Advantage+ on Meta: trust the algorithm or stay in control?

Short answer: if your pixel or Conversions API is collecting plenty of conversion data, your creative is strong and varied, and your product appeals to a lot of people, let Advantage+ handle the audience. If you're niche, local, B2B or on a small budget, manual targeting is usually still the better choice. Below I explain why, and how to test it yourself instead of guessing.

What is Advantage+, really?

Advantage+ isn't a feature, it's a label. Meta sticks it on anything where the algorithm takes over a choice you used to make yourself in Ads Manager. The parts that matter most:

• Advantage+ audience: you give Meta suggestions about who you want to reach, and Meta looks beyond them as soon as it thinks that will pay off.
• Advantage+ placements: the ad is shown wherever Meta thinks it will deliver the cheapest results, from Reels to Messenger to the Audience Network.
• Advantage+ creative: Meta crops images, adds music, rewrites text or changes brightness depending on where the ad appears.
• Advantage+ campaign budget: the budget sits at campaign level and Meta splits it between the ad sets.
• Advantage+ sales, leads and app campaigns: the bundled versions, where more or less everything above is switched on from the start.

Why does Meta push it so hard? First, its targeting data got worse when Apple started limiting tracking on iOS in 2021, and broad audiences plus machine learning are how it compensates. Second, automation scales: an advertiser who lets the algorithm drive spends more, with less friction. That doesn't make Advantage+ bad for you. It means you should judge it on your own numbers, not on Meta's recommendation pop-ups.

When does broad targeting on Meta work?

The algorithm is only as smart as the data it gets. Advantage+ audience works when three things are all true.

1. You have conversion data to learn from. Meta's own documentation says an ad set needs roughly 50 optimisation events in a week to exit the learning phase. If you get 8 purchases a week, the algorithm simply has too little to go on, and "broad" becomes an expensive shot in the dark. A half-broken pixel or a missing Conversions API makes it even worse. I cover why conversions go missing in the article on Consent Mode and server-side tracking.

2. The creative is strong and varied. This is the key insight of the whole article, so read it twice: when the audience is automated, your creative becomes your targeting. Meta shows an ad to the people who react to it. Make a video that only a 55-year-old homeowner with a private well understands, and that's who will see it, whatever you chose in the settings. Your script, image and first sentence do the job interest filters used to do. With one generic ad ("We help businesses grow"), the algorithm has nothing to work with, and you end up with whoever is cheapest to reach. There's more on that in my seven rules for ad creative.

3. The product appeals to a lot of people. E-commerce with broad consumer products, apps, subscriptions, courses. Buyers are everywhere in the feed, and Meta's job is to find the ones who are ready right now.

When should you set the audience yourself?

In my accounts there are five situations where I still switch Advantage+ audience off, or at least keep it on a short leash:

• Narrow niches. If you sell spare parts for one particular kind of industrial machine, there might be 3,000 relevant people in your country. Broad targeting will happily spend 90% of the budget outside that group before it finds its way home.
• B2B. Meta knows where people work about as well as it knows which book they're reading, which is to say badly. Filtering by job title and industry is a blunt tool, but it's better than nothing when the buyer is a procurement manager.
• Local businesses with a small area. A dentist in one suburb doesn't need the algorithm to "learn" anything. A radius, an age range and a good offer are enough, and there's no volume to learn from anyway.
• Small budgets. Below roughly €900 a month the learning phase rarely ends before the month does. A warm audience you built yourself (website visitors, video viewers) is often more precise.
• When you want to learn something. Advantage+ is a black box. You find out that it worked (or didn't), but not for whom. To see which customer group actually responds, you have to split the audiences up manually.

How do you use Advantage+ audience without losing control?

It's not either/or. Advantage+ audience has two levels of control, and it's worth knowing the difference.

Controls are hard limits Meta won't go beyond: location, minimum age, language and excluded custom audiences. Suggestions (audience suggestions) are softer: you add interests, age ranges, gender or lookalike audiences (people who resemble your existing customers), and Meta starts there but goes beyond them when it thinks that will pay off.

The practical recipe I use:

• Always add existing customers as an exclusion, via a customer list or an audience of buyers from the last 180 days. Otherwise the algorithm finds them first, since they're the easiest to convert, and you end up paying for sales you would have had anyway.
• Use your best existing audiences (customer list, website visitors, video viewers) as suggestions, not as the only audience. That gives the algorithm a starting point that's yours, not Meta's.
• Set a minimum age as a control if the product has a clear age limit. It's the only age setting Meta fully respects in this mode.
• Give every ad set three to five different ads with different angles, not just different colours. That's what decides who sees them.

How do you run a fair test of broad vs manual targeting?

Most "tests" I see aren't really tests. Someone ran manual targeting for three weeks, switched to Advantage+, saw the cost per purchase drop and drew a conclusion, while the season, the competitors and the creative all changed. That tells you nothing.

Do this instead:

1. Use Meta's built-in A/B testing tool (Experiments) so the two versions don't compete for the same people.
2. The same budget in both. Not 70/30.
3. The same creative, exactly the same ads. You are testing the targeting, not the creative.
4. At least two weeks, preferably three, so both have time to exit the learning phase. Don't stop after four days because one of them looks better.
5. Compare on cost per conversion and actual revenue, not on clicks or CPM. Broad is almost always cheaper per click, and that's completely irrelevant.

The caveat: don't rely on Meta's own report alone

Meta's reporting isn't neutral on this question. The default setting counts a conversion if the person clicked an ad in the last seven days, or merely saw one in the last day. Broad audiences reach more people, so more buyers happen to have seen an ad first, and Advantage+ looks better than it is. Not a conspiracy, just an attribution model that favours reach.

So check two other sources too: Meta traffic in GA4 (with UTM parameters on every ad) and, most importantly, your actual sales: orders, booked meetings, new customers in your CRM. If Meta says 40 purchases and the shop says sales are flat, you have your answer.

Setting up ads you can run yourself, with these checks built in, is what I do with Meta Ads: setup and coaching for a one-off fee of €850, in your own ad account. If you need more creative to work with, that's what ad production is for.

What to do next

1. Check how many conversions your pixel or Conversions API recorded in the last seven days. Fewer than 50 per ad set? Start with manual targeting or a warm audience.
2. Create an exclusion audience of existing customers and add it to every ad set, whichever approach you choose.
3. Count your active ads. Fewer than three different angles? Make more before you let the audience off the leash.
4. Set up an A/B test: same budget, same creative, two weeks. Decide in advance what will count as a win, so you can't move the goalposts afterwards.
5. Add UTM parameters to your ads and compare with GA4 and your actual sales before you decide.

Not sure where you stand overall? The quick analysis takes two minutes (five questions) and shows where your advertising is leaking the most right now.