Innovating in a Sensitive Market

The U.S. biometrics technology market is growing rapidly, valued at $7.55 billion in 2023 with an 18.2% CAGR expected through 2030.

At MDSV, we wanted to catch this wave and serve a niche that has yet to be served.

To introduce a new product into the biometrics market, we faced several key challenges:

1. User Trust: Convincing users to share sensitive biometric data.
2. Regulation: Persevering in a highly regulated space.

Starting with just three team members—a designer (myself), an engineer, and a CEO—we set out to explore a seamless, privacy-first biometric system that delivered real-world value.

Research and Early Analysis

After doing a round of preliminary research, we identified a huge opportunity in the age-restricted market. This includes products like alcohol, tobacco, cannabis, and e-cigarettes.

These four markets alone accounted for $2.6 trillion dollars in annual sales.

More specifically, we noticed that the retail pain point was even bigger.

• 80% of the time, stores failed to ask minors for ID
• 20% of the time, IDs ended up being fake
• Each violation cost the retailer $19,192

Existing products like X, Y, and Z trying serving these markets but had no success.

Designing the MVP

Our market research led us to our first product idea: an SDK for retailers to integrate into their user flow, simplifying the age verification process.

When designing our MVP, I focused on:

• Simplicity: A non-intrusive, seamless biometric scanning experience.
• Transparency: Clearly communicating privacy measures to build trust.

The collaboration between the team was led by our engineer who understood the technical constraints and limitations of a face capture feed.

Working closely with them, I sketched several simple interfaces that would serve as our MVP.

The anatomy of our face capture UX was fairly straightforward, only showing the video feed of the camera, the face capture window, and helper text to guide the user through the process.

Identifying the Error States

As simple as this UX was, we knew there were going to be lots of error states we needed to account for, for example:

• Face is too far
• Face is too close
• No face detected
• Face is not centered
• Not enough lighting
• Too blurry

Each error state was translated into an instruction in the UI, and it needed to be displayed long enough for the user to correct their mistakes, but without interfering with the smoothness of the existing UX.

Piloting the SDK With JUUL

After achieving a 95% confidence score and a +/- 5 Mean Average Error, we decided to begin looking for strategic partnerships that could take advantage of Arcoscan's technology.

We landed an agreement with JUUL to test our technology on their signup experience. JUUL's current signup process took users anywhere from 2-4 minutes, and they wanted to significantly reduce that process to less than a minute.

Over the course of three months, JUUL tested our SDK across 500 users. Initially, they saw a spike in abandonment rates, specifically around the "Liveness Check" where users have to make a gesture to prove they are human.

We managed to land a pilot program with JUUL to test our product, integrating the SDK into their onboarding process to streamline age verification.

Over three months, we tested on 500 new users and found that our the abandonment rate was still hovering at 50%, even with the Arcoscan technology embedded in onboarding.

Interviewing JUUL new users revealed issues with our face capture step. (e.g., centering and gestures)

Improving the Face Capture Experience

With face capture being the largest contributor in abandonment rate for JUUL, we decided to go back to the drawing board to provide a more seamless experience that required minimal effort from the user, and keeping the technical flow strictly in the backend.

Face capture served as our "liveness check" to ensure the user was human.

As a team, we had to decide how we were going to solve for a more seamless user experience, while keeping a version of liveness check live on the platform.

As a quick solution to the problem, we decided to extend the time-out period to 2 minutes and shortened the delay between instructions from 1600ms to 800ms.

After extending the time-out period and shortened the delay between instructions. (1600ms --> 800ms)

Results and Adoption

Incremental updates, including improved instructions, timeout adjustments, and backend fixes, reduced abandonment rates to 40%, with mobile app users showing better performance (21%) than web users (50%).

Unfortunately, the partnership faced adoption challenges, prompting us to reevaluate our approach.

We were able to reduce the number of steps, but trust in this new face capture technology brought the abandonment rate back up to parity.

After our partnership with JUUL, we began experimenting with interactive instructions that visually guided the user through Liveness Checks during face capture.

Finding New Markets

This project taught me critical lessons:

• User Trust: Transparency about data use is essential in sensitive markets.
• Iterative Development: Pivoting based on user feedback unlocked new opportunities.
• Collaboration: Close teamwork between design, engineering, and leadership drove our success.
• Flexibility in Vision: Sometimes, the best application of technology isn’t the one you initially imagine.

Our pilot program with JUUL showed that trust is still the number one barrier for new products entering the biometric space.

As a entry point into a new market, we are currently venturing into the point-of-sale ecosystem to integrate with retailers in-store.

Integrating into POS systems would allow for massive distribution, effectively putting Arcoscan in front of everyday shoppers.