Patterns of Variation

Variation is the language of processes. Every output you measure—cycle time, defect rate, cost, yield—carries the imprint of the forces acting on it. In the Analyze phase, your job is to interpret that imprint with clarity and discipline. Patterns of variation help you distinguish between noise and signal, between what is inherent to the process and what is being introduced by specific conditions, shifts, or subgroups. When practitioners skip this step, they often chase symptoms instead of causes. When they do it well, the path to root cause becomes far more direct. 

At its core, variation comes from two broad sources: common cause and special cause. Common cause variation is the natural, expected fluctuation built into the system. It reflects the current design, the current methods, the current environment. Special cause variation, on the other hand, is introduced by something unusual—an event, a shift, a change in materials, a specific operator, a particular machine. The Analyze phase is where you learn to separate these two with confidence. 

Patterns of variation show up in several ways. You may see differences across time, across tools or machines, across operators, across materials, or across environmental conditions. You may see patterns that repeat, patterns that drift, or patterns that spike. Each pattern tells a story about how the process behaves and where to look next.

A disciplined approach begins with stratification—breaking the data into meaningful subgroups. Instead of treating all data as one homogeneous mass, you examine it by shift, supplier, product type, machine, or any other factor that could influence performance. This simple act often reveals differences that were previously hidden. Averages can conceal a great deal; stratification exposes it. 

From there, you look for structure. Are certain subgroups consistently higher or lower? Are some more variable? Do patterns align with known process conditions? This is where tools like multi-vari charts, boxplots, and time‑series plots become invaluable. They help you visualize how variation behaves; not just how much variation exists. 

Patterns of variation also help you avoid false conclusions. Without understanding the underlying structure, it’s easy to misinterpret data. For example, a process may appear stable overall, but when stratified by operator, one operator may show significantly higher variation. Or a process may appear to have a trend, but when stratified by product type, the trend disappears because it was driven by a shift in product mix. 

In the Analyze phase, your goal is not simply to describe variation but to interpret it. You want to understand what the patterns imply about the process, what they suggest about potential causes, and where they direct your investigation. Patterns of variation are not the answer—they are the map that leads you to the answer. 

When you approach variation with this mindset, you move from reacting to data to truly understanding it. You begin to see the process as it is, not as you assume it to be. And that clarity is what makes the Analyze phase so powerful.