Key takeaways
Short answer: Common cause and special cause variation are the two fundamentally different kinds of variation in any process, and treating one as the other is a classic, costly mistake. Common cause variation is the inherent noise of a stable process — the sum of many small, random influences that are always present. Special cause variation is a specific, assignable disturbance — something changed. A process showing only common cause is in statistical control; a special cause signals something to investigate. Control charts exist precisely to tell them apart. For the related quality distinction, see precision vs accuracy.
Common cause variation — sometimes called natural or random variation — is the inherent noise of a stable process. It comes from the many small, ever-present sources of variation built into the way the process runs: slight differences in material, minor temperature drift, normal play in equipment, small human variability. No single source dominates; together they produce a stable, predictable band of variation around the average. A process exhibiting only common cause variation is said to be in statistical control: it is stable, and while you cannot predict any single output exactly, you can predict the range it will fall within. Common cause variation is a property of the system itself, and reducing it requires changing the system, not reacting to individual points.
Special cause variation — also called assignable variation — comes from a specific, identifiable disturbance that is not part of the normal process. Something changed: a new batch of out-of-spec material, a tool that broke, an untrained operator, a setting knocked out of adjustment, a machine fault. Special causes produce variation that stands out from the normal band — a point beyond the control limits, a sudden shift, an unnatural trend. Unlike common cause, a special cause has a findable, assignable reason, which means it can be investigated and eliminated at the source. A process with special cause variation present is unstable and unpredictable until that cause is found and removed.
The distinction defines whether a process is in control. Only common cause variation present means the process is stable and predictable — in statistical control. A special cause present means the process is unstable — its output is no longer predictable until the disturbance is removed. This matters enormously for how you should respond. A special cause is a signal to stop and investigate a specific, findable reason. Common cause is a signal that the variation is built into the system, and the only way to reduce it is to improve the system as a whole. Mistaking one for the other leads directly to the most common and damaging error in process management.
A filling line targets 500 ml. Day to day, individual bottles read 498, 502, 499, 501 — bouncing randomly within a tight band around 500. That is common cause variation: the normal noise of a stable process, and reacting to each wobble by nudging the setting would only make things worse. Then one afternoon several bottles in a row read 510, 512, 511 — a clear shift outside the normal band. That is a special cause: investigation finds a valve drifted out of adjustment. The fix is specific — recalibrate the valve — and the process returns to its stable band. The skill was not reacting to the random 498-to-502 noise, but reacting decisively to the assignable 510-plus shift.
The most expensive mistake in process management is treating common cause variation as if it were special — adjusting the process in response to normal random noise. This is called tampering, and it does not reduce variation; it amplifies it. Every time an operator nudges a setting to chase a random high or low reading, they inject a new disturbance into a process that was actually stable, widening the very variation they were trying to control. The mirror-image error is ignoring a genuine special cause as if it were just noise, letting an assignable, fixable problem run. Both errors come from failing to distinguish the two kinds of variation — which is exactly the problem control charts were invented to solve.
The distinction underpins the trustworthiness of the data behind OEE. Performance and quality both vary; knowing whether a dip is common cause noise or a special cause signal is what stops you from overreacting to normal fluctuation or missing a real problem. A run of micro-stops that is just common cause noise calls for systemic improvement, while a sudden special-cause spike calls for immediate investigation — and treating them the same wastes effort or misses the fix. Reading OEE trends with this lens connects directly to the six big losses: distinguish the chronic, built-in losses (common cause) from the sporadic, assignable ones (special cause), because they need different responses.
Fabrico trends performance, downtime, and quality over time, which is what lets a team separate the steady background noise of a stable process from the genuine signals worth investigating. Instead of reacting to every fluctuation, you can see when a metric has truly shifted — a special cause to chase — versus when it is just varying within its normal band, where the answer is systemic improvement rather than a knee-jerk adjustment. That discipline prevents tampering and focuses effort where it pays. Book a demo to see your process variation in context.
Common cause variation is the natural, inherent noise of a stable process from many small random sources. Special cause variation is a specific, identifiable disturbance from outside the normal process. Common cause is built into the system; special cause is assignable and can be eliminated.
A process is in statistical control when only common cause variation is present — it is stable and predictable within a known band. When a special cause appears, the process becomes unstable and unpredictable until the cause is found and removed.
Tampering is adjusting a stable process in response to common cause (random) variation, as if each wobble were a special cause. It does not reduce variation — it injects new disturbances and makes variation worse. Avoiding it requires recognising common cause noise.
Usually with a control chart, which plots the process over time with statistically derived limits. Points within the limits and showing no unnatural patterns indicate common cause; points beyond the limits or unnatural patterns signal a special cause to investigate.
Knowing whether an OEE dip is common cause noise or a special cause signal prevents overreacting to normal fluctuation and prevents missing real problems. It separates chronic built-in losses from sporadic assignable ones, which need different responses.
Login
Blog
