The interpretation layer uses about eight terms with specific meanings. They are worth defining once, in one place, so the AI and the reader use them the same way.
Baseline: the rolling 24-hour median of a parameter in your space. It adapts to your normal but is robust to brief excursions. The VOC and NOâ‚“ indexes use Sensirion's own 24-hour normalization, which is conceptually equivalent. Elevated: above the 75th percentile of your last 30 days, or above an absolute health threshold, whichever is higher. A soft signal. Spike: a change of more than 2 standard deviations from the rolling mean within 5 minutes. A real event, not sensor jitter.
Trend: a sustained directional movement for 6 or more hours. Overnight COâ‚‚ buildup is a trend; a 10-minute VOC pulse is not. Anomaly: a pattern that does not match any entry in the fingerprint catalog. Logged but not interpreted. Fingerprint: the multi-parameter signature of a known scenario (cooking, sleep, cleaning, wildfire infiltration).
Cross-reference: the dashboard checked your indoor reading against the outdoor feed (from the Google Air Quality API) and, where applicable, the Open-Meteo weather forecast. Confidence: how well the current signal matches the AI's expectation. Low confidence does not mean wrong; it means the data is ambiguous, sparse, or near the boundary of the model's domain. The AI surfaces low-confidence calls explicitly rather than hiding them.
References
- Sensirion - SEN66 datasheet and VOC index info sensirion.com
- Sensirion - NOx index info note sensirion.com
- EPA - AirNow: AQI Basics www.airnow.gov
- WHO Global Air Quality Guidelines (2021) www.who.int