What Are Meta Custom Audiences? Types, Lookalikes and How to Use Them for Conversions

What are Meta custom audiences and what do they do?

Meta custom audiences are ad audiences you build from your own data or from engagement on Meta apps, made up of people who already know your business. Website visitors, customer lists, app users and Instagram or Facebook engagers all feed Meta custom audiences. So instead of paying to reach strangers, you show ads to people who have already met you.

Meta describes a custom audience as a way for advertisers to build audiences from their own data sources or from Meta engagement data, for retargeting, new customer campaigns and other targeting options. In other words, the tool does far more than "show ads to cart abandoners". It also lets you exclude existing customers, seed lookalike audiences and steer Meta's AI toward the right people.

In this guide I walk through every custom audience type, based on Meta's own documentation. Then I cover lookalike logic, where custom audiences fit in the Advantage+ era, and your privacy obligations. I covered the creative and timing side of retargeting in a separate article; here the focus stays on the audience itself and on using the data correctly.

Which data sources can feed a custom audience?

Meta groups custom audiences by source. Specifically, each source collects data differently, matches people differently and suits a different job. Therefore it pays to check which source you are strongest in before you build anything.

In short, if you hold a strong customer list, start there. If your site gets solid traffic, start with website audiences. For new brands, engagement audiences usually provide the first pool of data.

How do website custom audiences work?

A website custom audience matches people who visit your site with accounts on Meta apps. The Meta Pixel or the Conversions API tells Meta which page someone opened and which event fired. You then write rules on top of those events: for example, people who viewed a product page in the last 30 days but did not buy.

Above all, the strength of this source is that it reads intent from behaviour. Someone who reads a blog post and someone who reaches checkout are not the same prospect. So instead of one "all visitors" audience, I recommend building behavioural layers:

• Hot layer: People who added to cart or started checkout but did not finish.
• Warm layer: People who viewed a product or service page, or checked pricing.
• Cool layer: People who only visited the blog or the homepage.
• Exclusion layer: People who recently bought or submitted a form.

However, these layers only mean something when your events fire correctly. Visitors who decline cookies, ad blockers and browser restrictions all shrink the audience below your real traffic. That is why pairing the browser pixel with the server side Conversions API gives you a more realistic audience size.

Meta's developer documentation also states a limit of 10,000 website custom audiences per ad account. In practice, you will never get close. The real problem is managing dozens of scattered audiences that nobody remembers creating.

How do you prepare a customer list audience?

A customer list audience comes from identifiers you already hold, such as email, phone number, name and city, which you upload to Meta. Meta compares them with its own user data and adds the matching accounts to the audience. Rows without a match, however, simply stay out.

Meta's developer documentation sets a clear rule here: you must hash the data with SHA256, and Meta supports no other hashing method. When you upload a file through Ads Manager, the hashing happens in your browser during upload. If you send the list through the API, you handle the hashing yourself.

When preparing the list, I suggest these steps:

1. Write phone numbers in full international format, digits only, including the country code.
2. Convert email addresses to lowercase and remove leading and trailing spaces.
3. Add as many identifiers per row as you can; multiple keys raise the match rate.
4. Remove test orders, staff accounts and cancelled records.
5. Split the list by value: bought in the last 90 days, bought twice, highest order value.

Match rates often come in lower than expected. The usual reason is that people use a different email or phone number on their Meta account. So a low match rate does not always mean a broken file; still, cleaning up formatting errors improves it noticeably.

Who benefits from app activity audiences?

An app activity audience uses behaviour inside your mobile app. The Meta SDK, or a mobile measurement partner, sends events such as app opens, registrations, add to cart and purchases. You then define audiences on top of those events.

For example, users who installed the app but never purchased, users who have not opened it in 14 days, or users who tried a paid plan and cancelled can each become a separate audience. As a result, you can send the right win back message to the right person.

That said, this source is not essential for every business. If you have no app, or only a small active user base, website and customer list audiences will give you far more data. On the other hand, for ecommerce and subscription businesses with an app, the most valuable audiences often live here, because app users tend to be the most loyal segment.

Keep in mind that iOS privacy settings limit measurement on the app side. When a user declines tracking, some events never reach Meta. Consequently, you should compare app audiences with the real records in your store or payment system instead of trusting them blindly.

How do engagement audiences use Instagram and Facebook data?

An engagement audience includes people who interacted with your content on Meta apps. Sources include your Instagram professional account, your Facebook Page, video views, lead forms, Instant Experience ads and events. The big advantage: it covers people who care about your brand but never visited your website.

For new brands, in particular, this source is gold. Your site may still get little traffic, yet if your Instagram content reaches thousands of people, you can build a meaningful audience from them. For instance, people who watched most of a video, saved a post or sent you a message each give you a distinct signal.

When building engagement audiences, I separate them like this:

• Deep engagement: People who messaged you, saved a post, or opened a lead form without submitting.
• Medium engagement: People who visited your profile or watched a large share of a video.
• Light engagement: People who interacted briefly with a single post.

That said, do not confuse engagement with purchase intent. Someone who liked a post may not be ready to buy. So showing engagement audiences trust building, product explaining content before a hard offer usually works better. If you want to lift engagement itself, start by measuring where you stand with the Instagram engagement rate calculator.

How should you choose the retention window?

The retention window sets how many days a person stays in the audience after the qualifying action. In practice, this setting directly affects both the size and the temperature of the audience. A short window gives you a hot but small audience; a long window gives you a larger but colder one.

Meta's developer documentation lists a range of 1 to 180 days for website audiences, while the FAQ section of the same page says the longest duration can reach 365 days. Engagement audiences also let you look back up to 365 days. For that reason, treat the option you see on the audience creation screen in your own account as the real upper limit.

To choose the window, look at your product's decision cycle. For a low price, impulse purchase, 7 to 14 days is often enough. In contrast, for considered purchases such as furniture, training programmes or B2B services, 60 to 180 days makes more sense.

In practice I recommend several windows for the same behaviour: 7, 30 and 90 days, for instance. Then you exclude them from each other to get a clean layered structure. That way each layer receives its own message and offer, and the hottest layer gets the highest budget.

What is a lookalike audience and how do you pick the source?

A lookalike audience takes a custom audience as its source and finds the people in your chosen country who most resemble it. According to Meta, you can build a lookalike once you have a custom audience with at least 100 people. So custom audiences also supply the raw material for finding new customers.

Put simply, the quality of the source decides the quality of the lookalike. If you use all site visitors as a source, the system looks for people like your visitors, and most of those never buy. By contrast, when you seed it only with buyers, repeat buyers or high value customers, the system focuses on a far more valuable profile.

My usual order of preference for sources looks like this:

1. Your most valuable customers: repeat buyers or top revenue contributors.
2. All purchasers from the last 180 days.
3. Leads who submitted a form and later became customers.
4. Visitors who started checkout or added to cart.
5. Instagram and Facebook users with deep engagement.

Meta's documentation says a lookalike can take 1 to 6 hours to populate. Also, Meta refreshes the members of a lookalike roughly every 3 days while it belongs to an ad set. So as long as the source stays current, the lookalike updates itself too.

How should you set the lookalike percentage?

The lookalike percentage shows how much of the chosen country's population you cover. According to Meta's help centre, 1% includes the people most similar to the source, while 5% gives broader reach with a less precise match. In Ads Manager you typically choose between 1% and 10%.

Still, a smaller percentage does not always win. In a large market such as the United States or the United Kingdom, even 1% covers a huge number of people, which is plenty for most budgets. However, if you target a small country or a narrow region, 1% may turn out too small.

For testing, try this approach: build three audiences from the same source, at 1%, 1 to 3% and 3 to 5%. Then compare them with the same creative and the same budget. Whichever range delivers the lowest cost per conversion earns the extra budget.

One more point, however. As Meta's targeting AI improves, the edge of lookalikes over broad targeting varies from account to account. So treat a lookalike as a hypothesis to test, not a belief. When you read the results, check whether the difference is statistically meaningful with the A/B test calculator.

Do Meta custom audiences still matter with Advantage+?

Yes, but their role has changed. With Advantage+ audience, Meta uses its own AI to find the best people. You add custom and lookalike audiences to this system as "audience suggestions". According to Meta, the system first prioritises people who match your suggestion, then expands to broader audiences when it expects better performance.

In other words, Meta custom audiences now act less like a hard wall and more like a starting signal for the AI in most campaigns. The cleaner that signal, the faster the system learns. For example, adding your best customer list as a suggestion reduces wasted spend on the wrong people in the first days.

On the other hand, some situations call for strict limits. Meta offers controls the AI will not go beyond, including minimum age, location, language and custom audience exclusions. If you do not want existing customers to see a new customer discount, you keep them out with an exclusion.

So Advantage+ did not kill custom audiences; it split them into two jobs. On one side sits the suggestion that guides the system, on the other the exclusion that protects your budget. When you use both deliberately, you get the speed of the AI without losing control.

How do exclusion lists protect your ad budget?

An exclusion stops an ad set from reaching a specific custom audience. In my experience, it is the most overlooked use of Meta custom audiences, yet it saves money faster than anything else, because new customer budgets often end up on people who already bought.

These are the exclusion scenarios I see most often:

• New customer campaigns: Exclude purchasers from the last 180 days and your customer list.
• Lead generation campaigns: Leave out people who already submitted the form.
• Cart reminder campaigns: Remove anyone who bought in the last 7 days; otherwise a customer who finished the order still sees "your cart is waiting".
• Subscription businesses: Keep active subscribers away from trial offers.

When excluding, keep the audience current. If you upload a customer list once and forget it for months, recent customers will still see new customer ads. That is why refreshing the list regularly, or setting up an automated flow such as CRM integrated lead tracking, makes exclusions actually work.

Factor exclusions into budget planning as well. In your social media advertising budget, I suggest keeping new customer spend and existing customer spend as separate lines.

How can Meta custom audiences increase conversions?

Above all, conversion gains come from matching the right audience with the right message. Showing the same ad to cart abandoners and blog readers means missing both. Therefore each layer deserves its own offer, its own creative and its own frequency.

Here is the basic pairing I use. For cart abandoners, you remind them of the product and answer the likely objection: delivery time, return policy, payment options. For people who viewed a product but did not add it to cart, you show social proof, reviews and real use cases. Finally, for blog only visitors, a guide or a free tool usually works better than a direct sales pitch.

Before you blame the audience, check your checkout too. A broken payment step wastes even the best audience; the checks in my guide on reducing cart abandonment help here.

Also watch frequency. When you give a small hot audience a large budget, the same person sees the ad several times a day and soon gets annoyed. In that case, lowering the budget or extending the retention window works better. That way your Meta custom audiences campaign drives conversions without hurting how people see your brand.

Which audience fits each stage of the funnel?

In practice, lining up custom audiences against funnel stages clarifies which audience does which job. The mapping below gives most ecommerce and service businesses a solid starting point.

• Awareness: A lookalike of your best customers as a suggestion; existing customers as an exclusion.
• Interest: Video viewers, profile visitors, blog readers.
• Consideration: Product or service page viewers, people who checked pricing.
• Decision: Add to cart, checkout starters, people who opened a form without submitting.
• Loyalty: Purchasers; a customer list for cross selling and repeat orders.

The goal of this structure is to show each person a message that fits their moment in the journey. I explain the overall funnel design and the metrics for each stage in how to build a conversion funnel. You will find the creative and timing side of retargeting in strengthening your sales funnel with remarketing.

The critical detail, then: audiences must exclude one another. Someone who added to cart also sits in your product viewer audience. Without exclusions, the same person lands in two ad sets and your budget ends up bidding against itself.

What should you do when an audience is too small?

Small Meta custom audiences are a common problem, especially for new sites and niche B2B businesses. When an audience is too small, delivery slows down, frequency climbs fast and learning takes longer. Fortunately, a few practical fixes help.

First, extend the retention window. Using a 30 day cart audience instead of a 7 day one trades a little temperature for more volume. Second, merge similar sources: combining add to cart and checkout starters into one decision audience makes sense.

Third, bring in engagement audiences. Even if your site gets little traffic, video viewers on Instagram and Facebook often form a much larger pool. Fourth, grow your customer list; newsletter sign ups, old quote requests and trade show contacts all count, provided you have a lawful basis for them.

Finally, use the audience as a suggestion rather than a hard limit. In Advantage+ audience, a small suggestion tells the system where to start, while delivery spreads into a wider pool. The real way to grow audience volume, though, is to grow traffic; the methods in driving website traffic from social media will help.

What privacy rules should you keep in mind?

First, the moment you upload a customer list, you process personal data. Hashing makes the data unreadable, but it does not remove your legal responsibility. So before uploading, check why you collected the data and whether advertising use fits that purpose.

Meta's Customer List Custom Audiences Terms require advertisers to confirm they have the necessary rights and permissions to collect and share the data. In other words, the responsibility sits with you, not with Meta. Under GDPR and similar laws, transparency, a valid lawful basis and rules on international data transfers all belong at the centre of that assessment.

In practice I recommend these checks:

• State advertising and targeting use clearly in your privacy notice.
• Remove anyone who never gave marketing consent or later withdrew it.
• Never pass the list around in shared folders or email attachments.
• Confirm that your cookie banner really controls when pixel events fire.

I cover the website side of this in detail in how to build a GDPR compliant website. This article is not legal advice; if your list contains sensitive data, review it with a lawyer.

Which audiences does the health and finance restriction affect?

According to Meta's developer documentation, beginning September 2, 2025, Meta flags any custom or lookalike audience that suggests specific health conditions or financial status and prevents it from running in ad campaigns. The documentation gives examples such as "arthritis", "diabetes", "credit score" and "high income".

In practice, this rule hits sectors like health, beauty, clinics, insurance, lending and investing hardest. For example, a clinic's customer list audience named "diabetes patients" falls squarely within scope. Even the audience name or the criteria behind it can cause trouble if they imply a sensitive condition.

To work within the restriction, keep audience names neutral and base them on behaviour, such as "Customers last 90 days". Also define audiences by the relationship with your business, not by a health condition or income level: people who booked, newsletter subscribers, contact form senders.

Meta's customer list terms also state that audience names and criteria must not rely on health information, financial information or other sensitive categories. So this is not only a technical block; it is a contractual obligation. If you work in a sensitive sector, designing your audience strategy around this framework from day one costs far less than a paused campaign later.

What are the most common custom audience mistakes?

When I take over accounts across different industries, I run into the same mistakes again and again. Specifically, most of them are strategy mistakes, not technical ones.

1. One audience for everything: An "all visitors, 180 days" audience is both too cold and too mixed.
2. Forgetting exclusions: A buyer keeps seeing ads for the same product for weeks.
3. A stale customer list: A list you upload once and never refresh breaks both exclusions and lookalikes.
4. Lookalikes from weak sources: A lookalike seeded from engagement hunts for likers, not buyers.
5. Ignoring event quality: If the purchase event fires twice, audiences and reports break together.
6. No naming system: Names like "Custom audience 14" turn into a mess nobody understands a few months later.

In short, what these mistakes share is a set and forget mindset. A custom audience reflects your site and your customer base as they change. Reviewing the audience list monthly, archiving unused ones and tracking size changes prevents most of these problems.

To see which messages competitors use for retargeting, take a look at the ad library search tool; comparing creatives across audiences often sparks ideas.

How should you measure custom audience results?

Campaigns built on Meta custom audiences usually deliver conversions at a lower cost than prospecting campaigns. However, that low cost can mislead you, because some of these people would have bought anyway. So instead of looking only at in platform ROAS, question the incremental impact.

I recommend tracking these metrics together:

• Cost per conversion and conversion rate by audience layer.
• Frequency and the decline in click through rate over time.
• The share of sales from new versus existing customers.
• The gap between platform data and real revenue in your store or CRM.

To compare conversion rates by layer, use the conversion rate calculator; to see ad returns, use the ROAS calculator. Where possible, set up a holdout group with Meta's experiment tools to measure the true contribution of your ads.

Also tag your traffic sources properly by adding UTM parameters to ad links. If you give each audience layer its own tag with the UTM builder, your analytics tool will show which layer truly drives sales.

In what order should you set everything up?

Also, if you start from scratch, do not try to build everything on the same day. Set up the data layer first, then the audiences, and the campaigns last. The order below mirrors what my team and I follow when we take over a new account.

1. Install the pixel and the Conversions API; test purchase, add to cart and form events.
2. Check that cookie consent controls the events correctly.
3. Clean the customer list, confirm its lawful basis and upload it.
4. Build website audiences with 7, 30 and 90 day layers.
5. Add Instagram and Facebook engagement audiences.
6. Create lookalikes from your most valuable customers.
7. Attach exclusion rules to every campaign, then go live.

Above all, a good audience setup depends on knowing who you sell to. That is why a target audience analysis before setup clarifies which customer segment carries the most value. Then your source audience choice rests on data, not guesswork.

How do my team and I approach custom audience strategy?

With my team, the first job in any Meta ad account is auditing the data flow: event accuracy, cookie consent, the origin of the customer list and how often it updates. Next, we build layered audiences by funnel stage, exclusion rules and lookalike tests. Finally, we compare platform data with store or CRM data to measure the real contribution.

We run this work as the paid side of our social media management service. If you run an online store, our ecommerce consulting gives a more complete result, because we tie the audience structure to your product catalogue and checkout flow.

To sum up, custom audiences are one of the most powerful and most neglected tools in Meta advertising. When your data stays clean, your audiences stay layered and your exclusions stay current, you are far more likely to get more conversions from the same budget.

Sources: Meta Business Help Center: About custom audiences, Meta: About lookalike audiences, Meta: About Advantage+ audience, Meta for Developers: Customer File Custom Audiences, Meta: Customer List Custom Audiences Terms.