The hardest part of running a Meta ad is not making the creative or setting the budget. It is answering a deceptively simple question: who should see this? Ads Manager gives you several different machines for answering it, and they are not interchangeable. A customer list, an interest audience and a lookalike solve different problems, and picking the wrong machine for the job is how budget quietly disappears.
This guide walks through the audience types Meta Ads Manager offers, what each is genuinely for, and where each one is created inside the tool. It is deliberately concept-led, because the exact menu names and the place a given control lives move around between interface versions. For the strategy of choosing between these types (which fits a cold launch versus a retargeting push), read it alongside how to define your audience; this piece is about the machinery.
Two places matter for building audiences. There is a dedicated Audiences section inside the Ads Manager area, where you create and save audiences ahead of time and reuse them across campaigns, and there is the ad set itself, where you can define or select an audience while setting up a campaign. Meta documents the current path to both in its guidance on ad targeting (Meta Business Help Center); the important thing to hold onto is that a saved audience is just a reusable definition, while an ad-set audience can be built on the spot.
Saved and core audiences: people who do not know you yet
A saved or core audience is built from attributes Meta already knows about its users: broad demographics such as location and age, plus interests and behaviours. You assemble it by combining these, and Meta narrows delivery to people who match.
This is your tool for reaching strangers who fit a profile. It is how a local furniture maker reaches people in their city who have shown an interest in interiors and home renovation, when those people have never heard of the brand. You build one in the Audiences section or directly in the ad set, layering the attributes you want.
Two cautions. First, resist the urge to stack a dozen interests until the audience feels precisely you. Over-narrowing starves delivery of room to find results, and Meta's own guidance increasingly favours broader definitions that give the system more to work with. Second, interest categories are Meta's inference about a person, not a declared fact, so treat "interested in sustainability" as a signal rather than a certainty. What counts as an available targeting attribute also changes as Meta retires categories, so check the current options in Ads Manager rather than planning around one you used last year.
Custom audiences: people who already know you
A custom audience is the opposite starting point. Instead of describing strangers, you supply Meta with people who have already interacted with your business, and Meta matches them to accounts. These are your warm audiences, and they are usually your most efficient spend because you are talking to people who already have context. Meta's overview of the type sits in the Business Help Center (About Custom Audiences), and there are three sources worth understanding separately, because their privacy implications differ sharply.
Website or app activity, via the pixel. If you have Meta's pixel or SDK installed, you can build an audience of people who visited your site, viewed a product, or added to cart, and then advertise to them (the mechanism behind retargeting). This depends entirely on your pixel and events being set up correctly, which is the subject of the pixel, events and the Conversions API. It also depends on the visitor having consented to that tracking, which is not optional.
Engagement. You can build an audience from people who engaged with your content on Meta's own surfaces: watched a video, opened a lead form, interacted with your Facebook Page or Instagram profile. Because this data lives inside Meta, it sidesteps some of the cross-site tracking questions the pixel raises, which makes it a reliable warm audience even as signal from outside Meta gets harder to collect.
A customer list. You can upload a list of contacts (emails or phone numbers) and Meta will match them to accounts. This is the source that demands the most care, and it deserves its own section.
The customer-list caveats, stated plainly
Uploading a customer list is useful and legally loaded, so be precise about how it works.
Meta does not want, and you should not send, raw email addresses or phone numbers in the clear. The data is hashed (converted into an irreversible fingerprint) before it is used for matching, so that Meta compares fingerprints rather than reading your customer list directly. Meta's tools hash the data during upload, and Meta documents the process and the accepted formats in the Business Help Center (About customer lists); read that before your first upload rather than guessing at the format.
Hashing protects the data in transit. It does not grant you permission to use it. You may only upload contact details you have a lawful basis to use for advertising, which in most markets means the person gave you consent that covers this use. A list of everyone who ever emailed you is not a consented advertising audience. Under regimes such as the GDPR and South Africa's POPIA, using contact data for ad targeting without an appropriate basis is a compliance problem regardless of how the file is hashed. If you are unsure whether your consent covers advertising, the honest answer is usually that it does not, and the fix is to ask.
Match rates are also never total. Some contacts will not map to an account, so a list of a few thousand becomes a smaller matched audience, and very small lists may not produce a usable audience at all.
Lookalike audiences: more people like your best ones
A lookalike audience asks Meta to take a source audience you value (purchasers, high-value customers, a custom audience) and find other users who resemble them. It is how you scale beyond the people who already know you without falling back on guessing at interests. Meta explains the mechanism and the source requirements in the Business Help Center (About Lookalike Audiences), and the lookalike audience glossary entry gives the short version.
The quality of a lookalike is set almost entirely by the quality of its source. A lookalike of your actual paying customers is a genuinely useful audience; a lookalike of everyone who ever visited your site is a lookalike of a vague crowd. The instinct that matters here is to feed it your best, smallest, most defined source rather than your largest one.
You also choose how tightly the lookalike hugs its source. A tighter match resembles the source closely but reaches fewer people; a looser one reaches more but resembles the source less. Meta expresses this as a spectrum rather than a single setting, and the sensible default is to start tight and widen only if delivery needs the room. The specific size options are Meta's to set and do change, so read them in the tool.
Advantage+ audience: letting the system find them
More recently Meta has pushed towards handing the targeting decision to its own system. Advantage+ audience inverts the usual flow: instead of you defining exactly who to reach, you provide optional suggestions (audiences that have worked, attributes you care about) and the system treats them as a starting point rather than a hard boundary, exploring beyond them where it sees results.
This can outperform tight manual targeting, particularly now that the system has more signal than any human layering of interests. It also means you are trusting the delivery system, which makes your pixel, events and conversion data more important rather than less, because that is what the system learns from. Advantage+ audience and the older manual controls are both present in the tool, sometimes as a choice at the ad set, and Meta's descriptions of what each does are the authority on the current behaviour. Check them in the Business Help Center rather than assuming the version you read about last quarter still applies.
The reasonable posture for most brands is not "manual is safer" or "let the machine do it", but to test one against the other in separate ad sets, exactly as you would test any other variable, and let the results decide.
Sense-check the size and the overlap
Two quick checks save a lot of wasted spend.
Size. An audience that is too small starves delivery; one that is too broad for your budget spreads it thin. Ads Manager shows an estimated size as you build, and while you should not treat that estimate as a promise, an audience of a few hundred people or one the size of a whole country are both usually signs to rethink.
Overlap. If you run several audiences at once and they contain many of the same people, your own ad sets end up competing against each other in the auction, which wastes money. Meta provides an audience overlap tool for exactly this check. When two audiences overlap heavily, the fix is usually to merge them or to exclude one from the other.
Where Acumin fits
Acumin does not build your Meta audiences for you: they are created in Ads Manager, against Meta's own data, and that is where they belong. What Acumin does upstream is help you work out who the audience should be in the first place, which is the decision the tool cannot make for you. The audience definition work and the research in Meta's Ad Library both feed the choices you then express as saved, custom or lookalike audiences inside the tool.
Do this today
Go to the Audiences section in Ads Manager and build one saved audience for a cold prospect and, if you have the pixel installed and consent in place, one custom audience of recent site visitors. Do not attach them to a campaign yet. Naming and saving the two audiences you will actually use forces the "who" decision to happen before the budget pressure of a live campaign does, which is exactly the order you want it in.