Meta targeting used to be about who you picked. In 2026, it is about what the model can infer from your creative and conversion signals.
If you keep building 2019 interest stacks, you get 2019 results.
Start With Broad Discovery Audiences
For cold discovery, restrict only what you must:
- geography
- language, if the song is language‑specific
- an age floor if genre fit needs it
Then let the model explore.
Geography and RPM: The Hidden Targeting Variable
Geography is not just about audience fit. It directly affects what each stream is worth. Based on Dynamoi's royalty data, per-stream payouts vary dramatically by country tier:
- Tier 1 (US, UK, Norway, Sweden, Switzerland): Spotify RPM ranges from $3.02 to $5.00+ per 1,000 streams
- Tier 2 (Mexico, Brazil, India, Southeast Asia): RPM drops to $0.50-$1.50 per 1,000 streams
- Tier 3 (many African and South Asian markets): RPM can fall below $0.30 per 1,000 streams
A $0.30 cost-per-save in a Tier 2 market looks cheap until you realize each stream pays 3-6x less than Tier 1. For revenue-focused campaigns, weight your geo targeting toward higher-RPM markets even if CPMs are higher. The math often works out: a $50 CPM in the US generating $3.02/1K RPM streams beats a $5 CPM in a Tier 2 market generating $0.80/1K RPM streams.
Broad works because:
- music taste correlates with hundreds of signals Meta already sees
- your creative is the real filter
When to Add Guardrails
Add guardrails only to fix clear mismatches.
Examples:
- a metal track showing to mainstream pop scrollers
- a Spanish‑language song over‑delivering to non‑Spanish territories
Guardrails can be:
- a handful of relevant interests
- excluded interests that are clearly wrong
- a narrower region split
Do not add guardrails because results feel “too broad.” Add them because results are wrong.
Advantage+ Audiences
Advantage+ audiences let Meta expand past your seed targeting if it thinks it can win.
For music, that is usually good. Taste clusters are wide. You want the model to find adjacent scenes you did not think of.
Use Advantage+ for discovery and intent campaigns unless you are in a strict regional or rights‑limited context.
Lookalikes Still Work, But Volume Rules
Lookalikes are useful only if you have real intent volume:
- at least a few thousand savers or engaged viewers
- clean server events, not bot clicks
If your seed is thin, broad will outperform a low‑quality lookalike.
Retargeting Windows for Music
People do not become fans overnight.
Use longer windows than ecommerce:
- 7 days for immediate follow‑ups after a drop
- 14 days for mid‑cycle reminders
- 30 days for catalog or tour pushes
Retarget on intent, not curiosity. A 95% video viewer is warm. A 3‑second viewer is not.
The Interest Targeting Mistakes to Avoid
- Stacking 20 micro‑interests. You collapse scale and force Meta into a narrow, often low‑quality pocket.
- Targeting only “big artists like yours.” You train the system to find fans of legacy names, not of your sound.
- Using interest targeting to compensate for weak creative. That never works for long.
Creative creates the audience, not the other way around.