Custom vs Lookalike Audiences in Meta Ads

Custom audiences and lookalike audiences are two of the highest quality targeting tools inside Meta Ads, and learning how to combine them changes the way you scale campaigns. You will see how to build them step by step, when to use each one, and how to avoid the mistakes that drain budget on cold traffic.

What is a custom audience in Meta Ads?

A custom audience is a group of people who already performed a specific action with your brand. This data is not public: it lives inside your tracking tools, mainly the pixel and the Conversions API, which means only you can use it [00:20].

A classic example is building a custom audience of everyone who bought from you in the last 15 days. That is a high value segment because it comes from your own event registry, not from generic interests or behaviors.

What is a custom audience? It is a list of people who already interacted with your brand (purchase, lead, add to cart, video view) tracked through your pixel, CAPI or Meta native sources, and reusable for retargeting or exclusion.

Which sources can you use to build it?

Meta splits the origins into two big groups [02:30]:

• Your sources: website, app activity, product catalog, customer list and offline activity. They depend on setups you own.
• Meta sources: video views, lead forms, instant experiences, shopping activity, your Instagram account, events, your Facebook page and Facebook posts.

For a customer list, you upload a CSV with fields like email, phone, country and name. If you also include an LTV (lifetime value) column, Meta can identify which users inside that list are more valuable and prioritize similar profiles later [04:10].

How do lookalike audiences work?

Lookalike audiences target new people who do not know you yet, but share characteristics with a seed audience you already trust. That seed is, almost always, a custom audience [01:05].

You need a seed of at least 100 people from the same country where you want to advertise. From there, Meta can project audiences that range from 1 million up to 10 million users [01:30].

How big should my seed audience be? At least 100 people from the target country. If you try to build a lookalike for the US using a seed mostly from Mexico, Meta will likely reject it.

What does the lookalike percentage mean?

The percentage defines how close the new audience is to the original one. A 1% lookalike is the most similar; as you move toward 4% or higher, you broaden the criteria and lose precision, but you gain reach [07:45].

• 1% to 3%: keep similarity high, ideal for most accounts.
• Unified ranges (0 to 3%): a single audience that covers the top tier.
• Split ranges (0 to 1%, 1 to 2%, 2 to 3%): only when you want to test which slice converts better.

Even a 1% lookalike already reaches around 1 million new people who behave like your buyers, so there is no need to push the percentage up just to chase volume.

When should you use each type of audience?

Use custom audiences in three clear scenarios [09:30]:

• Specific communication: talk only to people who added to cart and did not buy, for example with a tailored promo.
• Exclusions: in a cold campaign, remove people who already bought so you do not waste budget on them.
• Seed for lookalikes: when you want to scale, feed the lookalike with a quality custom audience.

Lookalike audiences fit two moments: when you want to scale and invest more, and when you need quality prospecting, reaching cold users that still resemble your best customers [10:40].

Should I exclude buyers from a cold campaign? Yes. Add a custom audience of recent buyers in the exclusions box so your prospecting budget only chases new users.

Time windows and naming matter

The day range you choose changes everything. Talking to someone who abandoned a cart 7 days ago is not the same as someone who did it a month ago. For high intent retargeting, use short windows (3 to 7 days). For broader exclusions, go up to 30, 60, 90 or 180 days [12:15].

Also, keep a clean naming convention. When you build several custom audiences, names like purchase_15d_personalizado save you hours later inside the campaign builder.

Best practices to keep your audiences healthy

For custom audiences, do not mix sources. If the audience is buyers, include only buyers; do not blend checkout starts or abandoned carts in the same list. You can mix them later at the campaign level.

Always exclude the final conversion in cold campaigns: in a sales campaign, exclude buyers; in a lead campaign, exclude existing leads. This protects your CAC and avoids serving ads to people who already converted.

For lookalike audiences, follow these rules:

• Use quality origins. If your seed is weak, the lookalike will be weak.
• Start with low percentages (1%). Move to 2% or 3% only if performance drops.
• Test different origins: buyers, add to cart, long page visits, high LTV customers.
• Compare them in separate campaigns, for example purchases vs purchases with LTV, and double down on the winner.

Which custom audience are you planning to use as your first lookalike seed? Share it in the comments and tell me which time window you would pick.