Previous blogs described issues with traditional AQL testing. In this blog, I suggest using a 30,000-foot-level predictive measurement system methodology in lieu of AQL testing whenever possible. A test that samples and makes process statements from a 30,000-foot-level metric vantage point can provide more insight with less test effort than AQL testing. Later blogs will describe how this methodology can be used not only in manufacturing but as the reporting methodology for an organizational-wide scorecard system, which is a component of an overall Integrated Enterprise Excellence (IEE) enhanced business management system.

To visualize 30,000-foot-level metric reporting, consider the view from an airplane. When the airplane is at an elevation of 30,000 feet, passengers can see a big picture view of the landscape. However, when the airplane is at 50 feet during landing, passengers view a much smaller portion of the landscape. Similarly a 30,000-foot-level control chart gives a macro view of a process Key Process Output Variable (KPOV) or Y, while a 50-foot-level control chart gives more of a micro view of some aspect of the process, i.e., Key Process Input Variable (KPIV) or X of the process in the relationship Y=f(X).

Reporting at the 30,000-foot-level can characterize the time-series behavior of an AQL characteristic or Y variable response, along with a predictive statement, when appropriate. In 30,000-foot-level reporting, subgrouping intervals are infrequent so that short-term variations, which might be caused by typical KPIV swings, will result in charts that view these perturbations as common-cause variability. As an alternative to AQL testing, each lot might become a subgroup within 30,000-foot-level reporting.

It is not the intent of the 30,000-foot-level control chart to provide timely feedback for process intervention and correction, as traditional control charts do. Example 30,000-foot-level metrics are lead time, inventory, defective rates, and a critical part dimension. There can be a drill down to a 20,000-foot-level metric if there is an alignment; e.g., the largest product defect type. A 30,000-foot-level individuals control chart can reduce the amount of organizational firefighting when reporting operational metrics. As a business metric, 30,000-foot-level reporting can lead to more efficient resource utilization and less playing games with the numbers.

In an IEE organization, we might also describe the drill down of 30,000-foot-level KPOV’s measures to other high-levels such as 20,000-foot-level, 15,000-foot-level, and 10,000-foot-level, noting that the 50-foot-level is reserved for the KPIV designation.

In training sessions, control charts are typically taught to identify in a timely fashion special causes within the control of a process at a low level; within IEE, I would describe this as the 50-foot-level. An example of this form of control in IEE is to identify in a timely way when temperature changes to an unacceptable, predetermined level. When this occurs, the process is adjusted so that the temperature variable problem is fixed before a large amount of product with unsatisfactory characteristics is produced. In IEE, pre-control charts are often a better charting choice than traditional control charts to identify when a timely process adjustment needs to be made so that the 30,000-foot-level metric remains satisfactory. These 50-foot KPIVs need close monitoring and their allowed tolerances could have been determined from a process improvement project.

Control charts are used as part of a high-level scorecard/dashboard, where this emphasis can lead to a reduction in firefighting the problems of the day. These charts can change the culture so that there is less common-cause variation issues attacked as though they were special cause. IEE can create a measurement system scorecard/dashboard so that fire preventive actions are created to address common-cause issues where products do not consistently meet specification needs, in lieu of the day-to-day firefighting of noncompliance issues.

Unlike 50-foot-level control charts, 30,000-foot-level reporting suggests infrequent subgrouping/sampling to capture how the process is performing relative to overall customer needs. When the sampling frequency is long enough to span all short-term process noise inputs such as raw material differences between days or daily cycle differences, I refer to this high-level control chart reporting as a 30,000-foot-level control chart. At this sampling frequency, we might examine only one sample or a culmination of process output data, plotting the response on an individuals control chart.

When someone is first introduced to this concept of infrequent subgrouping/sampling, a concern typically expressed is that this measurement does not give any insight to what we should do to improve. This observation is true; the purpose of this measurement is not to determine what should be done to improve.

The two intents for this infrequent subgrouping/sample approach are:

  • To determine if we have special-cause or common-cause conditions from a 30,000-foot-level vantage point.
  • To compile and analyze data so that they provide a long-term view of the process capability/performance metric of our process relative to meeting the needs of customers.

    At the initial state of evaluating a situation, this approach does not suggest that practitioners get bogged down trying to collect a great deal of data with the intent of identifying a cause-and-effect relation that can fix their problem. The methodology does suggest that this form of data collection should occur after a problem has been identified relative to meeting either an internal or external customer requirement KPOV, where the improvement team identifies the prioritization of effort for further evaluation; e.g., a supplier is consistently producing 3% defectives, as described through 30,000-foot-level metric reporting, and we need to work with the supplier to determine what it can do to improve its process so that the 30,000-foot-level control chart shifts to a new level of stability around 1%.

    The next few blogs will elaborate more on traditional process metrics, their issues and resolution.

    Reference: The content of this blog was taken from Section 12.2 of IEE Volume III