Formula used
Western Electric rules implemented
- Rule 1: any point with |x − CL| > 3σ.
- Rule 2: within any 3 consecutive points, at least 2 are beyond +2σ or −2σ.
- Rule 3: within any 5 consecutive points, at least 4 are beyond +1σ or −1σ.
- Rule 4: any run of 8 consecutive points all above or all below the centerline.
How to use this calculator
- Paste your measurements into Data points, keeping time order.
- Leave Centerline and Sigma on auto for quick checks.
- Use manual values when you have a fixed target or verified sigma.
- Select the rules you want to enforce for your signal policy.
- Press Submit and review the violations table above.
- Export results to CSV/PDF for audits, meetings, or CAPA notes.
Example data table
Use the “Load example data” button to try these points quickly.
| Index | Point | Comment |
|---|---|---|
| 1 | 10 | — |
| 2 | 10.1 | — |
| 3 | 10.2 | — |
| 4 | 10.3 | — |
| 5 | 10.8 | Sustained high run may trigger Rule 4 |
| 6 | 10.9 | Sustained high run may trigger Rule 4 |
| 7 | 11 | Sustained high run may trigger Rule 4 |
| 8 | 11.1 | Sustained high run may trigger Rule 4 |
| 9 | 11.2 | Sustained high run may trigger Rule 4 |
| 10 | 11.3 | Sustained high run may trigger Rule 4 |
| 11 | 10.2 | — |
| 12 | 10.1 | — |
Operational guidance
Signal detection in control charts
Western Electric rules convert visual chart behavior into repeatable triggers. Rule 1 flags extreme points beyond 3σ, which are rare under stable variation. Rules 2 and 3 capture smaller but persistent shifts by checking short windows: two of three beyond 2σ, or four of five beyond 1σ, on the same side. Rule 4 detects sustained bias with eight consecutive points above or below the centerline.
Understanding zones and sigma
The calculator uses your selected centerline (CL) and sigma (σ) to form zones at ±1σ, ±2σ, and ±3σ. When σ is estimated from a small sample, the limits can be noisy, so increasing the number of points improves stability. If your subgrouping strategy changes, update σ accordingly. For high‑resolution measurements, confirm that rounding does not artificially compress σ and inflate zone-based signals.
Interpreting overlapping violations
Because the rules use sliding windows, a single shift can trigger multiple findings across adjacent windows. This calculator de-duplicates identical point sets, but you may still see different windows reporting the same underlying event. Treat clustered violations as one investigation unless evidence suggests multiple causes. Start with the earliest violating point index; it often aligns with the time a special cause entered the process.
Setting centerline and sigma in practice
Auto mode calculates CL as the arithmetic mean and σ as either sample (n−1) or population (n). Use manual CL when you have a validated target, such as a nominal dimension, and manual σ when a qualified measurement system study has established repeatability. If your data contains known out-of-control periods, calculate CL and σ from a stable baseline segment before running the rule checks.
Exportable reporting for audits
Consistent documentation helps during audits and CAPA reviews. The CSV export preserves the rule label, side, threshold, point indices, and values for easy traceability. The PDF export produces a compact report that can be attached to nonconformance records or shift handovers. Include the chosen σ method, the number of points (N), and the decision outcome to keep comparisons consistent across runs. For trend reviews, save reports weekly and compare violation counts by line, product, and shift to identify recurring patterns early in teams.
FAQs
1) What data length is recommended for reliable signals?
More points improve sigma stability. For ongoing monitoring, 20–30 sequential points is a practical minimum for a baseline. Shorter series still work, but rule coverage is limited and limits may be sensitive to a few values.
2) Should I use sample or population standard deviation?
Use sample (n−1) when your points are a subset of ongoing production. Use population (n) when the data represents the full population for the period you want to model. Consistency matters more than the choice.
3) When should I enter a manual centerline?
Use a manual centerline when a fixed target is required, such as a specification nominal or a calibrated setpoint. This avoids shifting CL due to recent drift and helps you detect bias relative to the target.
4) Why do I see multiple violations from one shift?
Rules 2–4 use overlapping windows, so one special cause can trigger several adjacent windows. Review the earliest violating index and treat the cluster as a single event unless process evidence indicates separate causes.
5) What if sigma is nearly zero?
Near-zero sigma makes zone limits extremely tight, causing frequent triggers. Confirm measurement resolution, remove constant placeholders, and ensure you have real variation in the data. Consider manual sigma from a validated study if appropriate.
6) Do these rules replace engineering judgment?
No. They are standardized alarms that prioritize review. Always confirm the context: recent changeovers, tooling, material lots, operator shifts, and measurement system performance. Use the violations to focus investigation, not to assign blame.