Control charts on data nobody is managing

The frequency of the British electricity grid, live. Nobody inspects it, it just varies — so it shows what "normal variation" actually looks like.

Frozen limits on live data — and what happens to a chart when you sample too often.

A river drifts with the weather. Rain shows up as a run of rising points long before any alarm threshold is crossed.

Runs and trends: a shift caught by its pattern, not by a threshold.

Counts, not measurements. Same logic, different maths — and using the wrong chart is why most dashboards cry wolf.

Counts need a count chart: Poisson limits, no lower tail, no range panel.

Five European cities sampled on the same hour. The subgroup separates place-to-place spread from a real, continent-wide event.

Rational subgrouping: the spread inside a subgroup is the yardstick for movement between them.

Real hardware, real drift, real dropouts. A slow drift is invisible on a gauge and obvious on a control chart.

Slow drift, dropouts, and why a sensor that reports every minute should not be charted every minute.