Jan:1.42. Pick Run chart (median centerline; needs ≥ 10–12 points to test for signals) or SPC I-chart (individuals chart, mean ± 3σ from the average moving range). Press Plot & analyze. The tool draws the chart, highlights any special-cause points in red, and tells you in plain language whether to investigate a cause or leave the process alone. Everything runs in your browser; no data is stored or transmitted.
Plot Your Metric
The Special-Cause Rules Applied
| Rule | What it detects | Chart |
|---|---|---|
| Shift | ≥ 6 consecutive points all on one side of the median (points exactly on the median are skipped) | Run & SPC |
| Trend | ≥ 5 consecutive points all increasing or all decreasing | Run & SPC |
| Too few / too many runs | The number of times the line crosses the median is outside the range expected by chance (a non-random amount of clustering or oscillation) | Run |
| Point beyond a control limit | A single point outside mean ± 3σ — by itself strong evidence of a special cause | SPC |
| Astronomical point | A point obviously, strikingly different from the rest — flagged for your eye, judged by you | Run & SPC |
When none of these fire, the variation you see is common cause — the process's ordinary noise. You improve a common-cause process by redesigning it, not by reacting to individual points. When one does fire, that is special cause — a signal with an assignable cause worth investigating. Confusing the two is the most expensive mistake a CQI committee makes.
How the Chart Is Built
The run chart uses the median of your values as the centerline and applies the standard probability-based run rules (shift, trend, and the runs test) from Provost & Murray's The Health Care Data Guide and the Institute for Healthcare Improvement (IHI). The SPC I-chart (individuals chart) uses the mean as the centerline and control limits at mean ± 3σ, with σ estimated from the average moving range (σ ≈ MR̄ / 1.128, so the 3σ limits are mean ± 2.66 × MR̄) — the conventional individuals-chart formula. Use a run chart for a quick, assumption-light look and for detecting sustained shifts and trends; use the I-chart when you also want statistically derived control limits to flag single extreme points. Both need enough data — aim for at least 10–12 time-ordered points before trusting the signal tests. Every target discussed here is explained in the QAPI vs. CQI field manual, Section 5.2, and the aggregation layer is the QAPI Scorecard.
