Pin Me

Using Run Charts to Track Indicators

written by: Heidi Wiesenfelder • edited by: Michele McDonough • updated: 5/20/2011

Control charts aren't the only game in town for tracking business and process performance indicators. Find out how a run chart can help you understand the variation in your data and when special causes are present in a process.

  • slide 1 of 3

    About Run Charts and Indicators

    If you are reading this article you probably already know the importance of tracking certain metrics, known as performance indicators or even key performance indicators (KPIs), to understand how your business is performing. Tracking indicators over time allows you not only to view the random variation in your data but also to watch for trends or other signs that performance of a process or set of processes is changing.

    In Six Sigma, the tool of choice for viewing data over time is usually the control chart. However its simpler cousin, the time plot or run chart, provides much of the same information and can be easier to explain to employees and other business leaders.

    A run chart looks similar to a control chart but does not include the control limits. It does include all the data points along with a line representing the average, known as the center line, and any signals of special cause. It will not display outliers as does a control chart, but all the other special cause tests can be run and displayed using a run chart.

    Many programs used by Six Sigma practitioners, such as Minitab or the Excel plug-in QI Macros, can easily create run charts from your data. These programs have options you can set to choose the mean or median as the average, to select which special causes to test for, and more. Or read our article on creating a run chart in Excel.

    As for a control chart, the data for a run chart may be continuous. Numerical data that measures a physical characteristic or any other variable that can take any value along a continuum is known as continuous data. It includes common indicators such as revenue, cost per unit, cycle time, and widget length. Attribute percentage data and some attribute count data may also be appropriate for a run chart. (Read our article on data types if you need a refresher.) In most cases it will be obvious whether your data is appropriate for a run chart, and if you have the wrong type it will simply not make sense. Check with your Master Black Belt or other leader if you need guidance.

  • slide 2 of 3

    Understanding What Your Run Chart Says About Your Indicators

    Using a run chart, you can check for the following types of special cause:

    Too many runs above or below the center line:

    A "run" is a series of data points all on the same side of the center line. Points right on the average do not count as above or below. Depending on how the data are distributed over time, there may be very few total runs or very many. The expected number of runs depends on the number of data points you have. (The highest number possible would be the number of data points you have, if each point alternated above then below the center line and so on.)

    If you do not have software to run the special cause tests you will need to consult a chart that shows the expected number of runs, which will provide you with the minimum and maximum number considered acceptable for the amount of data you have.

  • slide 3 of 3

    If you get a special cause signal indicating you have too few or too many runs, you will need to investigate to find out why your data is varying in that way. Too few runs may mean that data is being collected and reported in batches, for instance with daily data it may indicate that the process being performed one way for a week and then another way the next week. Too many runs may be a sign of overcompensation, with a process being adjusted upward based on a low data point or downward based on a high data point.

    Too many points continuously increasing or decreasing:

    This is a sign of a trend in process performance, which means something is gradually shifting over time. The usual threshold is six points in a row. If your data shows this type of special cause, investigate to find out what is causing a performance shift in your process. If it is moving in the right direction, determine how to leverage it and make it sustainable. If it is moving in the wrong direction, determine how to correct and prevent it.

    Too many points on the same side of the center line:

    This is another sign of a process shift. A typical threshold for this test is eight points in a row. Follow the steps for the previous type of special cause to respond accordingly based on what the data shows.

    Too many points alternating up and down:

    If your data shows a large number of points in a row that are alternately higher and lower than the previous point, you may have bias or sampling problems in your data. Investigate the data collection process to find out what may be going on. The threshold for this test is often set at 14 points in a row.

    Note that for all these special cause tests you can set an option in your software to alter the number of runs or data points that is the threshold for a special cause signal.

privacy policy