Lookalike Audiences for Music Ads

You hand Meta a source audience, which is a list of your best existing people. Meta then goes and finds new people who resemble them and lets you advertise to that new group. In Meta's own words, a lookalike is a way to reach new people who are likely to be interested because they are similar to your best existing customers. That is the official mechanic.

The lookalike is much bigger than the source. Meta takes a few thousand of your people and expands that into a far larger pool of people it thinks are similar. That expansion is the value and the risk at the same time. Some of those new people really do resemble your fans. Some are loose guesses Meta made to fill the pool. Your job is to give Meta a source clean enough that the guesses stay close.

This is one of three ways to pick an audience for a music ad. The other two, hand-picked interests and broad targeting, are covered in the targeting pillar. Read that first if you have not decided which approach you are testing.

Here is the part people skip. A lookalike is only as good as the source you build it from. The percentage slider, the size, the location, none of it matters if the source is weak. If you feed Meta a list of random page likers or emails from some off-brand giveaway, you get a lookalike of the wrong people. Clean source in, useful lookalike out.

The best sources for music are people who already showed real taste. People who clicked out from your smartlink to a streaming service. People who watched most of a video. People who visited your link. Or an uploaded list of real fans and past buyers, not a cold list you bought or scraped. These are people whose behavior proves something, and that is what makes the lookalike off them worth having.

This is also where vlvtn earns its place. The pixel-and-server conversion event your smartlink fires when someone clicks out to streaming is exactly the kind of high-quality signal that makes a good source. The custom audience of real click-outs becomes the source you build the lookalike from. vlvtn owns the tracking layer, which is the layer a good source depends on.

Meta has hard requirements on the source. You need at least 100 people in it, and Meta recommends a source of 1,000 to 5,000. A hundred is the floor where it will let you build at all, not the number you want. People who are in the source get excluded from the lookalike automatically, so you are always reaching new people, and you can build up to 500 lookalikes off a single source.

Once the source exists, you pick a percentage. That percentage is of the population of a location, and it controls how tight the match is. Here is how I think about it from running these.

The rule of thumb is simple: smaller percentage is a closer match, larger percentage is more reach. That smaller-is-closer guidance is a practitioner read, not a number Meta guarantees. I lean tight when the source is clean and only widen when a campaign needs more volume than the tight version can deliver.

Most lookalikes I see from independent artists fail for one reason, and it is almost never the percentage. The source is too thin or too dirty. Someone has 80 fans and tries to build off it, or they dump in a list of people who liked a meme post two years ago, or they upload an email list that is mostly contacts who never cared about the music. Meta builds a lookalike off that and the lookalike is a lookalike of nobody useful.

The fix is not a clever setting. It is patience on the source. Run ads to a real conversion event for a while, let the pixel collect people who actually clicked out, and build the lookalike off that once it is a thousand real people deep. A lookalike built too early, off too little, is a guess on top of a guess.

Here is the actual flow in Ads Manager. The platform reshuffles names and screens often, so treat this as the action to take, not exact button text.

Open Audiences in Ads Manager. Choose Create audience, then Lookalike audience. Pick or create your source: a custom audience off your pixel click-outs, your video viewers, your smartlink visitors, or an uploaded fan list. Then set the audience size with the percentage slider, starting tight at 1 percent when the source is strong. You need admin or advertiser permissions on the ad account, and admin permissions on the Page or Pixel if the source is a Page or Pixel.

Then point the ad set's conversion event at your smartlink click-out, the pixel-and-CAPI event vlvtn fires. The conversion signal is what lets the lookalike learn and improve. Read cost per result in Ads Manager and let the data tell you whether this audience beats your broad baseline.

A lookalike can come out huge and loose, especially at a higher percentage. You can tighten it by stacking another constraint on top. Target only the people who are in your lookalike and who also fit a second criterion: a related artist, a festival page, a genre interest, or a geography. That slices the loose pool down to its most relevant part.

That overlaps with how interest targeting works, so if you want to go deeper on related-artist and genre stacking, read interest targeting for musicians. And if you are still deciding whether to bother with a lookalike at all versus just going wide, read broad vs narrow targeting.

A lookalike is a prediction, not a fact. Meta is guessing who resembles your fans, and some of those guesses are wrong. That is the whole reason you read cost per conversion and let the numbers settle it rather than trusting the audience on faith. A targeted impression is not a listen, and a click-out is not a play, save, or follow. What happens on Spotify is the song's job.

And no audience rescues weak creative. The best lookalike off the cleanest source still loses money against a video people scroll past. Targeting is a lever. Creative is the bigger one. Build the lookalike right, then put a good ad in front of it.