What Is a Multi-Vari Chart and How Does It Help You Find the Real Source of Process Variation?

If your process is producing inconsistent results, the first question is not "how do I fix it?" but "where is the variation coming from?" A multi-vari chart is a simple yet powerful graphical tool that helps you answer that question before you dive into complex statistical analysis.

What It Is

A multi-vari chart is a visual analysis tool used in quality engineering and statistical process control (SPC) to break down total process variation into three main components:

• Within-unit (or within-batch) variation – variation that occurs within a single product, batch, or time period.
• Between-unit (or between-batch) variation – variation that occurs from one unit or batch to another.
• Over-time (or temporal) variation – variation that occurs as time passes, such as shifts, days, or production runs.

The chart was popularized in the 1950s by Leonard Seder and is widely described in standard quality engineering texts (e.g., Statistical Quality Control by Grant & Leavenworth, and World Class Quality by Keki Bhote). It is a standard tool in the "Analyze" phase of DMAIC (Define, Measure, Analyze, Improve, Control) projects.

How It Works: Steps and Logic

The multi-vari chart is built from a sampling plan that deliberately captures variation at different levels. A typical plan looks like this:

1. Select time periods – e.g., 3 shifts, 5 days, or 4 production runs.
2. Within each time period, select multiple batches or units – e.g., 3 batches per shift.
3. Within each batch, take multiple measurements – e.g., 5 consecutive parts or 3 positions on a single part.

Then you plot the data as follows:

• X-axis: Time order (shifts, days, runs).
• Y-axis: The measured quality characteristic (e.g., dimension, weight, temperature).
• For each time period, you plot:

- A vertical line connecting the minimum and maximum of all measurements in that period (this shows total variation for that period).
- A short horizontal line or marker for the average of each batch.
- Small dots or lines for each individual measurement, often connected within a batch to show within-batch spread.

By visually comparing the vertical spread of the dots within each batch, the vertical spread of batch averages within a time period, and the movement of the overall averages across time periods, you can judge which source dominates.

A Worked Illustrative Example

Example data (illustrative only):
Suppose you measure the outer diameter (in mm) of a machined shaft. You sample 3 shafts per hour, for 4 hours.

Hour | Shaft 1 | Shaft 2 | Shaft 3 | Hourly Average

• 1 | 10.02 | 10.05 | 10.01 | 10.027
• 2 | 10.03 | 10.04 | 10.06 | 10.043
• 3 | 10.01 | 10.02 | 10.03 | 10.020
• 4 | 10.05 | 10.07 | 10.06 | 10.060

Plotting the multi-vari chart:

• For Hour 1, draw a vertical line from 10.01 to 10.05 (range = 0.04). Mark the average at 10.027.
• For Hour 2, range = 0.03 (10.03 to 10.06), average = 10.043.
• Hour 3, range = 0.02, average = 10.020.
• Hour 4, range = 0.02, average = 10.060.

Interpretation (illustrative):
The within-hour ranges are small (0.02–0.04), so within-batch variation is low. But the hourly averages move from 10.020 to 10.060 — a spread of 0.040, which is as large as the within-hour range. This suggests that time-based (shift-to-shift) variation dominates. You would then look for causes that change hour by hour: tool wear, operator change, material lot change, or ambient temperature drift.

Common Pitfalls

• Sampling too narrowly – If you only sample from one batch or one time period, you cannot separate sources of variation. The chart is only as good as your sampling plan.
• Over-interpreting small patterns – A multi-vari chart is a screening tool, not a formal hypothesis test. Use it to form hypotheses, then confirm with ANOVA or control charts.
• Ignoring measurement error – If your measurement system itself is noisy (low Gage R&R), the within-unit spread will be inflated and mislead you.

Closing

A multi-vari chart is one of the fastest ways to see where your process variation lives — and it requires no complex calculations. To build one in seconds from your own data, try the free multi-vari chart tool at https://www.6sq.com/tools/multivari/ and start separating signal from noise today.