Statistical methods are employed to continually improve process performance and efficiency. Statistical Process Control is applied when analyzing a process to correctly identify special cause from common cause variations. When identified incorrectly, one is either over adjusting by correcting for common causes or not taking action for special cause which leads to an unpredictable process and decreased efficiency. An effective means of detection must be employed that follows the standard SPC methodologies and rules.
When looking at SPC software it is critical to look below the surface. Care should be taken to ensure that the software adheres to the fundamentals of SPC. Assuming that a software employs correct methods of detection and calculations can be a very costly mistake. Key questions should be asked when looking at different SPC software vendors regarding the standards followed, statistical methods employed, as well as other capabilities of the software.
Control and Specification Limits
When monitoring SPC variable data there are two sets of limits defined, specification limits and control limits. Specification and control limits are distinct values that should not be viewed as one and the same. Specification limits are determined by customer requirements and determines if the product is acceptable to ship. A run chart is used to compare each individual observation value with the specification limits.
Control limits are based upon process performance and calculated using historical data. A control chart is used to compare the average of the observations, dependent upon sample size, with the control limits. Only in the event that the sample size is equal to one, are the run charts and control charts plotting the same data. Comparing an average of the individual observation values with the specification limits provides no information about the process, whether the point falls inside or outside the specification limits. For this reason, the specification limits should not be present on the control chart with a sample size greater than one.
Control Limit Calculation
Control charts have been proven as the most effective means of detecting special causes during production. However, if a control chart is not setup and monitored correctly, it is not only ineffective but can actually degrade process performance resulting in substantial loss through increased scrap, rework and downtime. Standards have developed in regards to the calculation and recalculation of control limits since they are critical in monitoring process control.
Control limits are calculated values based upon previous process performance, historical data. When control limits are initially calculated, all samples should be contained within the trial limits. If any sample(s) fall outside the limits, they should be withheld from the calculation. Once the control limits are established, the limits should be extended for ongoing monitoring of the process. Only in the event that a change is made to the process in efforts to reduce common cause variation, should control limits be recalculated.
Automatic Control Limit Calculation
The automatic calculation of control limits is prohibited by Standard SPC methodologies and should not be incorporated into any SPC software program. Control limits are critical decision aids when monitoring a process. If not calculated correctly, they will have damaging effects on the process and negate any attempts of successful process improvements. For this reason, when investigating an SPC software’s capabilities the methods for control limit calculation is key, but there are also several other areas to examine to ensure standard SPC guidelines are followed.
Demand that SPC software vendors supply supporting documentation as evidence in response to the questions listed below.
- Is the SPC software able to define a specific range of historical data to include in the calculation of control limits?
- Does the calculation of control limits use the estimated or calculated value of sigma?
- Can samples be excluded when calculating control limits with reason noted and if so is this visually indicated on the control chart?
- Will the software allow extending the calculated control limits?
- Is process status clearly indicated and displayed on the control chart?
- What control trend tests are available and which trend test rule was applied?
- Was a cause assigned and corrective action taken to trending or out of control data point(s)?
- When are control limits recalculated and logged in the report?
- Can multiple sets of control limits be calculated, extended and displayed on the chart?
- Can process data be reviewed to see how it has changed over time?
- Are process capability charts able to be generated in the software with both +/- 3? and +/- 6? displayed?
- Are automatic email alerts available in the software to notify the appropriate personnel of out-of-spec and/or out-of-control conditions?
The design and execution of Six Sigma projects require a series of sequenced events that absorb much time and energy from the personnel involved. Implementing an SPC software program is essential part of this process. Before deciding on an SPC software program to use, it is critical to thoroughly investigate the software and the statistical and probability methods employed. Not adhering to standard guidelines when analyzing SPC data negatively affects the process and has a detrimental effect on any process improvement efforts.