What is face match?
Face match compares two face images to judge whether they show the same person. One is captured at the moment of the check, such as a selfie during onboarding or a frame from a video session. The other is a reference the business already trusts: a document photograph, an enrollment photo, or an image in the customer record.
The result is a similarity signal about two images. It settles nothing about who someone is. You read it next to the document and record checks your journey already runs.
How a face match API works
An API turns that comparison into a step inside software you already run. Instead of a reviewer holding a document photo beside a selfie, your application sends both images and gets a structured result it can route on.
That matters where the same decision repeats: a lender opening applications daily, a delivery platform activating riders weekly, a marketplace re-checking sellers at a payout change. The outcome no longer turns on who picked up the file.
The two images a face comparison needs
Every request needs a captured image and a reference image, and the condition of both decides how useful the result is. Reference photos lifted from scanned documents are often small, textured, or years older than the person at the camera. Phone captures can carry glare, motion blur or heavy compression.
A response can carry a similarity indication and image-quality signals that explain why a comparison was weak. The exact fields depend on the data source, the record and your service contract. A pair that cannot be compared responsibly may be declined.
What a face comparison cannot tell you
A strong similarity result says the two images look like the same person. It says nothing about how the captured image reached you, because the comparison never sees the camera. A printed photograph, a screen held to the lens, or a stored file sent in place of a fresh capture can all compare well against a document photo.
The separate Face Liveness service is there to narrow that gap, and the order matters: check the capture first, compare it second. Where liveness is not part of a journey, read the result as weaker evidence and lean harder on the checks around it.
Decisions face match supports, by industry
- Banks, lending and NBFCs: tie the application to the person on the submitted PAN card or driving license, then look again at disbursal.
- Insurance: put a check between a claim and its payout, and support proof-of-life reviews on long-running pensions and annuities.
- Fintech and payment service providers: verify the signatory during merchant onboarding, and step up when an account is reached from an unfamiliar device.
- Marketplaces: hold a seller account to the identity behind it, including at a payout change.
- Logistics and fleet: establish at shift start that the person about to drive is the one enrolled against the license on file, which is where substitution tends to happen.
- Staffing and gig platforms: check the worker when a shift opens, to make a rented or shared account harder to keep hidden.
- Telecom and retail networks: check whoever activates a connection at the counter, which makes it harder to raise a subscription on someone else's documents.
- Travel, hospitality and real estate: match a guest at check-in to the document given at booking.
Designing a face check into your journey
Four decisions shape the outcome more than any later tuning:
- Choose which reference image is authoritative for each journey: the document photo, the enrollment photo or the customer record.
- Put the capture behind a guided step with framing and lighting prompts.
- Fix your thresholds before launch and record why each one sits where it does, to keep a later move deliberate.
- Define the route for people a photo cannot resolve.
Running face match next to your other checks
Face match works best as one gate in a sequence, before anything irreversible. Send weak comparisons to a review queue with both images and the reason attached.
Sibling resemblance, aged photos, spectacles, face coverings and poor scans all produce genuine borderline cases. A fresh capture resolves more of them than a retry.
Consent and data protection for face images
A face image is personal data about an identifiable person. Collect it only for a purpose the person has been told about and has agreed to. Record consent before the capture, keep it with the case, and honor a withdrawal.
Use the comparison only for the authorized purpose, and keep images no longer than that purpose needs.
A similarity result should not be the sole basis for refusing someone a service. Provide a human review path and a way to contest an outcome, and do not reuse onboarding images for any secondary purpose the person never agreed to.