Three concepts get conflated whenever someone asks "is this reading accurate." Accuracy is how close a reading is to truth, the gap between what the sensor reports and what a reference instrument would say. Precision is how repeatable the reading is, two measurements of the same air five minutes apart should match. Drift is how much one of those numbers slides over weeks, months, or years, even when the underlying air is unchanged. A sensor can be precise but inaccurate (consistently 50 ppm high on CO2), accurate but imprecise (centered on truth but noisy), or stable today and drifted next year. The dashboard exposes all three differently.
Per-channel accuracy on the Terrestream stack, taken from the Sensirion SEN66, Bosch BMP390L, and TI OPT3001 datasheets: PM2.5 is ±10 µg/m³ or ±10% of reading, whichever is larger, so a reading of 15 µg/m³ carries roughly ±10 µg/m³ uncertainty. CO2 is ±50 ppm + 5% of reading, so an 800 ppm reading carries about ±90 ppm. RH is ±2% between 20 and 80%. Temperature is ±0.5 °C between 15 and 30 °C. Pressure is ±0.5 hPa (50 Pa). Lux is ±20% typical, which is at the better end of consumer ambient-light sensors.
The VOC and NOx indexes from 1 to 500 are a different category. They are not absolute concentrations and have no datasheet-level accuracy claim, because the underlying MOX chemistry is non-specific by physics. What they do have is good repeatability: day-to-day in stable conditions the index typically holds within ±10 units, and the Sensirion gas-index algorithm normalizes against a rolling baseline of the room's own air. The number means "different from this room's normal," not "this many parts per billion of benzene." See optical CO2 vs MOX sensing for the physics behind why.
Where the reference-grade tier sits, for scale: regulatory EPA Federal Reference Method beta-attenuation monitors hit roughly ±1 µg/m³ on PM2.5, and laboratory NDIR + DOAS analyzers hit ±5 ppm on CO2. Terrestream is one to two orders of magnitude less precise than that, which is appropriate: those instruments cost tens of thousands of dollars and need climate-controlled enclosures and weekly calibration. The practical consequence on the dashboard is that you should not over-react to a 3 µg/m³ change in PM2.5, that is inside the noise floor; do react to a 30 µg/m³ change, that is clearly above it. Cross-sensitivities and drift are the other two pieces of the same picture.
References
- Sensirion - SEN66 environmental sensor module sensirion.com
- NIST - Indoor air quality sensors www.nist.gov
- EPA - Air monitoring methods for criteria pollutants www.epa.gov
- WHO Global Air Quality Guidelines (2021) www.who.int
