What Is an EWMA Chart? Definition and Uses

An EWMA chart — exponentially weighted moving average — is a control chart in which each plotted point blends the newest measurement with a weighted memory of everything that came before, with recent history counting most. Where a classic Shewhart chart judges every point in isolation, the EWMA accumulates evidence across points. That single design change makes it markedly better at detecting small, sustained shifts — the slow drifts that a standard chart can fail to flag for a long time.

How it works

The chart is governed by a smoothing constant, lambda, set between zero and one, which decides how fast the memory fades. A small lambda — values around 0.1 to 0.3 are typical — gives the chart a long memory and high sensitivity to small drifts; a lambda of one forgets the past entirely and reproduces the ordinary individuals chart. Control limits are computed for the smoothed statistic itself, slightly narrower in the early points and settling to steady-state width.

The trade-off is symmetry: what the EWMA gains on small shifts it concedes on large, abrupt ones, where the Shewhart chart’s point-by-point verdict reacts fastest. In practice the two are often run side by side — the Shewhart chart as the fire alarm, the EWMA as the slow leak detector.

A worked example

A plating line deposits a coating whose thickness slowly declines as the bath chemistry depletes — a drift of a fraction of a standard deviation per day, invisible in any single reading. The individuals chart stays silent; every point looks ordinary. The EWMA, weighting the recent run of slightly-low readings together, crosses its lower limit days before any raw measurement looks alarming. The team replenishes the bath on evidence rather than on schedule or on scrap, and the drift never reaches the customer.

• Choose lambda deliberately, before deployment, based on the size of shift you need to catch — not by trial and error afterward.
• Pair the EWMA with a Shewhart chart so large abrupt shifts are still caught immediately.
• It suits individual measurements well, which makes it valuable where subgrouping is impractical — chemical processes, transactional data, long cycle times.
• Train the chart’s readers: each plotted point is a weighted statistic, not a raw measurement, and the difference matters when investigating.
• Treat an EWMA signal like any other special cause signal — timestamp it, investigate it, and log what you find.

A Shewhart chart asks what just happened; an EWMA chart asks what has been happening. Small drifts only answer the second question.

The EWMA chart belongs to advanced statistical process control, which is Black Belt material. Our Black Belt program ($499, about 60 hours) covers time-weighted charts alongside the rest of the advanced toolkit, and the 150-question proctored exam expects you to know when a Shewhart chart is no longer enough.