Technology
Industries
Use Cases
3DiVi Inc., founded in 2011, is one of the leading developers of AI and machine learning (ML) technologies for computer vision.
Let's talk business
Just fill in the form and we will contact you shortly
On-premise license request
Just fill in the form and we will contact you shortly
Your request has been successfully sent. We'll get in touch shortly.
THANK YOU!
Let's talk business
Just fill in the form and we will contact you shortly
Don't Add Face Recognition to 2FA / MFA Until You've Done These 3 Things
Rushing into face recognition integrations without a clear strategy is recipe for high failure rates, poor UX, and user frustration.
With 14 years of experience in computer vision, we’ve learned that before any implementation, teams need to focus carefully on these three critical factors.
With 14 years of experience in computer vision, we’ve learned that before any implementation, teams need to focus carefully on these three critical factors.
1. Don’t Just Set Goals—Make Them SMART
“A well-defined goal is half the battle.” – Abraham Lincoln
Biometric projects often start with ambitious but vague objectives: “Improve security,” “Speed up onboarding,” or “Reduce fraud”.
The truth is if success isn’t clearly defined, you can’t optimize for it. SMART goals—Specific, Measurable, Achievable, Relevant, and Time-bound—turn ambition into actionable targets.
Biometric projects often start with ambitious but vague objectives: “Improve security,” “Speed up onboarding,” or “Reduce fraud”.
The truth is if success isn’t clearly defined, you can’t optimize for it. SMART goals—Specific, Measurable, Achievable, Relevant, and Time-bound—turn ambition into actionable targets.
Case in point: In 2023, New Zealand launched a national face authentication system as part of its digital identity program. The initiative aimed to allow citizens to verify their identity online—unlocking access to banking, government services, and more.
After early testing revealed face recognition failures in 45% of initial attempts, the project team set a new goal: Achieve a 90% verification success rate within 3 attempts. After adjustments, they hit 89%—a major leap forward.
Here’s how that aligns with SMART:
- Specific – The goal was clearly defined: 90% success within three attempts.
- Measurable – The success rate was measured at 89%, close to the target.
- Achievable – The pilot results (89%) suggest that 90% is realistic.
- Relevant – While exact comparisons are unavailable, a 90% success rate is likely higher than traditional offline identity verification methods.
- Time-bound – The project operates on a defined timeline, with phased testing and rollout.
Clear targets turn vague ambition into actionable strategy. Without them, technical teams risk building blind,—unable to measure progress or tie tech to business value.
After early testing revealed face recognition failures in 45% of initial attempts, the project team set a new goal: Achieve a 90% verification success rate within 3 attempts. After adjustments, they hit 89%—a major leap forward.
Here’s how that aligns with SMART:
- Specific – The goal was clearly defined: 90% success within three attempts.
- Measurable – The success rate was measured at 89%, close to the target.
- Achievable – The pilot results (89%) suggest that 90% is realistic.
- Relevant – While exact comparisons are unavailable, a 90% success rate is likely higher than traditional offline identity verification methods.
- Time-bound – The project operates on a defined timeline, with phased testing and rollout.
Clear targets turn vague ambition into actionable strategy. Without them, technical teams risk building blind,—unable to measure progress or tie tech to business value.
2. One Success Metric is not Enough
A 95%+ face match rate is the industry standard, but that number alone won’t tell you how well your system actually works.
Let’s take a look at Kazakhstan’s digital ID verification process. Now it requires users to perform multiple head movements as an extra layer of security. Sounds smart, right? But for elderly users or those with motor limitations, it’s a serious usability challenge.
So while your model might hit 95% success in lab conditions, real-world performance varies drastically depending on UI, user demographics, and device capabilities.
What to do instead:
Define multiple KPIs that reflect both system performance and UX, aligned with industry standards such as performance metrics set by NIST (National Institute of Standards and Technology).
Let’s take a look at Kazakhstan’s digital ID verification process. Now it requires users to perform multiple head movements as an extra layer of security. Sounds smart, right? But for elderly users or those with motor limitations, it’s a serious usability challenge.
So while your model might hit 95% success in lab conditions, real-world performance varies drastically depending on UI, user demographics, and device capabilities.
What to do instead:
Define multiple KPIs that reflect both system performance and UX, aligned with industry standards such as performance metrics set by NIST (National Institute of Standards and Technology).
3. A Two-Stage Approach to Better Results
Face authentication isn’t just about the algorithm. The process itself can be split into two distinct phases—and doing so will dramatically improve its accuracy and UX.
Phase 1: External Constraints
These are pre-authentication steps imposed by regulations or design choices—things like performing head turns, blinking, or entering a unique identifier.
These steps are critical, but they’re not part of the actual face match process. Instead, they should be optimized through UI/UX testing, not treated as part of the success rate metric.
Phase 2: Face Authentication Itself
This is where you measure biometric performance. A 95%+ success rate here is excellent in production environments.
By separating these two phases you avoid skewed performance data, get clearer insight into where failures occur and create better experiences, especially for first-time users.
Phase 1: External Constraints
These are pre-authentication steps imposed by regulations or design choices—things like performing head turns, blinking, or entering a unique identifier.
These steps are critical, but they’re not part of the actual face match process. Instead, they should be optimized through UI/UX testing, not treated as part of the success rate metric.
Phase 2: Face Authentication Itself
This is where you measure biometric performance. A 95%+ success rate here is excellent in production environments.
By separating these two phases you avoid skewed performance data, get clearer insight into where failures occur and create better experiences, especially for first-time users.
Face authentication can drastically improve security and UX—but only when it’s planned with precision.
Set SMART goals. Go beyond one-size-fits-all metrics. And split your process into phases that can be tested and improved separately.
Ready to upgrade your 2FA / MFA with face recognition? Let's talk.
Set SMART goals. Go beyond one-size-fits-all metrics. And split your process into phases that can be tested and improved separately.
Ready to upgrade your 2FA / MFA with face recognition? Let's talk.
