Voices Radio is an independent online station with a tight community of hosts and listeners. The product I inherited was a static marketing site bolted onto a third-party player. Three things forced a rethink at once:

A second channel and second studio opening in Hackney Wick, with its own identity to maintain alongside the original.

A mobile app launching in the next couple of months, with no companion design.

A platform migration from Airtime to RadioCult, opening up new functionality the old site couldn't expose.

The brief from the founders was a website redesign. My job was to make sure the redesign actually solved the right problem — not just refreshed the surface.

Approach

I planned the discovery phase to triangulate between three sources:

User interviews with three recruited groups — A: existing Voices community, B: net-radio listeners outside the community, C: B2B (later deprioritised). Recruitment via the Voices community chat, founder posts, and personal invites.

Competitive analysis across 11 stations covering direct competitors (NTS, Worldwide FM, Noods, Kiosk, IDA, Radio Panini, Bloop, Crop Manchester, Netil, Do! You! Radio, HÖR) and indirect (Spotify, YouTube).

Feature inventory — a comparison matrix of features each station offers, and which they don't.

What I was trying to learn

Why do listeners come to net radio at all when Spotify exists?

What pulls them to Voices specifically vs other London stations?

Which features feel essential, which are friction, and which are missing entirely?

Key research insights

After affinity-mapping the interviews, four insights kept resurfacing:

Listen-back is the dominant mode, not live. Most listeners can't catch shows live; they come back in the evening. The current site treats listen-back as a secondary archive.

Hosts are the unit of loyalty. People follow DJs, not channels. They want to know when their host is on, listen back to that host, and discover other shows in that orbit.

Discovery is broken. No browse-by-genre, no host pages, no surfacing of what's currently playing. The chat is scattered and rarely read by hosts on air.

Community is the real differentiator vs Spotify. Listeners came to Voices for belonging, not utility. The site doesn't communicate that — Spotify wins on UX, but loses on meaning.

Competitive analysis — what I learned

NTS sets the benchmark for feature breadth: live notifications, schedule of followed hosts, picked-for-you recommendations, integrated chat (read-only, redirects to Discord). Worldwide FM is essentially "NTS Lite" — same shape, fewer features. Most other stations have surprisingly thin apps, which reads as an opportunity for Voices rather than a problem.

A few specific patterns I noted:

NTS shows monthly cadence labels and "next on air" times for special shows — small, high-value detail.

HÖR makes track IDs a paid subscriber perk — interesting monetisation lever.

Worldwide FM's player is non-sticky and cuts when you navigate away. Bad — and an obvious thing for Voices to get right.

Noods has host landing pages but doesn't link to them from show pages, because there's no account/follow feature to anchor them. Feature interdependencies matter.

Persona — Dora "the Explorer"

I built one primary persona to keep the team aligned — Dora, 35, designer, London, aspiring DJ. Her behaviour is representative of the dominant pattern across interviews: Spotify is her guilty utility; NTS is her primary net radio. She follows hosts, discovers via genre, and comes back to listen back in the evening. She comes to Voices for community, friendliness, versatility, and the ability to watch friends' shows on video stream.

Jobs To Be Done

I structured JTBD across functional, emotional and social dimensions to make sure the redesign wasn't only about features:

Functional — Step out of Spotify's algorithmic bubble. Curate music by real people. Discover across broad genre-spheres.

Emotional (personal) — A sense of belonging. Discover artists you feel a connection to. Learn the personalities behind the shows.

Emotional (social) — Feel encouraged and safe to get involved. Discover events for like-minded people. Understand what music the community appreciates.

Problem statement

This single statement reframed the whole project. The redesign isn't "a new website" — it's a system that makes the connection between listener, host, show, and community legible.

Prioritisation — MoSCoW

I ran a MoSCoW with the founders to scope the first release. Resulting priorities:

Must — Discovery, Host Page, Show Page, 2nd Channel surfacing, About Page.

Should — Unified Chat, Accounts, Favourites.

Could / Won't — to be confirmed.

This kept the redesign honest — Accounts and Favourites are obviously valuable, but Discovery + Host/Show pages are the load-bearing changes. The rest can ship in a v2 once usage data exists.

From current → prospective site map

The existing site's IA flattened everything into a marketing page with a player bolted on. The schedule lived in one place, hosts didn't have homes, and the second channel had nowhere to live. The new IA splits surfaces by intent:

Listen now / schedule — live state across channels 1 + 2, plus schedule navigation.

Discovery — browse by host, show, and genre. Search across all three.

Host & Show pages — first-class destinations, not nested archive items.

About / Visit Us / Work With Us — separates the editorial brand story from practical calls.

Account / Favourites — anchored later but designed for from day one.

User flows I prioritised

From the JTBD I picked five flows as the redesign's success criteria. If these feel effortless, the redesign works:

Find a host and listen back to their latest show.

Find a relevant show by genre, topic, or artist name.

Discover shows that played a track I love.

Save a show for later listening.

Be notified when my favourite host goes on air.

The decisions where the why matters more than the what — these are the moments hiring managers can probe.

01 — Live show as the centre of gravity

The Live Now card is the primary object on the Radio tab and centres on first load. When the user scrolls past it, a sticky mini-player slides in so play/pause is always one tap away.

Why: Worldwide FM's non-sticky player was the most-cited frustration in the competitive analysis. If listeners come for live + listen-back, the player has to behave like the OS's now-playing UI — present, persistent, never an accidental cut-off.

02 — Schedule with day navigation in the header

Yesterday / Today / Tomorrow chevrons live in a floating chip. The schedule is the page's spine; on day change, the mini-player appears so live audio doesn't get orphaned.

Why: Listen-back is the dominant mode (research insight #1). The fastest path to "a show I missed yesterday" is one tap from the live view, not a separate archive section.

03 — Hosts as first-class pages

Each host has a real page with bio, tags (genre, role), Subscribe / Share actions, and a chronological list of their shows. Subscribe is the anchor for the future Favourites + notifications system.

Why: Hosts are the unit of loyalty (research insight #2). Noods proves the cost of not doing this — host pages exist but go nowhere because there's no account scaffolding underneath. I designed Host pages and Accounts together, even though Accounts ships later, so the IA doesn't need to break later.

04 — Discovery (in progress)

The discovery surface should let users find a show by genre, by topic, by artist who hosted, or by track played. Wireframing in progress — pattern choice (faceted nav vs progressive search) to be decided.

05 — Two channels, one identity (in progress)

The Hackney Wick channel needs distinct identity without fragmenting the brand. Decision pending — current direction is a shared design system with channel-level accent colours and lockups, surfaced contextually.

Note: This project is in active design. Below are the screens completed so far. The hi-fi prototype, design system, and outcome metrics will be added as they land.

Mobile app — Radio / Live

Today's view with the live show card, schedule of the day, day-switcher in the header. Scrolled state with sticky mini-player. Yesterday / Tomorrow views with the same affordances. Expanded scroll showing past + upcoming shows in a single navigable column.

Host page

Host imagery, name, tags (genre, role). Subscribe + Share as primary actions. About + chronological show list with show artwork and metadata.

In progress

Discovery, Show LP, Web home, About scroll — currently wireframing.

To fill once the design is complete and shipped. Likely beats: what I'd test first if I had a usability round before launch; where I made a call without enough data and what I'd revisit; what the team learned about our research recruitment / interview structure.

Pending launch. Plan to capture:

Adoption — app installs, returning listeners, % live vs listen-back sessions.

Engagement — follows per host, average session length, chat participation.

Qualitative — founder + community feedback on whether the redesign feels like Voices.