I don't have diabetes, so I'm not regularly comparing it to alternatives, so I can't comment on my personal accuracy against clinic measurements. Which is obviously a nice privilege to ignore for "hobby" usage.
A huge portion of glucose management (even for diabetics) is the actual trend of the data vs the actual blood glucose levels. These CGMs capture trends correctly, even if the absolute value is off. A prescription variant would have periodic calibrations as an option. If you eat a meal that spikes blood glucose level, the CGM probably won't smooth out a peak, so you'd still see the magnitude of the effect. The other thing to consider is that when read from the arm or a "fleshy" bit, a change in glucose level is about 20min delayed from blood levels.
The healthy range for a non-diabetic is ~70-140 mg/dL, and your body regularly can swing up more than +50mg/dL after a meal. This is a pretty big range and swing, and being 15% off (in either direction) is not particularly problematic. Your bodily response to the same meal can vary more than 25%. One test I did do is consume "glucose tablets" (which are basically candy), they're products meant to spike your blood sugar (for diabetics who get "low"), and all you can predict is that my levels go "up fast". Again, all that matters is the trend, especially if your body can self-regulate.
The FDA does study how accurate these products are, and they conform to certain expectations. That said, you're not wrong that it can be 25% off sometimes. According to a study (linked below), ~99% of the time, the device is within 30% of clinical test levels, and about 90% of the time, the device is within 15%.
A huge portion of glucose management (even for diabetics) is the actual trend of the data vs the actual blood glucose levels. These CGMs capture trends correctly, even if the absolute value is off. A prescription variant would have periodic calibrations as an option. If you eat a meal that spikes blood glucose level, the CGM probably won't smooth out a peak, so you'd still see the magnitude of the effect. The other thing to consider is that when read from the arm or a "fleshy" bit, a change in glucose level is about 20min delayed from blood levels.
The healthy range for a non-diabetic is ~70-140 mg/dL, and your body regularly can swing up more than +50mg/dL after a meal. This is a pretty big range and swing, and being 15% off (in either direction) is not particularly problematic. Your bodily response to the same meal can vary more than 25%. One test I did do is consume "glucose tablets" (which are basically candy), they're products meant to spike your blood sugar (for diabetics who get "low"), and all you can predict is that my levels go "up fast". Again, all that matters is the trend, especially if your body can self-regulate.
The FDA does study how accurate these products are, and they conform to certain expectations. That said, you're not wrong that it can be 25% off sometimes. According to a study (linked below), ~99% of the time, the device is within 30% of clinical test levels, and about 90% of the time, the device is within 15%.
https://pubmed.ncbi.nlm.nih.gov/35157505/