Technically, any value between 0°C and 50°C is allowed and would be transmitted and displayed on the app as core body temperature.
In practice, the algorithm will only predict core body temperatures which it has seen during training – since we train with real-world data, these are in the physiological range (35°C – 42°C). Work is being done to extend these regions, particularly in the lower ranges.
A caveat is if the algorithm is fed invalid data (e.g. when it is not worn on the body while laying on the table) it could lead to “extreme” (non-physiological) temperatures which are, of course, invalid as well. The off body scenario is a known issue for which a fix is coming in the near future via direct off body detection.
The insecurity about the value, especially when using invalid data, will be reflected in the data quality by varying colors of the plot (bad = light gray, best =dark green).