TWO SENTENCE SUMMARY: Degrading slowly over the last 9 months, during workouts, my FitBit shows my Heart Rate (HR) to be fifteen to thirty percent lower than actual, and it lags that way for up to 20+ minutes during the workout before it starts to catch-up. Not a quibble about a couple BPM delta compared to a medical gold-standard device, that’s a big enough error as to render it useless for tracking how much cardio exercise I’m getting, week to week.
BACKGROUND: I've had a FitBit Sense 2 (my first ever wearable) since Christmas, 2022. I will be 70 in a couple of months. Although I’ve had no major health events of any kind—no fainting, no angina, no coronaries, no strokes, etc.—my doctor said my sedentary lifestyle was starting to show up in some of my numbers (I work remote full time as a healthcare data scientist). She said to get back to the gym. Since Christmas, when I got my FitBit, I have worked my way progressively to where I can sustain 25 minutes on the elliptical at 8+ METs of exertion. According to the FitBit app, my resting HR wobbles around 64 BPM plus or minus 2, and my overall cardio fitness is excellent (forty-three to seven) for my age.
Even so, if you think this reads like a FitBit success story, that would be wrong. While my Sense 2 performed reasonably well for the first couple of months, its HR performance has been degrading steadily to the point that its HR readings are useless to track the amount of cardio exercise I’m getting.
FITBIT BEHAVIOR 9+ MONTHS AGO, WHEN I FIRST GOT IT: While the HR BPM readings never exactly matched the numbers on the exercise machines at the gym, early on the differences were small (a few BPM) and were easy to explain in terms of different approaches to filtering intrinsically noisy heart rate signals. The machines’ (treadmill, elliptical, etc.) handgrip HR monitors seem to use a simple moving average (a finite impulse response filter, or FIR) that lags a little and shows what my HR was on average a few seconds ago. The FitBit seems to use a moving average that assures that it won’t jump around (an infinite impulse response filter, or IIR, where the input is not just the noisy raw HR signal but also the filter’s current output) … an IIR designed to add stability and hysteresis. I’m guessing that FitBit chose this approach to improve its AFib detection. Still, even though the FitBit changed HR more slowly / less frequently, being only a few beats per second at variance from the exercise machines’ handgrips, neither the BPM deltas nor the seconds of FitBit lag were meaningful, neither for my fitness nor for my cardiological safety. I was generally satisfied with my Sense 2.
CURRENT FITBIT MISBEHAVIOR MAKES IT USELESS FOR TRACKING TIME SPENT DOING CARDIO: However, over the 9+ months, the real-time performance of Sense 2 HR monitoring has degraded to the point that, when I start working on the elliptical (a True brand machine with resistance level set to twenty of thirty possible), the FitBit shows my HR to be twenty to thirty BPM lower than the machine’s handgrip HR monitor. Again, I don’t think the handgrips are exactly accurate; but several machines agree with each other within a few BPM, which also agrees with my manual check on my carotid artery. Over the past several months, the FitBit has slowly drifted to the point that THE FITBIT IS FIFTEEN TO THIRTY PERCENT LOWER THAN ACTUAL, AND IT LAGS THAT WAY FOR UP TO 20+ MINUTES BEFORE IT STARTS TO CATCH-UP.
WHAT I’VE TRIED: To remedy this, I’ve tried everything I’ve seen mentioned in the Community here, including (a) every feasible combination of re-pairing the Bluetooth, restarting the Sense 2 device, stopping and restarting the FitBit app and rebooting my Android phone, (b) cleaning the Sense 2, (c) moving it from non-dominant to dominant hand, (d) moving it from top to bottom of my wrist, (e) telling the FitBit I’m on the elliptical at the beginning of my workout, and (f) turning off AFib detection. None of these resolve the problem. Given the slow drift over months and the FitBit being all electronics, I doubt the issue is mechanical. My joints won’t tolerate any kind of pounding so my FitBit never gets one. I don’t wear it in the water nor in the shower. I take it off before going into the sauna.
WHERE I’D LOOK FIRST IF I WERE THE DEBUGGER: It's as though (a) the FitBit has a broken AI algorithm that is trying to learn iteratively how best to set the coefficients on the IIR that is instead slowly increasing HR change detection hysteresis, thus degrading the HR monitor’s performance, and/or (b) FitBit is doing some sort of A/B testing by downloading experimental firmware versions to different users devices, as part of improving product performance, which would be fine with users’ permission but not otherwise (did we all agree to that in some fine print somewhere?)
Fitbit Product Experts Alumni are retired members of the Fitbit Product Expert Program. Learn more
Best Answer