chart is most appropriate. Let us see the chart drawn by the SPC software.

Please click on the picture to enlarge.

In this case, all the data points are within control limits. Therefore, no process stoppage / adjustment is required to eliminate any assignable cause of variation.

Now let us see the capability statistics.

Process Potential Index (Cp) : 1.0991

Process Capability Ratio (Cr) : 0.9099

Process Performance Index (Cpk) : 0.7144

Taguchi's Index (Cpm) : 0.7198

As Cp > 1, the process has the potential to just meet the product specifications.

However Cpk < 1 indicates that the process is off-centred, i.e, the overall process average is not at the target.

Under the assumption of Normal Distribution of data, the expected defective tablets (in this case, over weight tablets) is 16056 per million produced (or 1.6 %).

This indicates that the process requires finer adjustments to drive the average to target value.

Coming to the implementation aspect of SPC charts, we need to first study the process data and compute the control limits. These limits can be used for monitoring and controlling the process in the subsequent period. Periodic re-calculation of process variation is also necessary.

**Top**

**Engineering Process Control (EPC)**

Engineering Process Control (EPC) is fast gaining popularity these days. While SPC charts provide a good check against assignable causes of variation, EPC charts can be used for prediction and run-on-run adjustment of process average.

Consider the data on weight of tablets discussed earlier. You may ask two interesting questions.

What would be the process average at 16:00 Hrs ?

What amount of adjustment is required NOW to bring the process average to target value ?

Let us see what the SPC software says:

Please click on the picture to enlarge.

Process mean at 16:00 Hrs (predicted)= 1.01823

Process adjustment required now (at 15:30 Hrs) = -0.016

You may now adjust the process parameters to lower the process average by 0.016 grams.

**Top**

**Six Sigma**

Six Sigma is a business initiative first introduced by Motorola in early 1990s. Recent Six Sigma success stories come from companies like General Electric, Allied Signal, Sony etc. According to GE's 1997 annual report, Six Sigma initiatives contributed more than 300 million US Dollars !

In general, Six Sigma implementation involves the following SEVEN phases:

1. DEFINE the processes that contribute to the problem.

2. MEASURE the capability of critical processes.

3. ANALYSE the data.

4. IMPROVE the key product / service characteristics.

5. CONTROL the key process variables.

6. STANDARDISE the methods for best-in-class process performance, and

7. INTEGRATE the standard methods and processes with the product / service design stage.

The Six Sigma strategy involves extensive use of statistical techniques such as control charts, design of experiments, response surface methodology etc. in order to minimise process variations and product / service defects. These techniques need to be applied in a structured manner.

While reporting the process improvement, Six Sigma teams use certain numeric values, known as Six Sigma Metrics. The most common metrics are 'Defects Per Million Opportunities (DPMO)', 'Sigma Quality Level', and 'Yield'.

'Defects Per Million Opportunities (DPMO)' is the number of critical defects that the process is estimated to generate per million opportunities (operations or steps). In shop-floor process control, this is also called defective 'Parts Per Million (PPM)' pieces produced by a single process / operation.

'Sigma Quality Level' is an indicator of process centering, and, process variation viz-a-viz technical tolerance. A process at Six sigma quality level is expected to generate only 3.4 defective Parts Per Million.

'Yield' is the estimated percentage of defect-free items (probability of zero defects) churned out by a process.

Based on the quality characteristic under study (variable / attribute data type), one or more metrics may be used for process monitoring and reporting.

It may be noted that the six sigma metrics are just the indicators of process quality. Sustaining and improving the process performance require process monitoring and control schemes such as Statistical Process Control (SPC), Engineering Process Control (EPC) etc.

**Top**

**Conclusion:**

Six Sigma initiatives aim at reduction of process variations and defects. SPC and EPC are two important techniques for achieving these goals. Relatively inexpensive and easy to understand (requiring minimal support from external experts), it is a feasible proposition to implement these techniques in any organisation.

**Top**

**Useful SPC/EPC/Six Sigma resources**

The following products and services could be helpful in understanding and applying the techniques, and getting the results: