The audience types in one map
Every ad set on Meta describes its audience with some combination of the rows below. The distinctions that matter in practice are where the data comes from and whether the definition is a boundary or a hint.
| Audience type | Built from | Boundary or hint | Typical use |
|---|---|---|---|
| Core audience | Location, age, gender, language, detailed targeting (interests, behaviours, demographics) | Boundary under original audience options; hint under Advantage+ audience, except location, minimum age and language | Prospecting where no first-party data exists; narrowing a brand reach buy |
| Custom audience | Your data: website visitors via the pixel, app users via the SDK, customer lists, catalog interactions, engagement with your Facebook or Instagram presence | Boundary when included or excluded under original options; a suggestion when included under Advantage+ audience, a boundary when excluded | RemarketingAdvertising to previous visitors or engagers., suppressing existing customers, seeding lookalikes |
| Lookalike audience | A seed custom audience, page or pixel source, modelled against a chosen country or region | Boundary under original options; hint under Advantage+ audience | Prospecting among people who resemble known converters |
| Advantage+ audience | Meta's prediction, informed by pixel and Conversions API data, past campaign results and any suggestions you add | Hint by design, with a short list of documented controls | Broad prospecting once conversion data is flowing |
Two things are missing from the table on purpose. Placements are a separate decision from audiences and are covered in the creative section, because the format and the surface are chosen together. And the conversion events that feed all of these audiences live in the measurement section, because an audience built on a broken pixel is a guess dressed up as a list.
The shift from manual targeting to signal and creative
For most of its history the craft of Meta targeting was additive: stack interests, layer behaviours, exclude competitors' fans, split everything into its own ad set and let the results tell you which slice worked. That approach assumed the advertiser could out-guess the delivery system. Meta's documentation now points the other way. Detailed targeting options have been pruned, the Advantage+ audience option is the default on new ad sets for most objectives, and the Help Centre's own guidance favours broad definitions with a clear optimisation event.
The reasoning is mechanical rather than ideological. The delivery system predicts, for each person and each auction, how likely that person is to take the optimisation event. A narrow audience limits the pool it can search; a broad one lets it search everywhere and spend where the predictions are strongest. The prediction is only as good as the data behind it, which is why the two things that now decide performance are the signal (how many conversion events reach Meta, how accurately, and how quickly) and the creative (which is what the system uses to find the people who respond). Audience definitions still matter at the edges: suppressing buyers, keeping a retention message to known customers, honouring a geographic licence. They matter less in the middle, where the system is usually better at finding the next converter than a list of interests is.
In practice, that leaves advertisers with a different job. Instead of building the audience, they feed it: install the pixel properly, add the Conversions API, choose the optimisation event that reflects the real business outcome, and then make enough creative variants to give the system something to learn from. The targeting guides in this section are written with that frame. Each one explains the documented mechanics of an audience type, where it still earns its place, and where it has become a constraint that costs more than it saves.
The core guides
The core guides each own one decision.
- Custom audiences: sources, matching and uses covers the audiences built from your own data: website and app events, customer list uploads and how they are matched, engagement audiences from your Facebook and Instagram presence, the retention windows that govern each, and the consent requirements that come with them.
- Lookalike audiences: seeds, sizes and strategy explains how a Lookalike AudienceAudience similar to a seed list. is modelled from a seed, why the choice of seed matters more than the size setting, how value-based seeds change the model, and how lookalikes behave once Advantage+ audience treats them as suggestions.
- Advantage+ audience: suggestions, constraints and when to use it sets out the documented controls that remain hard limits, the inputs that are only hints, what broad delivery needs from signal, creative and budget before it works, and a fair test design for deciding between automated and defined audiences.
- Exclusions and audience overlap deals with keeping ad sets from competing with each other and keeping buyers out of prospecting: the difference between audience overlap and auction overlap, how to read the overlap tool, and exclusion rules that keep a funnel clean.
The author's remarketing expertise page describes how these pieces are used together on real accounts, across Meta and other platforms.
Further guides
Further guides take one of those decisions through the work step by step:
- Uploading a customer list to Meta is a tutorial for preparing and uploading a customer list: choosing the right segment, the identifier columns, formatting, hashing, the data origin, a value column and scheduled refreshes.
Audiences by funnel stage
The See-Think-Do-Care framing is a useful way to decide which audience type carries which message, because each stage has a natural data source.
See (awareness). The job is ReachUnique people who saw an ad. among people who could plausibly become customers. Broad core audiences or Advantage+ audience with a reach or awareness objective fit here; lookalikes built from a wide seed can narrow the pool where a product is genuinely niche. The reach and frequency calculator helps size a budget against an audience before launch.
Think (consideration). People who have shown a flicker of interest: video viewers, page engagers, visitors who read but did not act. Engagement and website custom audiences with short windows are the natural source, and lookalikes of engaged people extend the pool.
Do (conversion). Two different jobs. Prospecting for new converters is where Advantage+ audience or value-based lookalikes do their best work. Retargeting people who added to basket or started a form is a custom audience job, with exclusions for anyone who has already completed the action.
Care (retention). Customer list audiences and purchase-event audiences carry replenishment, cross-sell and loyalty messages, and the same lists are excluded from every prospecting ad set so the acquisition budget is spent on strangers.
The stage boundaries also tell you where FrequencyAverage times a person saw an ad. needs watching: the further down the funnel, the smaller the audience and the faster the same people see the same ad.
Common targeting problems
Most targeting complaints resolve to one of a handful of causes. The table sends each to the guide that handles it.
| Symptom | Likely cause | Where to look |
|---|---|---|
| Prospecting ads keep reaching existing customers | No customer list or purchase-event exclusion on the prospecting ad sets | Exclusions and overlap |
| A website custom audience is far smaller than site traffic suggests | Pixel firing on some pages only, consent blocking events, or a window shorter than intended | Custom audiences |
| Lookalike performs no better than a broad audience | Seed too small, too old, or built from a low-intent action | Lookalike audiences |
| Delivery reaches an age or region you thought was excluded | Those inputs were suggestions under Advantage+ audience, not constraints | Advantage+ audiences |
| Two ad sets show the same people and costs rise | Audiences overlap and ad sets compete; consolidate or exclude | Exclusions and overlap |
| Advantage+ audience spends but converts poorly | Weak or sparse conversion signal, or too few creative variants for the system to learn from | Advantage+ audiences and the measurement section |
| Retargeting keeps showing people the item they just bought | No purchaser exclusion, an unmatched purchase event, or an exclusion window shorter than the retargeting window | Retargeting ads reaching recent buyers |
| Retargeting audience is exhausted within days | Audience is small and nothing refreshes it; move the message up the funnel or widen the window | Custom audiences |
| A customer list or retargeting audience is too small to deliver or seed a lookalike | Few identifiers, a rule that seldom fires, a short window or consent limits | Why a custom audience stays too small |
| A customer list was uploaded once and no longer reflects who the customers are | No clear job for the list and no scheduled replacement | Uploading a customer list |
| Advantage+ audience prospecting keeps spending on past buyers | Missing or stale exclusions, buyer-based suggestions, or a goal that favours repeat purchases | Advantage+ audience reaching existing customers |
Where a problem is really about creative wear-out rather than audience definition, the creative section is the better starting point.
