This month we launched one of the most unusual systems we have ever been commissioned to build: KanadaRadio.com, an online radio platform where the DJs are real people sharing music on the social network Bluesky.
The pitch from the Kanada Team sounded almost simple. No fixed playlists. No songs stored on their servers. Stations that build themselves as the community posts music. Then the engineering questions start, and they do not stop.
A Database Doing the Work of a Radio Staff
A social network produces an enormous, messy stream of posts, and almost none of it is playable music. The platform has to watch that stream around the clock, pull out the posts that actually contain songs, verify each link still plays, and catalog every track by title, artist, and the person who shared it.
Then it all runs in reverse, on demand. A listener clicks a station and the system has to query thousands of constantly changing records and hand back a sequence that plays straight through. The sorting is the product. If the database gets it wrong, stations repeat themselves, hit dead links, or go quiet. When it gets it right, the listener never thinks about it at all, and that is the point.
The player had its own hard parameters: every track lives somewhere else on the internet, so the player is really a coordination layer. It queues media it does not control, detects when a track dies, skips without a gap, and reports the failure back so the catalog heals itself while people listen.
The Part Nobody Sees: Analytics With No Third Parties
Kanada Radio made a public promise to its listeners: no third-party ads and no third-party tracking. That ruled out the standard toolkit entirely. No Google Analytics. No ad network. No outside tags of any kind.
Keeping the promise still required knowing how the platform performs. So the Kanada Team commissioned us to build the measurement layer itself: an analytics program that runs on their own server. It measures traffic, listening time, and on-site interactions without advertising cookies or cross-site tracking. Daily rotating, one-way identifiers are designed to count visits without building a persistent profile that follows a listener from one day to the next.
A Complete Ad Platform, Built From Scratch
The same promise meant the ads had to be first-party too. We built a complete ad delivery, tracking, and management system: sponsors are created and scheduled from an admin console, placed into specific slots, weighted, given flight dates, and paused at will. Every impression and click is counted on the way through the platform's own server, and each sponsor gets a live results page with impressions, clicks, and click-through rate. No third-party verification tag is needed, because there is no third party involved.
The Takeaway
The listener sees a clean player and presses play. Underneath, a custom database is discovering music, checking links, building stations, measuring the audience, and running the ad program, all in real time. That is what we mean when we say the website is rarely where the work ends.
You can read the full engineering story in the Kanada Radio case study. Have an idea that needs custom software and serious database work? See our web applications page or contact us.
