Leo Bramwell

Recommendation systems are reactive. They show you new pop after enough other people have already found it.

If you only hear new pop through recommendations, you hear it late. The systems are good, but they are followers by design.

You can listen to our full playlist which contains the artist's music, and know more about the artist's work by scrolling down the page.

PLAYLIST

Recommendation Systems Are Reactive

Recommendation engines are frequently described as discovery tools, and structurally they are the opposite: they surface music after enough other people have already engaged with it.

The mechanism requires signal. A track needs plays, saves, completion rates and playlist adds before the system has anything to reason about. Until that data exists, the record is effectively invisible to the algorithm.

That means anything reaching you through recommendation has already been found by a substantial number of people. It is an excellent tool for catching up and a poor one for being early.

Follow Labels, Not Artists

The single most effective substitute is following labels, because a label is a curatorial filter operating ahead of any audience data.

Someone at a label heard the record, decided it was worth the cost of releasing, and staked their reputation on it. That judgement happens months before the first play, which is exactly the lead time recommendation cannot provide.

Labels also have identifiable taste. Once you know that a particular imprint reliably releases things you like, its release schedule becomes a more reliable filter than any personalised system, because it is not waiting for consensus.

Label-following has a practical wrinkle worth knowing: many imprints publish their schedule weeks ahead on their own site or newsletter, while streaming services only surface a record on release day. Subscribing to the newsletter is frequently the single largest gain in lead time available to a listener.

Radio Specialists Still Do the Work

Specialist radio shows remain one of the strongest early sources, and they are consistently underused by people who assume radio is a legacy format.

The relevant shows are hosted by people whose job is listening to enormous quantities of new music and selecting from it. They receive material before release and they play it before it charts anywhere.

Because these shows are archived and often released as podcasts, the constraint of broadcast time no longer applies. A weekly specialist show is a curated list assembled by an expert, available on demand.

Live Listings Are a Leading Indicator

Local venue schedules are a genuinely early signal and require no music knowledge to read.

Support slots in particular are informative, because promoters and headline acts choose openers, and those choices reflect judgements made by people close to the scene. An unfamiliar name opening for an artist you like is a recommendation with someone's reputation attached.

This is also the only source on the list that is geographically specific, which matters because scenes are still local before they are anything else.

Support slots also indicate direction rather than only quality. A promoter pairing an unfamiliar act with a specific headliner is making a claim about what that act sounds like, which is genre information you can act on before hearing a note.

Bandcamp Rewards Browsing

Direct-to-fan platforms are structured differently from streaming services in a way that suits early discovery.

Because artists upload directly and the platform does not gate releases behind editorial approval, material appears there first, often well before it reaches streaming. Tag-based browsing and label pages make lateral exploration straightforward.

The purchase-based model also means the signals you encounter are people spending money rather than passively streaming, which is a considerably stronger indication that a record is worth attention.

Accept a Worse Hit Rate

The honest qualification is that early discovery involves listening to a great deal that is not good, and this is not a flaw in the method.

Recommendation systems have a high hit rate precisely because they wait for consensus. Removing that filter means encountering unfinished, uneven and simply bad records alongside the good ones.

Anyone unwilling to accept that ratio is better served by the algorithm, which is a legitimate choice. Being early has a cost, and the cost is paid in time spent on records that do not work.

The ratio improves faster than people expect. After a few months of following the same sources, you learn which ones match your taste and which do not, and the hit rate rises without the sources changing at all.