Lookalike audiences and Advantage+ audience at a glance
A lookalike starts with a seed: a Custom AudienceAn ad audience built from the advertiser's own sources: customer lists, site visitors or engagers., a customer list, the people who follow a page, or events from the pixel or an app. Meta's help centre describes the steps: choose the source, choose a location, choose a size, and Meta finds people in that location who share qualities with the seed, excluding the seed members themselves. The lookalike audiences guide covers seed choice, value-based seeds, size tiers and stacking. The decision this page addresses is not how to build one but whether the ad set should be bounded by it.
Advantage+ audience is Meta's automated targeting option and, as the help centre documents, the default on new ad sets for most objectives. It splits ad set inputs into controls, which are honoured as hard limits, and suggestions, which tell the system where to start. Location, minimum age, language and excluded custom audiences are controls; included audiences, lookalikes, detailed targeting, gender and maximum age are suggestions. The Advantage+ audience guide sets out those controls and what broad delivery needs to work.
The two are not separate products competing for the same slot. A lookalike can sit inside an Advantage+ ad set as a suggestion, or it can define an ad set under the original options. The real comparison is between a lookalike used as a boundary and an Advantage+ audience that may or may not carry a lookalike as a starting hint.
Compare table
Lookalike audiences vs Advantage+ audience
| Row label | Lookalike audiences | Advantage+ audience |
|---|---|---|
| Audience source | A seed you choose, modelled into people who resemble it | The delivery system's predictions, starting from any suggestions you give |
| Boundary | A hard limit under the original audience options | Only the documented controls are limits; everything else can be crossed |
| Control | Seed, location and size are yours to set and read | Location, minimum age, language and exclusions; the rest is the system's |
| Data dependence | Depends on the quality of the seed | Depends on the conversion signal reaching Meta and the volume of events |
| Setup effort | Build and refresh seeds, choose sizes per market | Little setup; effort moves to signal quality, creative and exclusions |
| Scale | Limited by the size setting and the market's population | As wide as the controls allow |
| Testing | Easy to describe and hold constant as one arm of a test | Needs the A/B test tool, because its real audience is discovered afterwards |
| Existing-customer handling | Seed members are left out of the lookalike itself | Customers can be reached unless excluded with a custom audience control |
| Best-fit account | Thin conversion history, rules on who may be reached, or a controlled comparison | Strong pixel and Conversions API signal, varied creative, enough budget |
Two rows decide most cases: boundary and data dependence. A lookalike puts the decision about who sees the ad in the advertiser's hands, through the seed. Advantage+ puts it in the system's hands, through the conversion data. Whichever input is stronger in a given account usually wins.
When to choose Lookalike audiences
Choose a lookalike as a boundary when the account's own conversion history is too thin for the system to learn from. A new pixel, a new offer or a market with few purchases gives Advantage+ little to work with, while a good seed, such as a customer list filtered to repeat buyers or weighted by value, brings knowledge the system cannot yet have. The lookalike guide's advice on seeding from as far down the funnel as volume allows applies directly.
Choose it when a rule defines who may be reached. Under Advantage+ audience, gender and maximum age are suggestions, not controls, so a licence, a policy or a contract that limits the audience can only be enforced under the original audience options. A lookalike inside those options is a defined, describable population that the business can justify.
Choose it for controlled comparisons. A test of two seeds, two size tiers or two markets needs boundaries that do not move during the test. The lookalike guide notes that the create-multiple feature produces non-overlapping tiers from one seed, which is the cleanest way to compare sizes without the arms competing.
When to choose Advantage+ audience
Choose Advantage+ audience when the conversion signal is strong. That means purchase or lead events reaching Meta reliably, ideally through the pixel and the Conversions APIServer-side event sharing with Meta. together, with enough matching information to tie them to people; an optimisation event that reflects the business outcome; and enough of those events each week for the ad set to leave the Learning PhaseThe period after launch or a significant edit when an ad set's delivery is still being optimised.. With that signal in place, the documented behaviour is that the system finds converters a defined audience would have missed.
Choose it for prospecting at scale. A lookalike's size setting caps the audience, and the narrow tiers that perform best exhaust quickly under meaningful spend. Advantage+ removes the cap and lets creative do much of the targeting, so accounts with varied ads, different hooks, formats and angles, give the system more ways to find different people.
Choose it with suggestions that carry knowledge. A strong lookalike or customer list added as a suggestion gives the system a good starting point and costs nothing in reach. Keep exclusions as the backbone: excluded custom audiences are the one audience input that stays hard, which is how existing customers and recent converters are kept out of prospecting.
Decision table
| Situation | Choose | Why |
|---|---|---|
| New account or pixel with little conversion history | Lookalike from a strong seed, as a boundary or suggestion | The system has little of its own to learn from yet |
| Mature account with reliable pixel and Conversions API events | Advantage+ audience | Strong signal is what broad delivery needs |
| Rules limit gender or upper age of the audience | Lookalike under the original options | Those inputs are only suggestions under Advantage+ |
| Comparing two seeds or two size tiers | Lookalikes under the original options, in a split test | The arms need fixed boundaries |
| Scaling prospecting beyond what narrow tiers can supply | Advantage+ audience, with the lookalike as a suggestion | Removes the size cap while keeping the seed's knowledge |
| Prospecting spend reaching existing customers | Either, with a customer exclusion as a control | Exclusions are honoured in both |
| One creative and a small budget | Lookalike while creative and budget build | Advantage+ relies on creative variety and volume to learn |
Running a fair test between them
The usual mistake is to run a lookalike ad set and an Advantage+ ad set side by side in the same campaign and compare their reports. They overlap heavily, enter the same auctions for many of the same people and learn from each other's conversions, so neither report describes what its audience would have done alone. Meta's A/B test tool avoids this by assigning each person to one arm.
Hold everything except the audience constant: creative, optimisation event, placements, budget and schedule. Put the lookalike arm under the original audience options so it is a real boundary, and give the Advantage+ arm the same exclusions. Whether the Advantage+ arm also carries the lookalike as a suggestion is a choice to make deliberately and write down, because it changes what is being compared. Run long enough for both arms to leave the learning phase, and judge on the business result, such as cost per qualified lead or first-order margin, rather than on cost per click or cost per raw lead.
Read the outcome as a decision about the boundary. If the lookalike arm wins clearly, keep it as a fence. If the two are close, the lookalike is usually more useful as a suggestion inside a broad ad set, where it adds knowledge without limiting reach. Repeat the test after material changes such as a new Conversions API connection or a creative overhaul. The exclusions and overlap guide explains how to check overlap between ad sets outside a formal test, and the TikTok Ads versus Meta Ads comparison covers how the same choice between defined and automated audiences looks on another platform.
A candle maker tests its best lookalike against broad delivery
Pebblecroft Candles has run a 1% lookalike of repeat customers as a boundary for a year, at a cost per first order of about £22. It sets up an A/B test with £120 a day per arm for four weeks: arm A keeps the lookalike under the original options, arm B uses Advantage+ audience with the same lookalike as a suggestion, and both exclude the 6,500 existing customers. Each arm spends £3,360. Arm A records 146 first orders at about £23 each with an average order of £38; arm B records 168 at £20 with an average order of £35. At a 60% gross margin, arm B's orders return about £3,530 of margin against arm A's £3,330, so Pebblecroft moves prospecting to Advantage+ with the lookalike kept as a suggestion, and schedules a repeat test after its Christmas range launches.
Common mistakes
- Comparing two ad sets side by side instead of using the A/B test tool, so the arms share people and auctions.
- Treating a lookalike as a fence under Advantage+, when it has become a suggestion the system can cross.
- Relying on gender or maximum age to enforce a rule, when only the original options make them hard.
- Switching to Advantage+ on a leaky pixel, so the system learns from a partial record of who converts.
- Forgetting the customer exclusion and reading spend on existing buyers as prospecting success.
- Seeding lookalikes from page engagers or all visitors, then concluding that lookalikes no longer work.
- Judging the test on raw leads or clicks, which broad delivery often wins while losing on quality.
