Emmanuel Corona · UX Researcher, XR

Design

12-day instrumented field study · n=74 (capstone of a 3-phase mixed-methods lifecycle)

Evidence

GA4 + Firebase telemetry · think-aloud (n=9) · content analysis (n=36)

Analysis

Funnel analysis · thematic synthesis

Problem

A loved idea that users kept leaving

Resonate lets listeners drop timestamped comments on exact moments in a song, turning private listening into shared, annotatable musical meaning. The concept landed; the funnel didn't. Public-sharing bursts pulled in waves of new users who then cratered within a day.

67.2%of users abandoned at the first open / visit

16.4%returned the next day (1-day retention)

5.3%were still active by day 5

The open question was never whether people wanted Resonate. It was where they were leaving, and why.

Approach

Separate “won’t” from “can’t”

The diagnosis hinged on one fork: was this a value problem (people don’t want it) or an access problem (people can’t get in)? Those point to opposite roadmaps: one says redesign the product, the other says redesign the door.

So I instrumented Resonate with Firebase and Google Analytics to see exactly where users dropped in the funnel, then paired that behavioral signal with think-aloud sessions and a content analysis of what users did once inside. Quant locates the leak; qual explains it.

Process & artifacts

Triangulating the leak

I owned the quantitative pipeline end to end: event instrumentation (post_comment, filter_click, share), daily-active-user and retention tracking, and the conversion funnels from landing → view → play → share.

The funnel located it67.2% of users never made it past the first open. Of the few who reached a song page, 100% played it, so the leak sat upstream of any core interaction.

Think-aloud explained itWatching 9 first-time users attempt onboarding surfaced the wall: a Spotify developer-token retrieval step, required by Spotify’s SDK, that asks consumers to act like engineers.

Content analysis validated the valueAcross 36 comments, 35 were public. Users treated Resonate as a community, not a private notebook. The idea worked for the people who got in.

A mobile cliff appeared100% of mobile users abandoned before even viewing a song, a second, distinct failure the funnel made impossible to miss.

The qualitative evidence was blunt. A 60-year-old tester spent her entire session trapped in Spotify’s developer portal, never realizing she had to return to Resonate: “For developers? I’m not a developer. What are you talking about?” Users spent anywhere from 3 to 18 minutes on a single auth step built for the wrong audience.

Retention curve dropping from 100% to under 20% within 48 hours — The retention curve reads like rejection, until the funnel and think-aloud reframe it as an onboarding wall, not a verdict on the product.

Funnel: first open → view song → play song, with 67.2% lost at first open — The funnel localized the loss to a single upstream step, separating the access problem from everything downstream.

Impact

The churn was the door, not the room

67.2%First-open abandonment isolated as the single largest leak, and traced to developer-token onboarding, not the concept

100%of users who reached a song page played it. Core playback value held

100%mobile abandonment before viewing surfaced as a separate, critical failure

RoadmapRecommended consumer-grade Spotify OAuth, an embedded walkthrough + first-run tour, and social-discovery features to break the cold start

The study reframed an existential-looking retention chart into a scoped, fixable onboarding problem, and protected a validated core idea from being “fixed” in the wrong place.

My role

I was the data-analysis lead: I instrumented Resonate with Firebase and Google Analytics and built the funnel, retention, and event analysis that isolated the 67.2% first-open drop-off. I also worked as a UX researcher across the mixed-methods lifecycle (generative research, usability testing, and this 12-day field study), co-researching with Yoyo Xinman (co-researcher and PM) and Na Young Son (co-researcher).

The call I’m proudest of was methodological: pairing the behavioral funnel with think-aloud to separate “won’t” from “can’t.” That triangulation is what turned a frightening retention curve into a single onboarding bug.

Reflections & takeaways

The most useful thing I did was resist the obvious story. A curve that falls from 100% to under 20% in 48 hours looks like “nobody wants this.” The funnel and the think-aloud said the opposite: the people who got in loved it (100% view-to-play, 35 of 36 comments public, “love the concept, love the design”); they just couldn’t clear a developer-portal auth wall built for engineers. Separating a value problem from an access problem changes everything downstream: you don’t redesign the product, you redesign the door.

Resonate: mixed-methods UX research, Stanford CS 377U: Understanding Users, Spring 2025. Team: Emmanuel Corona (Lead UX Researcher & Data-Analysis Lead), Yoyo Xinman (co-researcher, PM), Na Young Son (co-researcher). 12-day field study, n=74.