Common cause describes variation in a process that comes from its normal, everyday operation rather than from any single identifiable, unusual event.
The idea
In statistical process control, variation falls into two categories: common cause (or “chance cause”) variation, the natural noise built into a stable system, and special cause variation, which comes from something specific and identifiable, such as a broken machine or a one-off event. Common cause variation affects every outcome the process produces and can only be reduced by changing the system itself; treating it as a special, one-off cause leads to overreacting to normal noise.
When to use it
- Interpreting fluctuations in metrics, quality data, or performance results before reacting to them
- Deciding whether a problem needs a systemic fix or a targeted, one-off fix
- Building control charts or dashboards that separate signal from noise
How to apply it
- Collect enough data points to see the normal range of variation in the process
- Plot the data over time, for example a control chart, and identify the expected range of common cause variation
- Investigate points that fall clearly outside that range as potential special causes
- For variation within the normal range, improve the system as a whole rather than chasing individual data points
Watch out for
- Treating ordinary common cause variation as a special event leads to overcorrection and adds noise, a mistake W. Edwards Deming called “tampering”
- Ignoring genuine special causes because they get lumped in with normal variation
- Small sample sizes make it hard to tell the two types of variation apart
Related models
- Root cause analysis — finding the underlying cause of a specific, often special-cause, problem
- Multiple causation — the related idea that outcomes often result from several contributing causes, not one
- Controlled experiment — a way to isolate whether a change is a real (special) cause of an effect
Sources
W. Edwards Deming’s work on statistical process control, building on Walter Shewhart’s distinction between common cause and special cause (assignable cause) variation, developed at Bell Labs in the 1920s.