Process Improvement: Problem Solver

1. Diagramming all the steps in a production process (flow chart) so every expected happening is understood;2. Measuring results of the production process (cause and effect diagram); and 3. Implementation of the Plan-Do-Study-Act (PDSA) cycle to refine processes and improve results.The primary diagramming tools are the flow chart and the cause-and-effect diagram. Useful measurement tools, some may

1. Diagramming all the steps in a production process (flow chart) so every expected happening is understood;

2. Measuring results of the production process (cause and effect diagram); and 3. Implementation of the Plan-Do-Study-Act (PDSA) cycle to refine processes and improve results.

The primary diagramming tools are the flow chart and the cause-and-effect diagram. Useful measurement tools, some may be unfamiliar, include the check sheet, the pareto chart, the histogram, scatter diagrams, run charts and control charts. The use of these tools to measure production processes and the application of statistical analysis to the measurements is known as Statistical Process Control (SPC).

Tools, Techniques Explained To illustrate, let's say we've identified a reproduction problem that is being expressed as low farrowing rate. Artificial insemination is used.

Flow charts, the pictorial representations of a process, have three steps:

1. Accurately draw all the steps that actually occur in a process (Figure 1);

2. Draw a flow chart of the steps the process should follow if everything is being done correctly; and

3. Compare the two charts to find where they are different.

Cause-and-effect diagrams, also called "fishbone" diagrams, are used to help identify all potential causes of a specific problem (Figure 2). The effect or problem is listed on the right side of the chart; the major influences or causes are listed on the left. The causes are usually categorized as: people, machine, method or material.

When constructing the cause-and-effect diagram, causes must be well defined so that the most likely causes can be selected for further analysis. Since the usual tendency is to attribute causation much too easily, other tools are necessary to determine causation vs. association.

Check sheets are the simplest data-gathering forms and are the logical starting point in many problem-solving situations (Figure 3). They begin the process of translating opinions into facts, identifying prevalence.

Pareto charts are used to display the relative importance of the problems (Figure 4). They deal only with characteristics of a product or service. They are a special form of a vertical bar graph that ranks problems by frequency of occurrence. It is important to remember, however, that the most frequent problems are not necessarily the most costly. Re-ranking based on cost may be more appropriate than frequency of occurrence in many cases.

Histograms go a step further than pareto charts by displaying the distribution of measurement data over time (Figure 5). This tool can begin to show the variation inherent in any process. Histograms help visualize variability (the range of outcomes) and skewness (whether outcomes are weighted on one side or another of the mean).

Scatter diagrams are used to analyze two variables to determine their relatedness (Figure 6). Scatter diagrams show possible cause-and-effect relationships and the strength of those relationships. Statistical tests can be applied to scatter diagrams to determine exact levels of correlation.

Run charts are the simplest way to chart a process over time (Figure 7). They are useful for displaying potential trends and observing long-range averages. A danger in using run charts is the tendency to see minor or normal variation as being significant.

Some simple rules can be used with run charts that don't require sophisticated statistical analysis. When nine consecutive points run on one side of the average this indicates that a potentially significant event has occurred or that the average has changed. When six consecutive points are either increasing or decreasing with no reversal, regardless of where they fall in relation to the average, it is unlikely that the change is due to chance. Finally, if 14 points in a row are alternating up and down, the occurrences are not likely due to chance.

Control charts are run charts with statistically determined upper control limits (UCL) and lower control limits (LCL) - three standard deviations plotted on the chart (Figure 8).

Control charts help determine how much of the variability in a process is due to random variation and how much is due to unique events or individual actions. Control charts can be refined by calculating "zones" based on 1, 2 or 3 standard deviations.

With control charts it is possible to begin determining chance occurrence vs. real trends, whether a system is "in control" or "out of control" and whether "common causes" (system causes) or "special causes" (human error or unique events) are responsible for suboptimal performance.