Lookalike audiences find new listeners who resemble your existing fans. The catch: Meta can only find people similar to the seed you give it. Feed it clickers, and it finds clickers. Feed it savers, and it finds people who save music.
Most music advertisers build lookalikes from the wrong actions, then wonder why results do not convert.
Why Seed Quality Beats Seed Size
Meta recommends 1,000 to 5,000 people in your source audience. But the composition of that seed matters more than hitting a number.
A seed of 500 users who saved your track to their library is more valuable than 5,000 who clicked a link and bounced. The model learns from behavioral patterns, not just demographic overlap. If your seed is full of casual clickers, Meta finds more casual clickers.
Tip Build separate custom audiences for each streaming platform. Users who click through to Spotify behave differently than those who choose Apple Music. Platform-specific lookalikes let you optimize each funnel independently.
Savers vs Clickers: What to Use as Your Seed
The best seeds come from downstream intent signals, not upstream curiosity.
| Seed Type | Signal Strength | When to Use |
|---|---|---|
| Saves and library adds | Highest | Primary seed for discovery campaigns |
| Follows on Spotify or YouTube | High | Strong for artist-building campaigns |
| 95% video viewers | Medium-high | Good fallback when save data is thin |
| Link clicks | Medium | Use only if you lack intent data |
| Page engagers | Low | Avoid for music campaigns |
If you are running campaigns through a smart link, configure your pixel to fire on the save or follow confirmation, not just the landing page load. The difference in lookalike quality is substantial.
Works when: You have at least 500 users who completed a downstream action. The lookalike will reflect genuine music intent.
Fails when: Your seed is mostly curiosity clicks from a viral Reel. Meta will find more scrollers, not listeners.
Audience Size: 1% vs 5% vs 10%
The percentage controls how similar the lookalike is to your seed. Lower percentages mean tighter similarity but smaller reach.
In most accounts, 1% lookalikes are the best starting point because they stay closest to your intent seed. For music, where taste clusters are specific, tighter is usually better.
1% Lookalike
The top 1% most similar users. In the US, this is roughly 2 million people. Start here for performance campaigns where cost per save matters. If 1% does not work, wider percentages rarely fix the problem - the seed itself needs attention.
2-5% Lookalike
A balanced tradeoff between reach and relevance. Use this tier when 1% is working but you need more volume, or when your genre has broader appeal. Each percentage point roughly doubles the audience size.
6-10% Lookalike
Best for awareness campaigns or broad genres like pop and hip-hop. The similarity to your source decreases significantly. Users in a 10% lookalike are the least similar to your existing fans.
The practical rule: start at 1%, validate performance, then expand if you need scale and can afford slightly higher CPAs.
Layering Lookalikes with Interests
You can layer a lookalike with interest targeting to narrow the audience further. This helps when your seed is strong but the lookalike still delivers to mismatched pockets.
For example, if you have a 1% lookalike from your Spotify savers but notice delivery skewing toward non-music demographics, add 3-5 artist interests as a filter. You are not replacing the lookalike logic - you are adding a constraint.
Warning Meta interest taxonomies shift over time, and niche genre interests can get lumped into broader buckets. Test whether your interest filter actually improves results before committing budget.
Keep the layered audience above 100,000 people. If you narrow too aggressively, you lose the scale advantages that make lookalikes useful in the first place.
When Lookalikes Outperform Broad Targeting
Lookalikes are not always the right choice. Meta's Advantage+ system now creates implicit lookalike-style expansions automatically. But there are specific situations where explicit lookalikes still win.
Use lookalikes when:
- You have clean, high-intent seed data from tracked conversions
- Your genre is niche enough that broad targeting finds the wrong ears
- You are entering a new market and want to mirror success from another region
- Your account is new and Advantage+ has limited conversion history to learn from
Use broad or Advantage+ when:
- Your creative is strong and you trust the algorithm to find the right viewers
- You have deep conversion history and the model already knows your ideal listener
- Your seed is thin or polluted with low-intent actions
- You are spending enough to exit the learning phase quickly with broad targeting
Advantage+ can outperform explicit lookalikes on CPM, but cheaper impressions do not always mean better music fans. Measure cost per save or cost per follow, not just reach metrics.
Building Your First Music Lookalike
Create a custom audience from intent events In Meta Ads Manager, go to Audiences and create a custom audience from your website or app events. Select the event that represents genuine interest - saves, follows, or 95% video views. Exclude anyone who already follows you.
Wait for sufficient volume Give the custom audience at least 500 users before building a lookalike. Thin seeds produce unreliable models. If you are under 500, run more discovery campaigns to grow your intent pool first.
Create the lookalike at 1% Select your custom audience as the source. Choose your target country and start with 1% similarity. Name it clearly:
LAL 1% - Spotify Savers - US.Exclude your source audience When building campaigns, exclude the original custom audience from lookalike targeting. This keeps your prospecting clean and prevents wasting budget on people who already engaged.
Test against broad and measure downstream Run the lookalike against an Advantage+ or broad ad set with identical creative. Compare cost per save, not just cost per click. Give each enough budget to exit the learning phase - roughly 50 conversions per week.
Common Lookalike Mistakes in Music Campaigns
Using page engagers as your seed. Someone who liked a post is not the same as someone who saved a track. The algorithm cannot distinguish music fans from casual scrollers if your seed does not either.
Building lookalikes before you have intent data. A new artist with 100 link clicks does not have a usable seed. Run broad discovery campaigns first, build intent volume, then create lookalikes from that pool.
Never refreshing the seed. Lookalikes are static snapshots. If your fanbase evolves - new single, different sound, expanded genre - rebuild your custom audiences and create fresh lookalikes.
Stacking multiple lookalikes in one ad set. This dilutes the signal. Test each lookalike separately to understand which seed performs best, then scale the winner.
Rule: If your lookalike is underperforming broad targeting, the problem is almost always the seed, not the percentage.