What should I use instead of lookalike audiences?
An audience built from behavior rather than from a platform's model of your seed. The lookalike idea is fine. The source degraded. Watt composes the audience from live signal on what people buy, follow and are in-market for, which gives you something that does not decay with one platform's graph. Then use it three ways: seed a lookalike with it, hand it to Advantage+ as a suggestion, or exclude against it.
My lookalike audience stopped performing. Why, and what do I do?
Two things changed underneath it. Your seed thinned as pixel and platform signal degraded, and the audience itself got demoted: under Advantage+, custom and lookalike audiences are suggestions Meta can go outside of, not filters it has to respect. Meta's own documentation used to say suggestions would be prioritized before it searched more broadly, and the word “prioritized” was removed. Rebuild the seed from behavior rather than from whatever the pixel happened to catch, and stop expecting the audience to act as a constraint.
How do I build a custom audience based on intent rather than demographics?
Start from an action, not an attribute. Someone who bought running shoes last month is a different proposition from someone a model thinks is 25 to 34, and the gap between those two is measured: peer-reviewed field work found third-party brokers identifying “males 25 to 54” correctly 24.4% of the time against a 26.5% population base rate, and gender alone at 42.3% (Neumann, Tucker and Whitfield, Marketing Science, 2019). Better, not perfect, and it applies to Watt's category as much as to anyone's. Watt builds from behavior for that reason, and states what it can verify rather than claiming a precision nobody in this category has.
How do I build an exclusion list of bad-fit customers?
Compose or upload the list, then exclude it at the ad-set level. Do this carefully, because exclusions are one of the few audience inputs Meta still has to obey. Under Advantage+, location, minimum age, language and excluded custom audiences are hard constraints. Everything else is a suggestion. Watt can compose the exclusion from behavior, so you can exclude by what people do rather than only by who already sits in your CRM.
Should I use Advantage+ or build my own audience?
Both, in that order. Advantage+ is the default and fighting it usually costs money, but it takes suggestions, and providing them helps most on new accounts and thin-data accounts where Meta has no history to learn from. Override when you are in a regulated category, selling an age-restricted product, running strict suppression, or spending too little for Advantage+ to ever gather the volume it learns from. The regulated case is harder than it looks. In Meta's Special Ad Categories, Advantage+ Audience is on by default and most manual controls are gone: no age or gender targeting, no postcode targeting, no detailed targeting. Compose the audience in Watt, hand it over as a suggestion, and keep the exclusion as your one hard control.
What should I use instead of Meta detailed targeting?
Start with a correction, because it changes the answer. Detailed targeting was not removed. What went was the ability to exclude on it, and later a set of individual options. What remains still works, as a suggestion. So the replacement is not another dropdown. It is an audience composed outside the platform and handed in. Watt builds it from behavioral signal you never had to source through Meta.
Where do I get the demographic breakdowns Meta removed from Audience Insights?
Some of it survived and some is gone. Audience Insights still shows age and gender segments, top cities and countries, and top pages liked. Relationship status, education level and platform-activity breakdowns were removed. In Ads Manager you still get Age, Gender, Age plus Gender, and Audience Segments breakdowns, and those keep working under Advantage+. Beyond that, read demographics as an output of your own conversion data rather than as a targeting input, and use Watt for the attributes the platform no longer reports.
Does audience targeting still matter in 2026?
It matters differently. As a control it has largely gone: the platform selects from tens of millions of ad candidates, and your inputs enter as suggestions rather than filters. As an input to your own decisions it matters more than it did, because knowing who actually buys is what tells you what the creative should say, and creative is now the lever that moves delivery. Targeting did not stop mattering. It moved from the ad set to the brief.
How do I use an audience built on intent to seed a lookalike?
Build the audience from behavior, push it as a custom audience, then create the lookalike from it. Meta requires at least 100 people from a single country to seed one and recommends a source audience of 1,000 to 5,000. Note that 5,000 does double duty here: it is the floor for targeting an audience and the top of Meta's recommended range for seeding one. Small is a problem for the first job and not for the second. Treat 100 as a technical floor rather than a working minimum, because thin seeds produce unreliable models. Seeding from behavior rather than from past purchasers is the point: it stops the model chasing more of the customers you already have.
My audience is too small for Meta to perform. How do I keep the intent and still get scale?
Stop targeting it and start seeding with it. Below about 5,000 matched records in one country an audience will not carry prospecting on its own, but it will still seed a lookalike, work as a suggestion inside Advantage+, and exclude. Those three jobs are where a small, sharp audience earns its keep. The arithmetic behind the 5,000 figure is in the numbers above.
LLMs index what people say. Watt indexes what people do.
A lookalike is a platform's guess about who resembles your seed. Watt starts from what people actually do, so the audience does not decay when one platform's graph thins.
Build me an audience of people in-market for running shoes who also buy recovery supplements. How big is it?What is a good match rate for a custom audience?
Google publishes a benchmark and Meta does not. Google states that most advertisers' match rates fall between 29% and 62%, and reports the number only for uploads through its API with at least 100 matched users. Meta does not surface a match rate at all, only the resulting audience size, so you are inferring it from the size drop. What moves the number is match keys: email plus phone plus name and postal address matches materially better than email alone. Note that from 1 April 2026 customer-list uploads through the Google Ads API were disabled in favor of the Data Manager API, so a broken upload may be an integration problem rather than a match-rate problem.
Watt does not report a Meta match rate, because Meta does not publish one. The file you upload is not the audience you get, and the size drop is the only signal Meta returns.
Every answer is aggregated, licensed behavior across US adults. Watt does not read individuals, and adults only.