Shift Detection Calculator

Calculator Inputs

Baseline observations

Use stable process data for the reference line.

Monitoring observations

Add the newer sequence you want to test.

Target mean

Process sigma

Control limit width in sigma

CUSUM reference value k

CUSUM decision interval h

EWMA lambda

Shift threshold in sigma units

Example Data Table

Observation	Baseline Value	Monitoring Value	Comment
1	50.1	50.2	Starts near the target.
2	49.8	50.4	Small upward movement begins.
3	50.3	50.5	Mean remains above baseline.
4	50.0	50.7	Shift becomes more visible.
5	49.9	50.8	CUSUM starts accumulating faster.

Formula Used

Mean shift: Shift = Monitoring Mean − Target Mean

Z score: z_i = (x_i − Target Mean) / Sigma

Control limits: UCL = Target Mean + z × Sigma, LCL = Target Mean − z × Sigma

Positive CUSUM: C⁺_i = max(0, C⁺_i−1 + z_i − k)

Negative CUSUM: C⁻_i = min(0, C⁻_i−1 + z_i + k)

EWMA: EWMA_i = λx_i + (1 − λ)EWMA_i−1

This calculator combines classic control chart limits, CUSUM accumulation, and EWMA smoothing. The blend helps reveal both sudden shifts and smaller persistent drifts that a simple average can miss.

How to Use This Calculator

Enter a stable baseline series from a process known to be in control.
Enter the newer monitoring series you want to test for a shift.
Leave target mean and sigma blank to estimate them from baseline data.
Set control width, CUSUM values, and EWMA lambda to match your sensitivity needs.
Press submit to show the result summary above the form.
Review the detailed observation table and export results as CSV or PDF.

Higher lambda reacts faster to recent changes. Smaller k and h make CUSUM more sensitive to smaller shifts, but they may increase false alarms.

Why Shift Detection Matters

Shift detection supports quality control teams when a process average moves away from its intended level. Early identification reduces scrap, protects customers, and helps teams trace root causes before drift becomes expensive.

Using only a single average can hide important evidence. By combining run length, control breaches, CUSUM, and EWMA, this tool offers a broader picture of process behavior across both large and subtle changes.

Frequently Asked Questions

1. What does this calculator detect?

It checks whether a monitored process mean has moved away from its target. It uses control limits, CUSUM, EWMA, and run patterns to flag possible shifts.

2. When should I use baseline data?

Use baseline data when you have a stable reference period. The calculator can estimate target mean and sigma from it, which helps when formal limits are not already set.

3. What is the benefit of CUSUM?

CUSUM is strong at detecting small sustained shifts. It accumulates evidence over time, so weak but persistent movement becomes easier to notice than with simple limit checks.

4. Why use EWMA too?

EWMA gives extra weight to recent data while still considering prior history. It smooths noise and helps you spot gradual drift without reacting too sharply to every point.

5. What values work for k and h?

Many teams start with k = 0.5 and h = 5 for standardized data. These are common settings for moderate sensitivity, though your process may require tuning.

6. Can I enter values separated by spaces?

Yes. The calculator accepts numbers separated by commas, spaces, semicolons, or line breaks, which makes it easy to paste data from many sources.

7. What does first signal observation mean?

It is the earliest monitoring point that triggered a limit breach or a CUSUM alarm. This helps identify when the process likely began shifting.

8. Is this tool for subgroup averages only?

No. You can use individual observations or subgroup means. Just make sure your target and sigma values are consistent with the type of data entered.