I once ran a trial comparing two FitBit models (one quite recent at the time) to a medical heart measurement device and manual counting. My conclusion was that

I once ran a trial comparing two FitBit models (one quite recent at the time) to a medical heart measurement device and manual counting. My conclusion was that FitBit's numbers were pretty much garbage: the manual and medical measurements generally agreed up to 3-4 beats, while the FitBit readings were off by 20 beats and more and fluctuating wildly in a short span of time. Their readings for step numbers might be more accurate, I don't know, but I concluded that the heart rate measurements probably contained more noise than signal.