Role & Team
I led this as Lead Designer for Google Fiber's customer-facing apps, and across this initiative I wore several hats in an agile, startup-style environment:
Lead Designer — the self-install MVP: defined every flow (happy path, edge cases, troubleshooting), directed the illustration and animation agency, partnered with the copywriter on all in-app text, and produced the final mockups and prototypes.
Lead Researcher — the large-scale benchmark study: designed the test framework, coordinated the research agency, and led planning and execution.
Director — the in-house video production that replaced the flow's static instructions.
I worked in close collaboration with product, engineering, a creative agency, a research agency, and internal Supply Chain, CPE, and Field Tech teams.
Background
Google Fiber, an Alphabet company, provides high-speed internet to homes and businesses across the US. Historically, getting a customer connected meant sending a technician to the home — reliable, but slow and expensive, and a major driver of customer acquisition cost (CAC).
Most eligible customers could technically install the service themselves. The problem was the experience: turning a hardware-and-software setup — physical install, cable management, extender placement, and software configuration — into something any customer could complete confidently, on their own, from a phone.
The Problem
The existing self-install path leaned almost entirely on static text and illustrations. In practice that wasn't enough for a hardware task:
Customers skimmed or skipped the written instructions, then failed at critical steps.
Extender placement — vital for good signal — was especially painful. Users had no reliable way to find a good spot, so they guessed, got poor signal, and called support.
Failed self-installs defaulted back to costly technician visits, undercutting the entire business case.
Process / Approach
I inherited wireframes built around step-by-step text and illustrations and began refining them with a copywriter and leading art direction with an illustration agency. But the more we detailed each step, the longer the copy got. I was convinced, per NN/g "People don't read" that customers would skim past it and fail at the hardware steps.
A benchmark study to test this text-and-image flow against the Google Home app (our 80% reference) was already scheduled, on a hard deadline: new Wi-Fi 6 hardware and the Fiber app were about to replace Nest for all customers. I flagged that we might be validating a flow I expected to fail — but the timeline held. So instead of fighting it, I used the study to get a definitive answer.
Research / Discovery
To find out whether the flow actually worked, I designed and led a large-scale benchmark study — the most operationally ambitious research of the project.
What we tested
The text-and-image MVP, head-to-head against the Google Home app. Home set the reference point at an 80% success rate; our own MVP's success rate was the unknown we needed to measure.
How we tested
Rather than a lab, we recreated the real conditions of a home install:
40 participants total — 20 on the Google Fiber app, 20 on Google Home.
Two 2-bedroom apartments rented and furnished to simulate real homes, running simultaneously.
Participants wore GoPro headsets with live closed-circuit monitoring.
Two weeks to complete, coordinated across the research agency, Supply Chain, CPE, and Field Tech teams (unit rentals, furniture, equipment, account setup).
What we learned
A quantitative matrix — install steps across the columns, users down the rows — pinpointed exactly where the experience broke. The answer to our open question: the MVP succeeded only ~40% of the time, far below the 80% reference. Two insights explained why:
Users bypass text. Many participants skimmed or skipped written instructions, causing failures across multiple steps.
Extender placement frustrated users. People couldn't reliably find a good spot for the extender, the single biggest source of frustration and a likely driver of support calls.
A qualitative report also inform the team of multiple areas of improvement that were categorized and resolved over time.
Solutions
1. Instructional videos replace static text — Insight 1
Anticipating the benchmark's result, I had parallel-pathed a solution: I shot the entire install on my own phone, cut a rough video prototype, and preference-tested it against the text version on usertesting.com. Presenting both studies together made the direction clear — and because the scrappy phone footage tested well, we realized we didn't need an agency to produce the finals. An engineer on the team who shot video on the side had pro equipment and offered to collaborate; I storyboarded and directed, and we shipped the video-led flow.
Before (MVP): the flow relied on illustrations and text that users skimmed or ignored, contributing to a low success rate in testing.
Now (Post-MVP): short, contextual video snippets guide customers through each step. I spearheaded Google Fiber's first-ever in-house, end-to-end video production — prototyping and A/B testing a video version against the MVP, securing stakeholder buy-in, running pre-production, production, and post-production, redesigning the flow around video, and coordinating development. What started as an amateur prototype became a professionally lit, full production shot on a built set.
Results of production data in the first month: ~84% of installs completed without a support call — up from ~40% in benchmark testing, and above the 80% reference.
2. Real-time extender placement — Insight 2
Before (MVP): placing the extender was trial and error. Users unplugged it, guessed at a spot, plugged it back in, waited ~2 minutes for it to boot, and only then found out if it was too close, too far, or disconnected — and repeated the whole loop. Signal was measured router-to-extender, invisible to the user.
Now (Post-MVP): the app calculates signal strength from the router to the phone in real time, so the customer walks the space and finds a good location live — sync, unplug, find the spot with the phone, plug back in. No boot-wait, no guessing, no repeat loop.
Result: 60% reduction in extender-related support calls vs. MVP.
Reflection / Learnings
The biggest lesson was that hardware-centric flows have to be tested in real environments — a lab can't surface the spatial, physical problems (like walking an apartment to place an extender) that decide whether a self-install succeeds. Renting and furnishing real apartments was expensive and logistically heavy, but it's what exposed the two insights that made the redesign work.
If I did it again, I'd validate the critical physical interactions even earlier — running focused tests on high-risk steps before a full MVP launch, rather than after. Future directions I explored in Google Fiber Labs point the same way: an AI network assistant and a LiDAR home-scan to make placement and troubleshooting even more precise.
The team working on the video shoot 🙌
Full presentation
Watch this YouTube video for a deep dive into this project.