This figure illustrates the basic assumptions of a pre-control chart. Source: Jim L. Smith


Ask any machinist how statistical process control (SPC) helps him manage his processes. The response is not likely to be positive. It is not uncommon for control charts to be placed around machines and other equipment, showing a history of how certain process parameters or key process characteristics (KPCs) have been performing. But for SPC to actually achieve its intended goal, to offer adjustments prior to making defective product, three things must be established:
    1) a real-time effective measurement system.
    2) a realistic tolerance linked back to customer enthusiasm.
    3) a known dial with a predictable response.

    In terms of capability, a process falls into three categories:

    1) stable and capable where virtually no adjustment is required.
    2) incapable.
    3) capable and drifting.
There are situations in which SPC helps the machinist manage processes by using pre-control, but there also are those in which it does not work. The intent of process control is to signal when to adjust a process so that ‘out-of-print’ product is not made in the first place. This responsibility falls with the machinist. The success of process control is usually dependent on the machinist’s interpretation of the data and the ability to react to that data in a timely manner. A key point is that the success of implementing process control is not dependent on the actual collecting and plotting of the data. Success is more about how that data is used in decision making.

If we ask machinists how SPC helps them manage their processes, we are likely to get a wide range of responses. Some would feel it is useful; some would feel it is a waste of time. The machinist feels that the time used to collect data when no action is required is wasted or redundant. Many machinists are likely to rely on a process or quality engineer to tell them when adjustments are needed. Their response is typically dependent on the type of process they are attempting to control and the method of process control being used.

X-bar and R charts are typically more valuable to process and quality engineers than machinists who must make decisions of adjustment. Machinists can become confused when a point falls out of control, but is still within print specifications. They may be confused as to when to adjust with a process that is moving toward the control limits. It also may require frequent recalculating of control limits. If this is done automatically (by a computer), then that is not much of a problem. If it is done manually, however, there can be an abundance of difficulties.
    Effective process control requires three major components:

    1) a clear goal, or specification.
    2) an effective measurement system to determine where one is relative to that goal.
    3) a known dial with a predictable response to steer that process within its specifications.
These components are necessary with whatever process control is being used. The usefulness of this, however, is dependent on the type of process one is trying to apply. If a machinist is asked to apply SPC to a process that is incapable, the machinist will likely become frustrated. Too many adjustments would be necessary, resulting in over-adjusting the process. These processes are in need of an investigative strategy to find and correct the variables contributing to the variation and to improve capability.

In the case of a capable and drifting process, this is exactly the usefulness of process control. Machinists can observe where processes are performing in relation to the goal (specification) and adjust accordingly with a predictable response. These are processes in which pre-control can help the machinist with the decisions of when to adjust. The goal of process control is to give the machinist a tool that is easily implemented and easy to use after it is in place, taking the guesswork out of decision making.


    This figure illustrates a pre-control chart. Source: Jim L. Smith

    What is Pre-Control?

    Many statisticians disagree about pre-control as an SPC tool. Much of the disagreement about pre-control revolves around understanding the objectives and assumptions of its use, contrasted with the objectives and assumptions of control charts.

    Constant emphasis is needed as a reminder that pre-control is not a quality improvement tool, but rather a technique to keep a technician from producing scrap and rework. With this said, pre-control is statistically based and it is simple to understand and implement.

    Pre-control was developed in 1954 by a team at Rath and Strong. The team was led by Dorian Shainin and included Warren Purcell, Charlie Carter and Frank Satterthwaite. Shainin, Purcell and Carter had extensive experience in manufacturing, and Satterthwaite was a statistician. Pre-control was developed in response to the need for developing SPC for short-run processes where X-bar and R charts were not practical. It turned out that pre-control worked well for short- and long-run processes.

    The technique begins with the assumption that the process is centered within the middle half (50%) of the specification limits; then it tries to detect shifts that might result in parts that fall outside the specification limits. Pre-control tells the machinists:
      1) when the process can be run (after five consecutive pieces pass).
      2) when to adjust the process.
      3) when sampling is safe to use.
      4) when the process needs attention.
      5) when to leave the process alone.
    Pre-control works very well for high-capability processes; however, we do not need to know the process capability in advance when applying pre-control. Pre-control makes no assumptions about the capability of a process, but it does require that five consecutive checks fall in the middle half of the specification before sampling begins. This essentially qualifies the process as ‘OK to run’ without collecting a large number of data points. This is especially valuable for small lot sizes. If a process is not capable, getting five consecutive pieces in the middle half of the specification is difficult, so it essentially becomes a 100% inspection to ensure quality protection.

    Pre-control is an active process control tool used to quickly check the status of a process. It does not use X-bar and R charts. It identifies when to continue running, adjust a process, or stop and fix a process. Through the use of pre-control setup and running rules, the user can quickly decide which of the previous actions is appropriate. It is a powerful yet simplistic technique that can reduce set-up time, process adjustments, and scrap and rework. The specification limits (not control limits) are divided into five zones. There is a central green zone, which includes ± 50% of the specification tolerance. Two yellow regions (each representing 25% of the specification tolerance) with one on each side of the green zone. Two red zones are on each side of the yellow zones.

    These zones can be established for bilateral tolerance or unilateral tolerances with a limit, and unilateral tolerances with no limits. The simplicity of pre-control allows easy calculation of these zones. For a bi-lateral tolerance, the green zone represents ± 50% of the tolerance and, if the process is stable and capable, 86% of the data points should fall within the green zones and the yellow zones will contain 7% on either side. For a unilateral tolerance with a limit, the green zone is the 86% of the tolerance closest to the limit, and the yellow the 14% furthest from the limit.

    If Cpk = 1.0, it means that the tolerance equals the process spread and the mean coincides with the tolerance mean. In such cases, and assuming normal distribution, we can expect that 86% of the readings will be in the green zone and 7% in each of the yellow zones. Therefore, we can expect one out of 14 readings (7%) in the yellow zone.

    Following the laws of probability, the chance of getting two consecutive readings in a yellow zone will be (1/14 x 1/14) or 1/196 or 0.51%. This is the foundation of pre-control. Considering all four possible permutations of the consecutive two pieces, the chance is 4/196 or approximately 2%.

    Pre-Control Rules

    1.Set up the pre-control chart.Divide the specification tolerance range with pre-control (PC) lines at ¼ of the acceptable tolerance zones.

    2.Start the process.If any check in the first five pieces falls outside the specification limits, reset the aim.

    3. When five consecutive pieces fall within the green zone, begin sampling.

    4.Sampling.Continue following the sampling plan (normally 25 pieces but that is only a general rule as the process performance should govern). If the first piece is within target (green zone), run (no need to measure second piece). If first piece is outside green zone but within yellow zone, check the next piece. If the second piece is within green zone, adjust and go back to start.

    A Real-Life Application

    A senior quality engineering manager is faced with managing quality in a short-run machine shop. There is no available automated or computerized process control system as traditional X-bar and R charts are not feasible because of the small lot sizes. A tool is clearly needed to ensure the factory is producing quality products. Traditional 100% inspection is only partially effective, as well as cost-prohibitive. After careful consideration, the decision is made to implement pre-control on KPCs and an outgoing product-sampling program.

    Training is to be a key to success. Special pre-control charts are created with the traditional zones colored for easy detection. Next, a course curriculum is developed in about four hours. Several classes are conducted of 12 to 15 students, which include supervisors and machinists in the methodology. Class length is less than one hour and the machinists are able to begin using the charts immediately without other support.

    A Little More

    Pre-control charts are not without controversy. In standard control charts, control limits are based on common cause variations. With pre-control, the zones are based on specification tolerances. For processes with lower capabilities, this could result in higher type I errors with pre-control charts (about 2%) when compared to standard control charts, which have a fixed type I error of 0.27%.

    The danger with higher type I errors is that more false alarms result in unnecessary process adjustments. However, with higher capability, type I errors are reduced to about the same as with standard control charts.

    Although pre-control is very simple to use, it is not a substitute for control charts. The purpose of control charts is to monitor processes to detect the presence of assignable causes. Pre-control is a simple technique that helps to prevent the manufacture of defective parts. It does not require any sophisticated control charting.

    Dorian Shainin understood the value of understanding the capability of processes. When needed, process capability studies should be conducted using a statistical approach. However, Shainin recommended that technicians use pre-control wherever possible as a means to limit the manufacture of defective parts and to protect against delivery of poor product to customers. Pre-control is a simple, straightforward tool that can pay big dividends very quickly.

    Tech Tips

    Pre-control begins with the assumption that the process is centered within the middle half (50%) of the specification limits, detecting shifts that might result in parts falling outside the specification limits. Pre-control tells the machinists:

  • when the process can be run (after five consecutive pieces pass).
  • when to adjust the process.
  • when sampling is safe to use.
  • when the process needs attention.
  • when to leave the process alone.