EWMA Control Chart: The Most Sensitive Chart for Small Shifts
What is the EWMA Control Chart?
The EWMA chart weights historical observations with exponentially decaying weights: z_i = λx_i + (1−λ)z_{i−1}, where λ is the smoothing constant (0 < λ ≤ 1). Smaller λ gives history more weight and makes the chart more sensitive to small shifts. The pointwise variance is σ²·λ/(2−λ)·[1−(1−λ)^{2i}], so the control limits start narrow and widen toward the steady-state value σ²·λ/(2−λ).
When to Use It
Use EWMA when a process mean shifts slowly within about 1.5σ: continuous chemical or pharmaceutical processes, individual measurements (one per batch) with σ estimated from moving ranges, and sensitive monitoring of large automated samples. λ is typically 0.05–0.3 with limit width L = 3, and L can be lowered for extra sensitivity. Because EWMA is insensitive to parameter-estimation error, it works well as a complement to conventional charts.
How to Use It (Step by Step)
Paste the observations in time order, set λ and the limit width L, and the tool computes pointwise control limits and marks any excursion. Read the AI interpretation of the shift direction, magnitude and likely causes, then confirm the special cause on the floor before acting. Remember that early points sit inside narrower limits by design — that is normal, not an anomaly.
Key Formulas / Example
z_i = λx_i + (1−λ)z_{i−1}; control limits μ0 ± L·σ√(λ/(2−λ)(1−(1−λ)^{2i})) with σ estimated from MR̄/d2. Example: λ = 0.2, σ = 1, L = 3 gives steady-state limits of about ±1.0σ around the target, reached after roughly 20–30 points.