Tech · Uni-Ubi · 2016–2018

Uface Smart Reception

A facial-recognition reception system across device, mobile app, and back-office, replacing manual front-desk check-in with fast, contactless recognition for offices, communities, and more.

Role

UX / UI · IA · Device flow

Platform

Device · App · Admin

Client

Uni-Ubi

Background

Contactless reception for the full physical-access scenario.

Uni-Ubi is an AIoT-based smart-city technology provider offering full-scenario IoT solutions. The Uface Smart Reception system serves offices, communities, hotels, and similar facilities, enabling ultra-fast facial-recognition sign-in, visitor management, and integrated attendance tracking. The goal was to streamline the front-desk experience for both employees and visitors, replacing manual check-in with automated, contactless recognition.

Visit website →

The Challenge

The design needed to unify scattered workflows, ensure smooth hardware–software integration, and strike the right balance between security and convenience, all while serving managers, employees, and frequent visitors in a single, intuitive system.

Uface Gen 1 — device, attendance app, and back-office admin system

Fragmented user flows

Attendance, visitor check-in, and approval processes were handled through separate, inconsistent systems.

Hardware–software complexity

Facial-recognition devices needed to communicate seamlessly with the mobile and backend systems.

Security vs. convenience

Prevent unauthorized entry while keeping check-in fast and seamless, without forcing a trade-off.

Diverse user needs

One solution had to serve managers, employees, and frequent visitors without adding complexity.

My Role & Responsibilities

Worked with the PM to create the information architecture unifying attendance, approvals, visitor management, and personal center.

Designed UI/UX for the mobile and backend admin system, ensuring consistency across devices.

Collaborated with hardware engineers to refine device–UI interaction flows for instant recognition feedback.

Design & Architecture

Based on preliminary investigation and research, the app is structured into four main modules: Attendance, Process Management (leave requests and overtime applications), Find (attendance rankings and visitor management), and Personal Center.

Information architecture / module map

App home / module overview

Key Scenarios

Attendance

The attendance module gives employees and managers a clear, interactive way to track work hours. Inspired by a traditional clock face, it uses color-coded indicators for check-in and check-out, making history instantly visible. Users can view personal summaries, see statistics as a histogram, and track up to three work periods per day, ideal for flexible schedules. The design balances function and clarity so data reads at a glance.

Attendance — personal data, three-period & blank status

Visitor Management

In many workplaces, visitor management is either too strict (causing delays) or too loose (compromising security). Uface needed a flow that verifies visitors quickly while preventing unauthorized access, handled in four steps:

Pre-registration

Employees register guests in advance with name, company, date/time, and reason.

Photo verification

Guest photos are uploaded ahead of time for facial recognition on arrival.

Instant host alert

The host is notified the moment their visitor checks in at reception.

Time-limited access

One-time permissions valid only during the scheduled meeting window.

Visitor registration, approval flow & photo verification

A/B Testing — Approval flow

Choosing the faster approval pattern with data, not opinion.

Variant A

Open-then-approve

Users open each request to a detail page, then approve. Steadier and lower-error, but slower for high volume.

Variant B

Inline actions

Approve / reject directly from the list. Much faster, with a small risk of mis-taps that guardrails can absorb.

Variant B — inline approve / reject flow

12.5s → 4.5s

Median time per approval

64% faster

Variant B vs. Variant A

0.2% / 0.6%

Mis-tap rate (A / B) — not significant after guardrails

Decision: Ship Variant B with guardrails, a confirmation modal, a required reason on reject, and key fields shown inline. Experiment used 50/50 user-level randomization (sticky buckets), same approval type first, run over ≥2 work weeks for sufficient samples.

Testing Findings & Iteration

Finding 01

Recognition feedback wasn't salient outdoors

Problem

In field tests at construction sites and school gates, bright sunlight and ambient noise made it hard to tell whether recognition succeeded.

Gen 1 — indoor

Subtle success/failure cues were fine for indoor offices, but couldn't adapt to the new outdoor environment.

Gen 2 — outdoor

Enlarged, high-contrast success/failure states plus audible cues (beeps + voice prompts) for unmistakable confirmation.

Gen 1 vs Gen 2 feedback states

Finding 02

Manual entry for visitor info was cumbersome

Problem

Adding a visitor meant filling six fields one by one. When details came from chat or email, users had to copy–switch–paste repeatedly, causing delays and errors.

Original

A blank form (Name, Gender, Company, Arrival, Departure, Reason) with no auto-fill and no clipboard detection.

Improved

Detect structured text on the clipboard, prompt "Paste the info just copied?"
Auto-parse key fields, with per-field check / uncheck.
One-tap confirm to populate; all fields stay editable.

Visitor flow

From one by one input to one tap copy

UI Works