Case study 02
coUniverse
A UX audit of a co-op RPG that turns personal growth into a multiplayer quest — reading the first-run experience end to end, then testing it hands-on, to find what would confuse or lose a new player before their first quest.
Role
UX Auditor / Product Designer
Engagement
Freelance, UX design & product audit
Timeline
4 months + present
01 — The engagement
Reading a new player's first ten minutes
coUniverse is a co-op RPG that turns personal growth into a multiplayer quest — players build an avatar, get matched into a "crew," and progress through story-driven quests that move between virtual gameplay and real-world missions, guided by an AI companion, Celestine.
I was brought in during the app's early version to review the first-run experience: onboarding (avatar, "the Reading," path choice, check-in) and the core navigation that follows (Atrium, New Log, Chat with Celestine). Any friction here has an outsized cost — a confusing or stalling first ten minutes is the difference between a player who reaches their first quest and one who doesn't.
Main screen — the door, and the floating-label CTAs around it.
02 — The UX challenge
Two different kinds of problem, both hiding in the same flow
Reading the flow end to end surfaced two distinct categories of issue, not one. The first was sequencing: moments where the order of the flow works against the emotional beat it's trying to create. The second was interface convention: patterns — floating labels, decorative-looking CTAs, autoplaying audio — that ask a first-time player to learn a custom convention where a familiar one would already work.
The tension underneath both
Where does this product ask a brand-new player to do extra work — emotional or interpretive — that the design itself could have absorbed instead?
03 — Onboarding sequence
The emotional peak lands before we know who the player is
The clearest example: the player touches a stone and receives "the Reading" — the moment the mountain responds to them — before Celestine has even asked their name. Right after that peak moment, they fill out name, avatar, and location like a form. The high point lands before the player feels like a person in the story, not after.
The emotional payoff is scheduled before its setup.
Celestine can't say the player's name during the biggest moment of the opening, because she doesn't have it yet — the flow spends its best beat on a stranger instead of a named character.
Grab the player's name early — skip avatar and location for now — so Celestine can say their name during the Reading, then finish the rest of check-in afterward. Smallest change, biggest payoff.
Two smaller issues sit in the same part of the flow. There's an unfinished node still labeled "draft" between the space shot and the mountain shot — the very first thing a new player sees, so it needs to be finished or removed; deleting it is the safest fix, since the flow works fine without it. Later, players are asked to choose a path — Inner, Outer, or Relational — framed as a real decision, but all three lead to the exact same place with the exact same result. The cheapest fix is a different visual right after the choice, so it at least visibly registers; the better long-term fix is to make the choice actually matter later, e.g. Celestine referencing it down the line.
04 — Method
How this was reviewed
A note on rigor, stated once: this was an expert UX review, not a moderated study with recruited participants. Findings came from reading the flow end to end and then testing it hands-on, on device — not from sessions with other users. I'm flagging that once here rather than re-qualifying every finding below.
What was reviewed
- Onboarding: entry, avatar, the Reading, path choice, check-in
- Core navigation: Atrium, El Cono, My Space, New Log, Chat with Celestine
How
Read the flow first to map sequence and intent, then tested hands-on on a physical device to catch anything that only shows up in real use — timing, audio behavior, whether controls actually register as tappable.
What I was looking for
- Where the sequence works against the emotional beat it's building toward
- Where a custom interface pattern is being learned instead of a familiar one reused
What it surfaced
A bug (a "Not now" button that doesn't skip anything), a priority audio issue, and a navigation pattern that's already solved correctly elsewhere in the same app.
05 — Navigation & structure findings
A pattern the app already knows how to do right
The main screen's four core actions — Step into the Atrium, I have a Key, New Log, Chat with Celestine — are floating text labels over the door art. They read as decorative, not actionable; a first-time player won't reliably register them as tappable. The same floating-label pattern repeats on El Cono's room-select screen ("Campus" hovering over the scene), so this isn't a one-screen problem.
El Cono — the floating-label pattern repeats here too.
My Space — a standard tab bar, already built and working.
What makes this an easy fix rather than a hard one: My Space already uses a standard tab bar (Today / Starwheel / Cosmic Blueprint / Logs) elsewhere in the same app. Extending that pattern to the main screen and El Cono is largely a reuse, not a new build.
06 — The key recommendation
Don't teach a new pattern when the app already knows the right one
Problem
Floating text labels over art don't read as tappable to a first-time player, and the pattern repeats across multiple screens.
Insight
The app doesn't need a novel solution here — My Space already has a working, familiar tab bar for the exact same job.
Decision
Adopt the standard tab bar system-wide, on the main screen and El Cono, instead of inventing a new interaction model.
Principle
Reuse the pattern the app already got right. Don't make players learn a new one.
The strongest visual asset on the main screen — the door itself — is currently just background art. Making the door the entry point (tap it, it opens into the Atrium) turns the app's best asset into its primary CTA, which is both more immersive and more intuitive than a floating label ever was.
07 — Priority fix
Audio that plays before the player says it can
Celestine's voice plays automatically on open, at full volume, regardless of the phone's own settings. This was tested with a sleeping baby nearby — the phone wasn't even on silent, and the audio still caught the tester off guard. Autoplaying voice on launch breaks a basic expectation around audio control, and the stakes are higher than usual here: guided meditation is a core feature of coUniverse, not a side one, so getting default audio behavior right matters more than it would in a typical app.
Current
Voice plays automatically on open, at full volume, with no regard for the phone's own volume or silent settings.
"The app decided I wanted sound on."
Recommended
Voice off by default. Player opts in via tap-to-play, or a visible mute toggle.
"I decide when I want sound on."
The guided meditation player — audio is core to the product, which raises the stakes on getting autoplay and volume defaults right.
Priority, not just important
This is flagged above the sequencing and navigation findings because it breaks trust on first launch, with zero user control by default — and because audio is central to what the product actually does.
08 — Visual consistency
Three screens, three different design systems
New Log and Chat feel visually outdated next to the Atrium, which breaks immersion on its own. Zooming out further, the pattern is bigger than those two screens: the Resonance Gate, Earth Grid, and Duo each read like a different product — frosted glassmorphism, a map-and-card interface, and flat single-accent illustration, none of them matching the Atrium's painterly gold-and-light language.
Resonance Gate, Earth Grid, and Duo — three visual systems that don't speak to each other, or to the Atrium.
The fix isn't just "make it consistent" as a slogan: bring New Log and Chat into the Atrium's visual language first, using a lighter variant suited to more functional, utility screens, then run a full design audit to consolidate everything into one library. AI tooling can speed up the audit and mapping work itself, but the fix still needs manual design QA — this isn't something to fully automate.
09 — Scope & confidence
What's grounded in a screenshot, and what isn't yet
Every finding in Sections 05–08 is anchored to an actual screen: the main screen, El Cono, My Space, Resonance Gate, Earth Grid, Duo, and the meditation player. The onboarding findings in Section 03 are not screenshotted the same way — the unfinished "draft" node, the stone-reading scene, the path-choice screen, and the contact-info screen with the "Not now" button are documented from reading and testing the flow, not from screens I can show here. I'm noting that distinction directly rather than presenting both at the same level of evidence.
10 — What's next
What I'd need to take this further
11 — Priority summary
What to fix first, and why
Priority fix
Audio autoplay — breaks user trust on first launch, with no user control by default.
High impact, low effort
Delete the unfinished "draft" step. Grab the player's name before the stone-reading moment. Fix the "Not now" button bug.
Structural
Replace floating-label navigation with a standard tab bar. Make the door the primary CTA on the main screen.
Ongoing
Visual-consistency audit across Atrium, New Log, and Chat — AI-assisted mapping, manually QA'd fixes.
12 — Contributions
What this engagement actually asked of me
End-to-end flow review
Read the full onboarding and core navigation flow, then tested it hands-on on device to catch what only shows up in real use.
Pattern diagnosis
Separated sequencing problems (the flow working against its own emotional beats) from interface-convention problems (custom patterns replacing familiar ones), rather than treating every issue the same way.
Prioritization
Ranked findings by real risk and effort — flagging the audio issue above structural navigation work, and separating a one-line bug fix from a multi-screen design-system audit.
Actionable recommendations
Paired every finding with a concrete fix and, where relevant, a cheaper interim option — not just a list of problems.
Next, I'd want the remaining onboarding screenshots to ground Section 03 the same way the navigation findings are grounded — see Section 10.