CO₂ as an air quality indicator
Elevated indoor CO₂ is a reliable proxy for poor ventilation and the build-up of exhaled air.* For every 400 ppm above the outdoor baseline (~420 ppm), roughly 1% of the room air consists of air previously exhaled by occupants. Most recent scientific recommendations suggest keeping indoor CO₂ below 800 ppm. Because CO₂ rises with occupancy and falls when fresh air is supplied, it captures how well a space is ventilated relative to how many people are using it.
* In rare cases, other CO₂ sources within a building (e.g. gas appliances or other combustion processes) can contribute to elevated readings and affect results.

Median
The median is the middle value when all measurements are sorted from lowest to highest — half the values fall below it and half above. It is more representative than the arithmetic mean for CO₂ data, which is often skewed by occasional very high readings.

Percentiles and the box (IQR)
The 25th percentile (bottom of the box) means 25% of measurements were below that level; the 75th percentile (top of the box) means 75% were below it. The box therefore spans the middle 50% of all values — the difference between these two percentile values is called the interquartile range (IQR) and shows where the middle half of locations sit.

Whiskers
The lines extending above and below the box each reach the most extreme actual data point that is still within 1.5× the IQR from the box edge — so the whisker tip is always a real measurement, not a calculated boundary. They show the range of values that are spread out but not extreme enough to be considered outliers.

Outliers
Any individual measurement that falls beyond the whiskers is plotted as a separate dot. These are rare readings that differ vastly from the rest of the distribution.

n = number of locations (or visits)
In most split modes, each data point represents one distinct physical location (identified by its OpenStreetMap element). Multiple visits to the same place are averaged into a single value, so n reflects how many different locations contributed — not how many individual measurement sessions were recorded. The exception is Split by: Single Locations, where each bar is already one location and n counts the individual visit averages that make up its box.