CUSUM and EWMA Charts
You'll learn how CUSUM and EWMA charts detect small, sustained shifts in process behavior that traditional Shewhart control charts might miss.
The Challenge with Traditional Control Charts
Shewhart control charts (like X-bar charts) excel at detecting large, sudden shifts in a process. However, they're less sensitive to small, gradual drifts or persistent changes that accumulate over time. That's where CUSUM and EWMA charts shine.
CUSUM Charts: Cumulative Sum
CUSUM tracks the cumulative sum of deviations from a target value. Think of it like monitoring your bank balance: a single small withdrawal might not alarm you, but if you're consistently spending $20 more than you earn each day, your cumulative debt reveals a problem quickly.
The CUSUM chart accumulates these small deviations. When the cumulative sum crosses a decision threshold (upper or lower limit), you've detected a sustained shift. It's especially useful for catching:
- Gradual process drift
- Small but consistent changes in mean
- Shifts of 0.5 to 2 standard deviations
EWMA Charts: Exponentially Weighted Moving Average
EWMA applies weighted averaging where recent observations matter more than older ones. It smooths the data while remaining responsive to changes. The smoothing parameter (lambda, λ) controls sensitivity:
- Small λ (e.g., 0.2): More smoothing, detects smaller shifts
- Large λ (e.g., 0.8): Less smoothing, faster response
Like a weather forecast that weighs today's temperature more heavily than last week's, EWMA adapts to evolving patterns while filtering noise.
When to Choose Which
Use CUSUM when you need to detect the exact time a shift began. Use EWMA when you want smoother monitoring with tunable sensitivity to different shift magnitudes.
Key Takeaway: CUSUM and EWMA charts are specialized control charts designed to detect small, sustained process shifts that Shewhart charts might miss, using cumulative sums and weighted averages respectively.