Yan Luo

10 years of experience UX researcher. I turn messy, high-stakes ambiguity into product decisions teams can act on.

From Concept to In-Market Success: 5 Studies on Live and Emerging Products

Case files

Each one follows the same arc: the challenge, how the study was run, what it found, and what changed because of it.

More work

Google Translate · Usability test · N=8 + 1 pilot · 2021

Slide from the original deck titled 'Impact – Strategic – Side by side view as key opportunity for web,' showing Google Translate's image viewer with the Spanish original and English translation side by side on desktop, and the full translated view on mobile.

Made the default view match what users already expected

A pre-launch usability test on Google Translate's new image-upload flow. Every participant expected translations to open side-by-side by default — that became the shipped design. Findings also went into Google's onboarding materials for new researchers.

Read the report (PDF) ↗
About

Notes on approach

These case studies span a decade—from evaluating pre-production VR hardware to observing AI shopping agents in the wild. The methodology shifts with the question: a Wizard-of-Oz rig when tech isn't real yet, a MaxDiff survey when user needs require strict trade-offs, or a diary study when trust and habits need to be observed rather than asked about.

What never changes is the definition of done: a study isn't complete when the deck is shared, but when it steers a decision—whether triggering a contextual feature prompt, shifting cross-functional roadmaps, or defining a controller that lands on store shelves, the goal is always tangible, measurable impact.

All case files

Case file 01

Established a Secure Payment Experience for Agentic Commerce

Client
Visa
Method
Diary study
Participants
25
Date
April 2026
Read the full report ↗

Contribution and impact

Built a multidimensional agentic commerce evaluation rubric, mapped critical Human-in-the-loop safeguards, translated findings to key strategic priorities for Visa — driving adoption by deploying tokenized payments and establishing default user verification.

Slide from the original Visa deck titled 'The idealized flow is a sequence of delegation followed by user authorization,' showing a 6-step flow split into the 95% shopping-consultant phase and the 5% deciding-and-purchasing phase.
From the original deck — the idealized agentic purchase flow participants converged on: the AI handles discovery and comparison, the person keeps final sign-off.

Challenge

As "agentic commerce"—AI agents discovering and purchasing on a user's behalf—moved from concept to reality, Visa needed to define its role in making autonomous transactions trustworthy enough for adoption.

Approach

A 5-day diary study tracking 25 participants across four AI shopping tools: Amazon, Microsoft Copilot, ChatGPT, and Perplexity. Each shopper completed a live transaction using a primary tool and evaluated a secondary tool, capturing discovery, evaluation, and checkout experience.

Key findings

  • Solution. AI agents are helpful research partners, but inability to complete transaction within the AI tool frustrates users.
  • Trustworthiness. Concerns about over-automation making some hesitant to share their payment information.
  • Accuracy. Displaying accurate product prices upfront was expected, but AI failed to do so.
  • Ease. Users want the AI platform to save personal information for easier shopping.
  • Perceived security — Payment. Tokenized payments and credit cards are the ideal payment methods for AI platforms.
  • Perceived security — Verification. Verification of purchases is required, with biometrics considered the most secure method.

Implementation and influence

  • Product Execution. Initiated strategic digital wallet partnerships to deploy tokenized payment within agentic workflows, separating sensitive financial data from AI platforms.
  • UX/Design Strategy. Provided Visa brand reassurance during payment entry to reinforce consumer protections and alleviate data-sharing concerns.
  • Authentication Standard. Adopted Visa Payment Passkeys (utilizing biometrics) as the default verification, ensures users retain the control.
All case files

Case file 02

Defined Google TV's India localization strategy

Client
Google
Method
Diary study, interviews, MaxDiff survey
Participants
24 diary · 12 interview · 1,000 survey
Date
March 2022
Read the full case study →

Contribution and impact

Uncovered Indian TV viewers' core viewing behaviors and built a MaxDiff vs. "Well-Met" framework to prioritize critical unmet needs. By establishing content aggregation and subscription bundling as primary market differentiators, this work guided the launch and localization strategy for Google TV's smart UI (Amati) in India and introduced contextual Google Assistant prompts during personal profile use, driving a 23% increase in Google Assistant engagement.

Quant key findings: needs prioritization 2x2 — MaxDiff importance vs. how well-met each need is, sorted into Top Differentiators, Table Stakes, Value Adds, and Innovate quadrants.
From the original deck — the MaxDiff-vs-"Well-Met" 2x2 used to sort Indian TV viewers' needs into top differentiators, table stakes, value adds, and innovate bets.

Challenge

Amati, the next generation smart TV UI, was heading into one of Google TV's most important markets with no clear read on which localized needs would actually move the needle for Indian household.

Approach

  • Phase 1 (Qualitative). A 5-day diary study (N=24) across four geographic regions with 12 follow-up interviews to uncover TV viewing habits, needs, and first impressions of the Amati concept.
  • Phase 2 (Quantitative). A MaxDiff survey (N=1,000) mapping need importance against current market satisfaction ("Well-Met" analysis) to define critical product differentiators.

Key findings

  • Family-friendly discovery. Shared, multi-generational households turn family-friendly discovery into a requirement, creating hesitation around voice assistants misinterpreting queries.
  • All-in-one is helpful. Almost everyone already juggles 5+ streaming apps — "everything in one place" was the single strongest hook.
  • Discount + convenience = delight. Together they drove real delight — neither alone was enough.

Implementation and influence

  • Drove Measurable Feature Adoption. Introduced contextual "Try Google Assistant" prompts within personal profiles, lifting Google Assistant engagement by 23% in the following quarter.
  • Defined Go-to-Market Differentiators. Positioned cross-platform content aggregation and subscription bundling as primary market differentiators to anchor regional product and marketing strategies.
  • Localized Privacy & Family Controls. Directly influenced the Google TV PRD to incorporate profile locks, family-friendly content tagging, and regional language filters tailored to communal Indian households.
  • Cross-Organizational Influence. Scaled research impact across Google, with Indian media consumption findings cited by cross-PA roadmaps including YouTube and Pixel Tablet. The mixed-methods framework was also integrated into the research handbook for new hire onboarding.
All case files

Case file 03

Shaped the design of Sony's Aim Controller

Client
Sony PlayStation
Method
Wizard-of-Oz prototyping, lab experiment
Participants
24 per game
Date
May 2016
Read the full report ↗

Contribution and impact

De-risked complex hardware concepts ahead of physical production, directly steering the hardware roadmap toward developing and launching the dedicated PS VR Aim Controller in 2017. Furthermore, uncovering critical interaction hurdles with sensor gloves—such as bimanual ambiguity—convinced leadership to halt standalone glove development and instead integrate capacitive finger touch detection on existing controllers to simulate natural finger tracking.

Research design slide describing the Wizard-of-Oz VR glove prototype method: participants wore mock VR gloves framed as functional, human operators simulated glove output, with pros and cons of the approach listed.
From the original deck — the Wizard-of-Oz prototype method used to simulate VR glove input before the hardware existed.

Challenge

Prior to launching PSVR, Sony faced fundamental uncertainty around interaction modalities—specifically whether players would favor sensory gloves, traditional controllers, or dedicated VR peripherals, and whether those preferences would diverge across game genres.

Approach

Wizard-of-Oz prototyping simulated VR glove interactions before the hardware was built, paired with a fully counterbalanced within-subjects lab experiment — 24 participants per game — across racing, shooter, and fighting genres. Followed quantitative measurement with interviews to probe mental transportation, controller preferences, and overall enjoyment.

Key findings

  • Participants strongly favored physical controllers that mirrored natural, real-world interactions.
  • Dedicated tactile controllers (like the Aim Controller) significantly elevated mental transportation, which led to higher overall gameplay enjoyment.

Implementation and influence

  • Steered Shooter Hardware Strategy. Guided the hardware roadmap by establishing the core case for dedicated, tactile shooter controllers, directly shaping the direction that led to the PS VR Aim Controller launch.
  • Resource Optimization. Redirected product development away from low-desirability standalone gloves, translating gesture insights into practical enhancements like capacitive finger touch detection on current hardware.
  • Organizational Mindset Shift. Aligned engineering, product, and design around user desirability thresholds, establishing a team-wide framework for balancing novel technology with intuitive tactile interaction.
All case files

Case file 04

Segmented health consumers to shape wearable and health app roadmaps

Client
Google Health
Method
Survey, video diaries, interviews
Participants
2,000 survey · 60 diary · 6 interview
Date
2020
Read the full report ↗

Contribution and impact

Built a six-segment model of US health consumers from a 2,000-person survey, video diaries and deep-dive interviews. The segments were mapped directly to wearable and health app roadmaps — validating proactive tracking and fitness gamification on Fitbit — and became a standardized screener that cross-functional teams use to recruit and segment participants.

Quadrant map of the six consumer health segments, plotted by health status from poor to excellent and by approach to care from reactive to proactive.
From the original deck — the six segments mapped by health status and by approach to care, from reactive to proactive.

Challenge

Early health app development at Google Health relied on a four-profile consumer model that was too shallow to guide product decisions, and industry segmentation frameworks didn't reflect how people actually move through their health — their health status, their needs, and how they use healthcare. The team needed a rigorous, Google-specific segmentation that product teams could also use for recruiting.

Approach

  • Phase 1 (Quantitative). A 15-minute survey of 2,000 US adults on healthcare attitudes and behaviors, product use, and digital health adoption, used to build the segments.
  • Phase 2 (Video diaries). Ten participants from each of the six segments recorded five short day-in-the-life videos.
  • Phase 3 (Deep dives). Six 60-minute one-on-one interviews, one per segment, selected from the diary participants.

Key findings

  • Healthcare avoiders. The Practical Preventers are in average health and skip the doctor because they feel fine; the Anxious Avoiders live with chronic issues and feel overwhelmed by the system.
  • Healthcare receivers. The Deliberate Doers are in good health and proactively track diet and exercise; the Mindful Managers manage their conditions and stay open to care.
  • Caretakers. The Overburdened Attendants care for someone else while their own health slips under financial strain; the Unbothered Allies keep well-maintained conditions and stay realistic.

Implementation and influence

  • Wearable and Health Ecosystem Roadmaps. Mapped segment-specific needs directly to product roadmaps — validating proactive tracking and fitness gamification on Fitbit for the Deliberate Doers, while defining the need for continuous, passive symptom monitoring in the health app ecosystem for the Mindful Managers.
  • Standardized Typing Tool. Built and deployed an operational screener across cross-functional teams, accelerating user recruitment and cohort segmentation for subsequent product testing.
All case files

Case file 05

Designed the safety-first interaction model for in-car commerce

Client
Visa
Method
In-car study, 90-minute sessions
Participants
13 drivers
Date
2026
Read the full report ↗

Contribution and impact

Tested six agentic-commerce concepts — two interaction models (all-at-once vs. step-by-step) crossed with three authentication methods (biometric, verbal passphrase, passive) — through live in-car testing with 13 drivers. Defined the checkpoint-based hybrid flow, hardware and telemetry guardrails, and brand-positioning recommendations Visa needs to scale agentic commerce safely into the vehicle.

Scorecard comparing the all-at-once and step-by-step interaction models on driving focus, error recovery, and purchase certainty — step-by-step scores strong across all three, all-at-once scores moderate to weak.
From the original deck — driving focus, error recovery, and purchase certainty scored across the two interaction models tested in-car.

Challenge

Before agentic commerce could extend into the vehicle, Visa needed to understand whether drivers would trust an AI agent to complete a purchase hands-free — and what interaction and authentication model could keep that trust intact without compromising driving safety.

Approach

90-minute in-person sessions with 13 frequent drivers inside a Genesis GV80, evaluating six voice-based prototype concepts — two interaction models (all-at-once vs. step-by-step) crossed with three authentication methods (biometric, verbal passphrase, passive) — while stationary and driving a controlled loop.

Key findings

  • Pacing. All-at-once felt faster, but step-by-step scored stronger on driving focus, error recovery, and purchase certainty — especially once mistakes or distractions occurred.
  • Authentication tradeoff. Biometrics ranked highest on security and confidence, but users struggled to tap the screen precisely while driving; verbal passphrase felt safer hands-free but less secure with passengers present.
  • Verbal-first, visual-backup. Drivers wanted a hands-free voice flow by default, but instinctively glanced at the screen the moment they hit an error or felt uncertain.
  • Scalability ceiling. Users welcomed routine, low-consideration purchases (food, pharmacy) but rejected high-consideration categories (electronics, apparel) for a voice-only flow.

Implementation and influence

  • Defined the Interaction Model. Recommended a four-checkpoint hybrid flow — combining the speed of all-at-once with the step-by-step model's built-in confirmation pauses — as the default in-car purchase flow.
  • Established Safety Guardrails. Specified hardware placement requirements (biometric sensors within the driver's immediate reach) and contextual telemetry rules that pause payment prompts during active maneuvers like turns or hard braking.
  • Shaped Authentication Strategy. Recommended a dynamic, context-aware authentication model — verbal passphrase when driving alone, biometric or facial recognition when passengers are present — bound by system guardrails rather than user self-assessment.
  • Informed Brand & Scalability Strategy. Defined where Visa branding should anchor trust in the flow, and set the purchase-category and price-cap boundaries for scaling agentic commerce into non-food verticals.