The Facematch Verification module determines whether two face images belong to the same person. It compares a selfie photograph captured during the onboarding journey against a reference face image, typically the portrait photograph extracted from the subject’s identity document or read from the document’s NFC chip. While liveness verification confirms that a real person is present, facematch verification confirms that the person present is the same person shown on the identity document.
This page contains documentation for the Facematch Verification module, including its variants, capabilities, and the specific result values or scoring thresholds each variant uses.
The module has two variants. Both compare a selfie against a document portrait, but they differ in result type and the level of control over matching thresholds.
| Variant | Result type | Threshold control | Review tier |
|---|
| NIST | Binary (Match / No Match) | None | Not available |
| V1 | Numeric score (0–100) | Configurable | Available |
Facematch Verification - NIST
The NIST variant uses a face recognition algorithm that has been independently evaluated under the NIST Face Recognition Vendor Test (FRVT) programme.
FRVT is a series of large-scale evaluations conducted by the US National Institute of Standards and Technology that measures the accuracy and performance of face recognition algorithms across standardised datasets.
An algorithm evaluated under FRVT has been tested against millions of facial images under controlled conditions, providing independent assurance of its accuracy and demographic fairness.
This variant compares the selfie image against the anchor image (the portrait extracted from the identity document) and returns a binary match or no-match result.
Capabilities
The module returns a single capability: the facematch result.
Facematch result
This capability indicates whether the face in the selfie matches the face in the reference image.
| Code | Label | Description |
|---|
Match | Match | The two face images are determined to belong to the same person. The selfie captured during the journey matches the portrait photograph on the subject’s identity document. |
No Match | No Match | The two face images are determined to belong to different people. The selfie doesn’t match the document portrait. This may indicate that the person presenting the document isn’t the document holder or a low-quality document photograph is captured. This is the default value. |
Default outcomes
| Outcome | Condition | Description |
|---|
Success | Facematch result is Match | The selfie matches the document portrait. The subject is confirmed as the document holder. |
Fail | Facematch result is No Match | The selfie doesn’t match the document portrait. Depending on your use case, this may route to manual review, a request to recapture the selfie, or rejection. |
ERROR | Default (no conditions matched) | The facematch check couldn’t be completed due to a system-level issue. This may occur if the selfie or anchor image couldn’t be processed, or if no face was detected in either image. |
Facematch Verification V1
The V1 variant uses a newer comparison engine that returns a granular similarity score rather than a binary match result. This gives you more control over how strictly you enforce the match, since you can configure the acceptance threshold to suit their risk tolerance.
The variant compares the selfie image against the anchor image and produces a score from 0 to 100, where a higher score indicates greater similarity between the two faces. Default outcomes are configured around two thresholds (80 for success, 70 for review), but these can be adjusted in the Journey builder.
Capabilities
The module returns a single capability: a facematch score.
Facematch score
This capability returns an integer similarity score indicating how closely the selfie matches the reference face image.
| Detail | Description |
|---|
| Type | Integer |
| Range | 0–100 |
| Default | 0 |
- Significant time elapsed between the document photo and the selfie (ageing).
- Changes in appearance such as facial hair, glasses, hairstyle, poor lighting.
- Camera quality during selfie capture, low resolution or damage to the document photograph.
- Differences in head angle or facial expression between the two images.
A score of 0 indicates that the comparison couldn’t produce a result. This occurs if no face was detected in the selfie, if the anchor image was missing or unreadable, or if an error prevented the comparison. Evaluation rules should treat a score of 0 as an inconclusive result rather than a confirmed mismatch.
Default outcomes
The module uses score thresholds to classify the facematch result into one of three outcomes.
| Outcome | Condition | Description |
|---|
Success | Score ≥ 80 | The system has high confidence that the selfie and the document portrait belong to the same person. The subject is confirmed as the document holder. |
Review | Score ≥ 70 and < 80 | The system’s confidence is moderate. The faces may belong to the same person, but the similarity score is not high enough for automatic acceptance. Common causes include poor lighting, a significant age gap between the document photo and the selfie, or partial face occlusion. This outcome typically routes to manual review, where a human operator visually compares the two images. |
Fail | Score < 70 | The system has low confidence that the faces belong to the same person. The selfie and the document portrait are likely of different individuals. This outcome typically routes to rejection or escalation. |
ERROR | Default (no conditions matched) | The facematch check could not be completed due to a system-level issue. |
The default score thresholds (80 for success, 70 for review) are starting points. You can adjust these thresholds in the evaluation configuration to match their risk tolerance. Lowering the success threshold increases acceptance rates but allows weaker matches through. Raising it improves security but may increase the number of genuine users routed to manual review, particularly for documents with older photographs.
Choosing between variants
The two variants serve the same purpose but differ in how they report results and how much control they give the you over the matching decision.
| NIST | V1 |
|---|
| Result type | Binary (Match / No Match) | Numeric score (0–100) |
| Threshold control | None: the algorithm applies its own internal threshold | Configurable: You set the success and review thresholds |
| Review tier | Not available: result is either match or no match | Available: scores between the review and success thresholds route to manual review |
| Algorithm assurance | Independently evaluated under NIST FRVT | Not specified |
| Best suited for | If you want a definitive yes/no determination with independently validated accuracy | If you want granular control over acceptance thresholds and a manual review tier for borderline cases |
{
"resourceId": "<resourceId>",
"context": {
"subject": {
"biometrics": [
{
"selfieImage": "<base64-encoded string>",
"anchorImage": "<base64-encoded string>"
}
]
}
}
}
| Field | Required | Description |
|---|
selfieImage | Yes | Base64-encoded selfie image captured from the end user. |
anchorImage | No | Base64-encoded reference image used for comparison. This is always sourced from a document processed earlier in the journey. |
biometrics | Yes | Accepts at least one item. |
