Strategy

A music marketing plan that survives contact with a release

Most music marketing plans are a budget split across channels. The ones that work are a sequence, because in 2026 the thing that decides whether a record travels is not how many channels you bought but whether one sound page filled up fast enough for the algorithms to notice.

What actually changed

Discovery used to run through gatekeepers, and the platforms now publish how their recommendation systems work instead. An editor, a programmer, a playlist curator decided what got in front of people, which meant marketing was fundamentally the business of persuading a small number of well-placed humans.

Now the first filter is behavioural. A feed shows a clip to a few hundred accounts, measures what they do, and either expands or stops. Nobody was persuaded. Something was measured. Which means the job of a marketing plan is no longer to reach decision-makers, it is to generate enough real behaviour on one asset that the systems expand it on their own.

That has one uncomfortable consequence. If the record does not hold attention, there is nobody to convince instead. The old model let a well-connected campaign carry a weak record. The current one does not.

The four things a plan has to specify

Not channels. These four.

• The hook, as a timestamp. Which seconds of the record are being marketed. Written down, to the second, before anything is commissioned. A plan that says "promote the single" has not started.
• The lane. Which genre-specific slice of creators the record routes to, and why that slice. Country needs different accounts from drill, and a plan that treats "creators" as one pool will spend most of its budget on irrelevance.
• The window. Start date, wave structure, and end date. Short-form starting seven to ten days before release, sustained three to four weeks after, because a spike teaches recommendation systems nothing and a slope does.
• The number you are judging it on. Pick it before you start. Sound creations and save rate are the useful ones. Total views is the easiest to inflate and the least predictive.

How to split a budget, and the honest caveat

The caveat first, because it matters more than the split. There is no correct percentage, and anybody handing you one has not asked what the record is. It depends on whether the record is a priority single with one release window or a catalogue track being reintroduced, and those are genuinely different problems.

What is generalisable is the principle. Spend most of it on the thing you still own when the money stops.

Paid social has one job worth paying for in a music context: amplifying a clip that has already proved it holds attention organically. Running ads on an untested clip is buying reach for a guess.

Planning for a roster rather than a single

A label planning across ten releases has an advantage a single artist does not, and most labels leave it on the table.

Clearance set up once against the catalogue means every subsequent release inherits it, so release two launches in hours rather than weeks. That turns a per-release negotiation into a standing capability, and it is the difference between running four campaigns a year and running twelve.

The second roster advantage is learning. Each campaign tells you which hook shapes travel for that artist and which creator lanes convert. Run them in sequence rather than in parallel and campaign four is briefed with real knowledge instead of assumption. Run them all at once and you learn nothing.

The five ways plans fail

1. Everything launches on release day. The sound page starts empty on the one day it needed to be full.
2. Clearance runs in parallel with promotion. Then a publisher takes three weeks and the campaign either waits or posts on uncleared audio.
3. The sound page splits. Creators import audio instead of selecting the official sound, so four hundred creations become five separate pages of eighty. Getting your music onto TikTok properly covers the fix.
4. One burst, then silence. Three days of everything, then nothing. A spike resolves inside a week.
5. Judging it on views. The easiest number to inflate, and the one that correlates least with whether anybody listened. Streams and saves covers what to read instead.

Questions