CLIENT
City of Atlanta
City of Atlanta
ROLE
UX Designer
UX Designer
YEAR
2024
2024
LOCATION
Atlanta, US
Atlanta, US
Overview
Overview
Overview
Our UX team came together with one burning question: “How can we make Google Home feel genuinely effortless, personal, and fun?” Spurred on by a Senior UX official at Google and a seasoned Google UX designer, we set out to tackle frustrations that users have voiced for years—like clunky onboarding, shallow personalization, and a voice assistant that frequently misunderstands. Instead of small, incremental tweaks, we took a bold approach to re-think each stage of the user journey, weaving together insights from real people, competitor analysis, and our own imaginative leaps.
Our UX team came together with one burning question: “How can we make Google Home feel genuinely effortless, personal, and fun?” Spurred on by a Senior UX official at Google and a seasoned Google UX designer, we set out to tackle frustrations that users have voiced for years—like clunky onboarding, shallow personalization, and a voice assistant that frequently misunderstands. Instead of small, incremental tweaks, we took a bold approach to re-think each stage of the user journey, weaving together insights from real people, competitor analysis, and our own imaginative leaps.
Defining the Problem
Defining the Problem
Defining the Problem
Google’s marketing promises an ecosystem that seamlessly handles daily chores, from controlling lights to playing music. Yet when we interviewed and observed actual users—particularly older millennials and Gen Xers who aren’t super tech-savvy—we saw three pain points crystallize:
Google’s marketing promises an ecosystem that seamlessly handles daily chores, from controlling lights to playing music. Yet when we interviewed and observed actual users—particularly older millennials and Gen Xers who aren’t super tech-savvy—we saw three pain points crystallize:
01
01
01
Onboarding
Onboarding
People expected a speaker that would talk them through setup, but discovered an app-based labyrinth. “I wanted to just say, ‘Help me set up!’ but it never did,” one interviewee explained, lamenting the 15 to 20 minutes it often took to get fully functional.
People expected a speaker that would talk them through setup, but discovered an app-based labyrinth. “I wanted to just say, ‘Help me set up!’ but it never did,” one interviewee explained, lamenting the 15 to 20 minutes it often took to get fully functional.
The onboarding process took more than 20 screens.
The onboarding process took more than 20 screens.
02
02
02
Personalization
Eager to customize routines or link multiple accounts, users found themselves hunting around the Google Home app. “I had no idea how to do bedtime routines for me and my spouse,” complained another. The device itself gave zero prompts—pushing them back to the phone for complicated tasks.
Personalization
Eager to customize routines or link multiple accounts, users found themselves hunting around the Google Home app. “I had no idea how to do bedtime routines for me and my spouse,” complained another. The device itself gave zero prompts—pushing them back to the phone for complicated tasks.
03
03
03
Voice Assistance
Instead of the advanced “smart” system Google’s brand suggests, many discovered it only recognized specific phrasing. “I have to reword everything to get a response,” one father said. Or, “If I mumble or have a different accent, it’s hopeless.” Multi-step tasks, like “Set 3 alarms for 7, 7:30, and 8,” often failed entirely
Voice Assistance
Instead of the advanced “smart” system Google’s brand suggests, many discovered it only recognized specific phrasing. “I have to reword everything to get a response,” one father said. Or, “If I mumble or have a different accent, it’s hopeless.” Multi-step tasks, like “Set 3 alarms for 7, 7:30, and 8,” often failed entirely
Key Opportunity
Key Opportunity
Key Opportunity
Satisfy a more demanding demographic—those not automatically comfortable with voice technology. If we can make Google Home truly intuitive for them, we’ll almost certainly please younger, more tech-forward watchers too.
Satisfy a more demanding demographic—those not automatically comfortable with voice technology. If we can make Google Home truly intuitive for them, we’ll almost certainly please younger, more tech-forward watchers too.
Our Journey & Mentorship
Our Journey & Mentorship
Our Journey & Mentorship
From the get-go, we had two remarkable mentors from Google: a Senior Staff UX Manager and a Google UX designer, who regularly challenged us to keep user empathy at the forefront. At every step—surveys, interviews, competitor testing—they’d nudge us with “But how does that feel to a 45-year-old parent?” or “What will make them say ‘Wow, that was easy!’ the very first time?” Their guidance gave our team’s decisions a sharper focus, ensuring we never lost sight of the human stories behind the data.
From the get-go, we had two remarkable mentors from Google: a Senior Staff UX Manager and a Google UX designer, who regularly challenged us to keep user empathy at the forefront. At every step—surveys, interviews, competitor testing—they’d nudge us with “But how does that feel to a 45-year-old parent?” or “What will make them say ‘Wow, that was easy!’ the very first time?” Their guidance gave our team’s decisions a sharper focus, ensuring we never lost sight of the human stories behind the data.
Competitive Benchmarking
Competitive Benchmarking
Competitive Benchmarking
Before we got too deep, we studied how the competition addresses voice interactions and setup:
Before we got too deep, we studied how the competition addresses voice interactions and setup:
Amazon Echo (Alexa)
Amazon Echo (Alexa)
Known for broad “Skills,” though not necessarily for frictionless personalization.
Known for broad “Skills,” though not necessarily for frictionless personalization.
Apple HomePod (Siri)
Apple HomePod (Siri)
Polished hardware synergy, but locked into the iOS ecosystem and can still struggle with context.
Polished hardware synergy, but locked into the iOS ecosystem and can still struggle with context.
Yandex Alice
Yandex Alice
A revelation for us—especially its fluid conversation style and advanced accent handling. Even though it’s region-specific, we saw how an AI could proactively interpret users’ half-finished commands without frustration.
A revelation for us—especially its fluid conversation style and advanced accent handling. Even though it’s region-specific, we saw how an AI could proactively interpret users’ half-finished commands without frustration.
These insights showed us that big leaps in voice recognition and proactive user assistance are possible, but rarely realized in everyday practice.
These insights showed us that big leaps in voice recognition and proactive user assistance are possible, but rarely realized in everyday practice.
Understanding User Frustrations
Understanding User Frustrations
Understanding User Frustrations
We combined surveys (reaching a broad audience) with one-on-one interviews to unearth deeper stories.
We combined surveys (reaching a broad audience) with one-on-one interviews to unearth deeper stories.
A father of two in his early 40s told us: “I wanted a device that says, ‘Hi! Let’s get you set up!’ Instead, I got this silent speaker. It’s like opening a puzzle box.”
A father of two in his early 40s told us: “I wanted a device that says, ‘Hi! Let’s get you set up!’ Instead, I got this silent speaker. It’s like opening a puzzle box.”
Another millennial, a teacher, criticized how personalization was “all or nothing” in the phone app. “I only discovered advanced routines by accident,” she said. “There was no voice clue to guide me—just an ocean of app settings.”
Another millennial, a teacher, criticized how personalization was “all or nothing” in the phone app. “I only discovered advanced routines by accident,” she said. “There was no voice clue to guide me—just an ocean of app settings.”
In short, these people wanted to trust Google Home but felt the device let them down at each step.
In short, these people wanted to trust Google Home but felt the device let them down at each step.
We also ran a buy-a-feature exercise where we gave participants “$100” to spend on desired improvements. Overwhelmingly, they allocated big money toward intuitive setup flows, better multi-user personalization, and a more “human-like AI” that can flex with them.
We also ran a buy-a-feature exercise where we gave participants “$100” to spend on desired improvements. Overwhelmingly, they allocated big money toward intuitive setup flows, better multi-user personalization, and a more “human-like AI” that can flex with them.
Usability Tests: How We Went Deeper
Usability Tests: How We Went Deeper
Usability Tests: How We Went Deeper
Onboarding: "Unbox and set up the device. Let us know each time you feel stuck or forced back to the phone."
Onboarding: "Unbox and set up the device. Let us know each time you feel stuck or forced back to the phone."
Personalization: “Set up bedtime routines for multiple family members—any confusion or hidden features?”
Personalization: “Set up bedtime routines for multiple family members—any confusion or hidden features?”
Voice Assistance: “Try simple commands (time, weather), then multi-step tasks or less standard phrasing. Notice if it falters.”
Voice Assistance: “Try simple commands (time, weather), then multi-step tasks or less standard phrasing. Notice if it falters.”
Participants repeatedly bumped into the same issues:
Participants repeatedly bumped into the same issues:
No one could connect the smart plug. Every participant got lost in the menus. Some tried for 25 minutes before giving up entirely.
No one could connect the smart plug. Every participant got lost in the menus. Some tried for 25 minutes before giving up entirely.
The assistant wasn’t really helpful. Hands-free tasks? Didn’t feel so hands-free. In one scenario, a participant had to grab their phone just to see a recipe the speaker claimed it could help with.
The assistant wasn’t really helpful. Hands-free tasks? Didn’t feel so hands-free. In one scenario, a participant had to grab their phone just to see a recipe the speaker claimed it could help with.
Even confident users got stuck. The menus were so layered and inconsistent that everyone made wrong turns
Even confident users got stuck. The menus were so layered and inconsistent that everyone made wrong turns
The phone and speaker fought for attention. Participants didn’t know which device was responding. One looked back and forth between the two and said, “I can’t tell who I’m talking to anymore.”
The phone and speaker fought for attention. Participants didn’t know which device was responding. One looked back and forth between the two and said, “I can’t tell who I’m talking to anymore.”
The system didn’t guide anyone. If anything, it made people feel like they were the problem. Users were ready to abandon the experience entirely.
The system didn’t guide anyone. If anything, it made people feel like they were the problem. Users were ready to abandon the experience entirely.
One middle-aged participant put it bluntly: “After all that, I don’t even feel it knows me better. Why go through the trouble if it never evolves?”
One middle-aged participant put it bluntly: “After all that, I don’t even feel it knows me better. Why go through the trouble if it never evolves?”
We also mapped the app’s information architecture and ran unmoderated usability tests to catch lingering friction points. Most participants failed key tasks, and our card sorting results showed little overlap with the current IA, pointing to a deeper issue with discoverability, not just navigation.
We also mapped the app’s information architecture and ran unmoderated usability tests to catch lingering friction points. Most participants failed key tasks, and our card sorting results showed little overlap with the current IA, pointing to a deeper issue with discoverability, not just navigation.
Proposed Solutions
Proposed Solutions
Proposed Solutions
01
01
01
Redesignin the Onboarding Process
Redesignin the Onboarding Process
People expected Google Home to guide them, not leave them in silence. To rebuild that first impression, we made onboarding conversational and intuitive.
People expected Google Home to guide them, not leave them in silence. To rebuild that first impression, we made onboarding conversational and intuitive.
We streamlined the mobile flow—removing unnecessary screens and combining redundant steps—reducing the original 12-screen process to just 5. This was based on direct observations of where users hesitated or got confused.
The redesigned speaker flow began with a voice prompt: “Hi, I’m your Google Home. Want help getting set up?” letting users feel welcomed from the start.
Instead of defaulting to app menus, the speaker narrates each setup step while the app quietly supports tasks like Wi-Fi and terms of service, establishing a clear role division and maintaining a smooth, synchronized experience.
We also implemented progressive disclosure: deeper personalization and multi-user features are introduced post-onboarding, only after users have built initial confidence.
02
02
02
Designing A Voice Guided Experience
Designing A Voice Guided Experience
Making the Google Home app easier to navigate was important—but it wasn’t enough. Our testing also revealed that users were less burdened by voice inputs. So we proposed giving users a choice. Voice, offered a more direct, effortless way to get things done—if it actually worked.
Making the Google Home app easier to navigate was important—but it wasn’t enough. Our testing also revealed that users were less burdened by voice inputs. So we proposed giving users a choice. Voice, offered a more direct, effortless way to get things done—if it actually worked.
Getting It Right
Getting It Right
We didn’t just want to add a voice only option—we wanted to make it feel right. Through open-ended Maze prompts, we collected feedback about how people wished their assistant would sound, act, and respond. That shaped how we defined a better experience: conversational, fast, and actually useful.
We didn’t just want to add a voice only option—we wanted to make it feel right. Through open-ended Maze prompts, we collected feedback about how people wished their assistant would sound, act, and respond. That shaped how we defined a better experience: conversational, fast, and actually useful.
To test it, we ran 8 voice scenarios with 10 participants—each featuring:
To test it, we ran 8 voice scenarios with 10 participants—each featuring:
Google Assistant’s current response
Google Assistant’s current response
Yandex Station’s “Alice”
Yandex Station’s “Alice”
A custom voice response we wrote
A custom voice response we wrote
Users weren’t told which was which, but most chose ours. They liked the tone, clarity, and how quickly it got to the point. That was a turning point—it showed that good voice UX isn’t just about AI—it’s about the way it listens, thinks, and responds.
Users weren’t told which was which, but most chose ours. They liked the tone, clarity, and how quickly it got to the point. That was a turning point—it showed that good voice UX isn’t just about AI—it’s about the way it listens, thinks, and responds.
Building Our Prototype
Building Our Prototype
Building Our Prototype
Based on our findings, we knew our redesign needed to address fragmentation, clarity, and the assistant’s overall tone. We co-developed conceptual features to tackle these gaps—then brought in a teammate with technical prototyping experience to help bring those ideas to life.
Based on our findings, we knew our redesign needed to address fragmentation, clarity, and the assistant’s overall tone. We co-developed conceptual features to tackle these gaps—then brought in a teammate with technical prototyping experience to help bring those ideas to life.
Instead of using complex voice prototyping tools, a rapid, low-budget prototype was built by repurposing a Google Home—installing a mini speaker and linking responses via a soundboard—and integrating a remote-controlled Figma phone prototype for dynamic interaction during a 'Wizard of Oz' test.
Instead of using complex voice prototyping tools, a rapid, low-budget prototype was built by repurposing a Google Home—installing a mini speaker and linking responses via a soundboard—and integrating a remote-controlled Figma phone prototype for dynamic interaction during a 'Wizard of Oz' test.
Though scrappy, the result was a working proof of concept that let us test the full experience, without building full-stack functionality.
Though scrappy, the result was a working proof of concept that let us test the full experience, without building full-stack functionality.
The Result
The Result
The Result
01
01
01
Voice-Guided Onboarding
The moment they power on, it greets them. A friendly, “Hi, I’m your Google Home. Want my help setting up?” No more dead silence.
The phone app remains an option for Wi-Fi or T&Cs, but the device narrates each step, bridging trust from the start.
Voice-Guided Onboarding
The moment they power on, it greets them. A friendly, “Hi, I’m your Google Home. Want my help setting up?” No more dead silence.
The phone app remains an option for Wi-Fi or T&Cs, but the device narrates each step, bridging trust from the start.
02
02
02
Seamless Personalization
TInstead of burying routines in a phone sub-menu, watchers say “Help me set my bedtime routine,” and the speaker gently clarifies each detail.
If the user consistently does certain tasks, the device asks if they want an automated routine to lighten daily friction.
Seamless Personalization
TInstead of burying routines in a phone sub-menu, watchers say “Help me set my bedtime routine,” and the speaker gently clarifies each detail.
If the user consistently does certain tasks, the device asks if they want an automated routine to lighten daily friction.
03
03
03
Advanced Voice Assistance
Code-named “Gemini,” our concept integrates a more robust AI (inspired by Yandex Alice’s fluid approach).
Multi-step or vaguely phrased requests become feasible. “Set three alarms: 7, 7:30, and 8 a.m.” is recognized in one breath.
Multi-user families can say “Dad wants a different news briefing than Mom,” and the device tracks each voice profile’s preferences over time.
Advanced Voice Assistance
Code-named “Gemini,” our concept integrates a more robust AI (inspired by Yandex Alice’s fluid approach).
Multi-step or vaguely phrased requests become feasible. “Set three alarms: 7, 7:30, and 8 a.m.” is recognized in one breath.
Multi-user families can say “Dad wants a different news briefing than Mom,” and the device tracks each voice profile’s preferences over time.
04
04
04
A More Humanized Tone
Subtle humor or empathy—like if you say “I’m bored,” it can respond with “Sorry to hear that! Up for music, a game, or checking new local events?” vs. a static, unhelpful remark.
Optional “chattiness” toggle for those who want purely functional responses vs. those who enjoy banter.
A More Humanized Tone
Subtle humor or empathy—like if you say “I’m bored,” it can respond with “Sorry to hear that! Up for music, a game, or checking new local events?” vs. a static, unhelpful remark.
Optional “chattiness” toggle for those who want purely functional responses vs. those who enjoy banter.
Validation & Findings
Validation & Findings
Validation & Findings
We brought participants back to test the redesigned flows—this time observing not just what they did, but how they felt.
We brought participants back to test the redesigned flows—this time observing not just what they did, but how they felt.
Results at a Glance:
Results at a Glance:
The SUS score experienced a dramatic transformation, leaping from a poor rating to one classified as excellent—a giant leap.
The SUS score experienced a dramatic transformation, leaping from a poor rating to one classified as excellent—a giant leap.
Onboarding
Onboarding
Down from ~17 minutes to ~5.
Down from ~17 minutes to ~5.
“I was done before I expected,” said one user, beaming at how the speaker itself guided them.
“I was done before I expected,” said one user, beaming at how the speaker itself guided them.
Personalization
Personalization
People found routines far quicker to set up verbally than rummaging through nested menus.
People found routines far quicker to set up verbally than rummaging through nested menus.
Voice Recognition
Voice Recognition
Even older testers with heavier accents or more casual speech patterns said the prototype “understood me first try!”
Even older testers with heavier accents or more casual speech patterns said the prototype “understood me first try!”
Those moments of ease—of trust being restored—were the real success metrics.
Those moments of ease—of trust being restored—were the real success metrics.
Conclusion
Conclusion
Conclusion
