Multi-vari uses a specific data sampling plan, which graphically highlights the major variation cause in the output characteristic of your Six Sigma process while allowing the process to operate in its normal fashion and without requiring any process disruptions. The major cause of output variation is isolated into three categories:
Positional
Cyclical
Temporal
When you know which category of variation dominates the output of your process, you can concentrate on potential factors that fall under that category and eliminate factors that belong to the other categories. If you find that the major variation in your process output is coming from a temporal source, you can discount all factors that are positional or cyclical; the true root cause must be a temporal factor.
Positional variation
The positional variation category is sometimes called within unit variation. That’s because it’s defined by the magnitude of variation coming from within a single unit. Differences among these measurements are evidence that a positional variation factor is influencing the output of the process.
You may need to define a “unit” differently for different process situations. The basic requirement for a unit is that the output characteristic must be measureable multiple times at different points on the unit. That may be measuring the same characteristic at different locations on the unit.
Cyclical variation
The cyclical variation category is sometimes called between unit variation. It’s defined by the magnitude of variation that occurs between consecutive units drawn from the process. Large variation between units means that the factor driving process performance must be one that falls under the cyclical category. In the axle example, the magnitude of the variation that you observe in the diameters between consecutively produced axles is cyclical variation.
Temporal variation
The temporal variation category is sometimes called time-to-time variation. When you look at the magnitude of variation between segments of the process separated by a significant amount of time, that is temporal variation. If this type of variation is large, the factor driving process performance must be one that belongs to the temporal category; otherwise, the factor must be from another category.
Here’s the step-by-step procedure for pulling intermittent data from a running process:
Select or establish a continuous-type data measurement of process output performance.
This scale may be in units of time, dollars, inches, grams, but whatever it is, it must be a continuous data type.
Explore the historical values of your selected output metric to understand what the magnitude of variation has been in the process.
After you begin multi-vari sampling of your process, you continue until you’ve observed approximately the same magnitude of variation that you’ve seen historically. That way, you’re sure to have monitored the process long enough to have captured the activity in the input factors that is driving variation in the process output.
Define what constitutes a unit in your multi-vari study.
Remember that your defined unit must allow two or more measurements of the process output in different “locations” within or on the unit.
Collect two to five measurements from within the unit defined in Step 3 on three to five consecutive units.
Allow some time to pass — enough that potential factors have a chance to exert new influence on the process.
Repeat Steps 4 and 5 in three to five consecutive-unit intervals until you’ve captured at least 80 percent of the historical process variation.
Simply compare the range of the historical data to the range of the multi-vari data. If they’re approximately equal, you’ve captured enough multi-vari data. If not, keep collecting.
Create a multi-vari chart and analyze and interpret the chart for a primary source of variation.