What it measures
False Rejection Rate (FRR) measures how often a biometric system incorrectly rejects a legitimate match — a real, matching face that the system fails to recognize as a match.
Why it's a trade-off, not a target to zero out
Tuning a system to never reject a real match tends to also make it more willing to accept near-matches that aren't real — pushing FRR down usually pushes the false acceptance rate up. The two have to be balanced, not optimized independently.
What drives it up in practice
Poor lighting, low-quality captures, and outdated reference photos all increase false rejections in real-world use — which is part of why capture quality and liveness checks matter as much as the matching algorithm itself.