Most audio players are built around songs — three to five minutes, skip freely, shuffle a queue. Sermons don't work that way. A 45-minute message has a structure, and losing your place in it is far more disruptive than losing your place in a playlist.
Koinonia's audio player is built around resuming exactly where you left off, every time, without needing to scrub back through minutes of audio to find your spot. Playback speed and a sleep timer matter more here than they do for music, because people fall asleep to sermons and want the app to stop, not loop.
The sleep timer specifically came out of listening to what users were actually doing, not what we assumed they'd want. A meaningful share of Koinonia's evening listening happens as people are falling asleep — which meant a player that kept playing all night, or worse, auto-advanced to a completely different message at 3am, was actively working against how the app was being used. The sleep timer isn't a nice-to-have convenience feature; for a chunk of the audience, it's closer to a requirement for the app to behave the way they expect at the time of day they use it most.
We also treat playlists differently — a sermon playlist is closer to a reading list than a party mix. Order matters, and the player respects that instead of assuming shuffle is always the right default.
Resume-position accuracy turned out to be a harder engineering problem than it looks. A user might start a message on a phone during a commute, lose signal partway through, and come back to it that evening — the player has to reliably reconcile local playback position with whatever got synced (or didn't) while offline, without ever silently losing someone's spot. Getting that wrong even occasionally erodes trust fast: a listener who loses their place twice stops trusting the resume feature at all and starts manually noting timestamps, which defeats the entire point of building it.
None of this is complicated engineering in the abstract. It's just building the player around how people actually listen to a 45-minute message — often while tired, often interrupted, often specifically trying to fall asleep — instead of assuming every audio experience should behave like a music app built for a very different listening context.
