The Case for Fewer, Better Recommendations

Published August 12, 2026

Most recommendation systems, across music, video, shopping, are judged internally by one number: how much of what they showed you got engaged with. That number quietly rewards volume. Show more, catch more, keep the session going. It's worth asking what that optimization costs the person on the other end of it.

Volume and quality pull in opposite directions

A system built to maximize engagement has no incentive to stop once it's found something good. More is always better by that measure. That's why "recommended for you" sections tend to grow rather than sharpen over time. An endless list is a safer bet for engagement than a short, confident one, even when the short list would serve the listener better.

A long list quietly shifts the work onto you

Fifty recommended songs isn't fifty decisions made for you. It's fifty decisions handed to you, each one small, but adding up. A shorter set that a person chose does something a long list structurally can't: it makes the selection for you, so what's left is listening.

What a smaller set has to get right, because it can't hide behind volume

When there's no long tail to fall back on, every pick has to earn its place. That's a constraint, and a healthy one. It's much easier to bury three mediocre picks inside fifty than inside seven.

This isn't really about music

The same pattern shows up anywhere recommendation exists. More isn't a proxy for better. It's often the opposite. A platform that commits to a small, finite, well-chosen set is making a quiet bet that restraint serves the person using it more than volume does.

Read more

For what this looks like in practice, see the case for a five-minute daily music habit, or for the reasoning against algorithmic personalization specifically, read why algorithm-free music discovery is having a moment.