What is Face Comparison?

Face comparison is a similarity score between two facial images—typically one live capture, the other a document photograph—to determine if they are the same person. It is the core of remote onboarding and recovery. A model encodes both faces as embeddings, then a distance metric determines how similar is “close enough”. Set the threshold too low, and fraud comes in the door. Set it too high and legitimate users turn away.

But what makes it reliable? Controlled capture, repeatable framing, and authentic liveness. Passive signals can detect replays, active challenges foil pre-rendered clips. Liveness checks explain more. Couple comparison with document genuineness and metadata sanity checks. If the MRZ, holograms, and barcodes all tell the same story, your decision will be more robust.

Operationally, store the score, the threshold, and the environment metadata (camera class, device ID, glare flags). When risk increases, mismatched address or high-limit product, raise the threshold or request additional evidence in a more comprehensive identity verification workflow. Reserve human review for exceptional cases, but don’t convert every edge case into a cliff.

It’s not mathematics. It’s evidence, it’s context, and it’s restraint—together.