Statistical Process Control (SPC) is a tools that helps to look at the variability of the process and assess if the process are in controll or out of controll.
SPC is composed by:
- The mean of the process;
- Upper control Limit / Lower control limit are the variance from the mean (UCL is set on +3 Sigma; LCL is set on -3 Sigma).
Remember that the control limit and specification limit are different. The control limit is the process’s voice and comes from the data distribution, his means, and his variance. The specification limit is the customer’s voice and comes from his requirements.
Example: The voice of the customer can be that the estimated time of a project will be about 28 days, +/- 2 days. The specification limits are 26 days and 30 days in this case. When we monitored the process, we took a mean of 27 and a standard deviation of 1. We have a control limit of 24 and 30 days in this case.
In a process you can have two different kind of variability:
- Chance variability: Is normal and always present in a process;
- Assignable variability: This kind of variability it’s the one that brings the process out of control, and we want to find in our SPC.
When we analyze the SPC, we want to assess the assignable variability. For do that we need to take care about:
- Outlier: If one point is out of the control limit, the process is out of control, and we need to investigate.
- Trends: Long-term trends, for example, the value starts to come near the lower control limit (or upper). This can bring the process out of control and need further investigation. Eight data points increasing or decreasing can signal a process out of control.
- Shift: if the process shift from the target value, that need further investigation. For example, 7 data point on the same side of the average can be a signal of a process out of control;
The causes of an assignable variability can vary from the problem with machinery, passing from the employer or the time to time variation, and even the error with the measurement itself.
It’s essential to take the process in control because this help to avoid defect.
Based on the different type of variable we can use different type of control charts for depict the SPC:
- Discrete variable (the one that has a finite number of value, like 1,2,3 or true/false);
- Count defect
- Constant sample size: C-Chart
- Variable sample size: U-Chart
- Classify that something is a defect
- Constant sample size: NP-Chart
- Variable sample size: P-Chart
- Count defect
- Continuous variable (the one that has an infinite number of value, like the weight, the temperature, etc.):
- Sample size 1 to 2: IMR
- Sample size 3 to 9: X-R
- Sample size >10: X-S
Why do we talk about sample size? We need to get a sample because we can’t collect all the data to monitor our process. This sample can be of different sizes. For instance, if you don’t have considerable variability in a single day (unit of measure), you can get a sample of size1.