Are lookalike audiences dead in the Advantage era?

Lookalikes were the engine of every Meta account for years. Today the algorithm largely does that work itself. This is what still works, what got replaced, and how to test it honestly in your own account.

Lookalike audiences are not dead, but their role as a growth engine is over. What a lookalike used to do, finding people who resemble your buyers, the algorithm now largely does itself through broad targeting and Advantage+, with more data than you could ever fit into a seed. Situations remain where lookalikes keep their value, and they are easy to fence off. The rest of this article: what changed, what still works, and how to test it honestly in your own account.

What made lookalikes so strong in the past?

For years, a lookalike was the smartest shortcut in media buying. You gave Meta a list of buyers as a seed, and the system found the people who statistically resembled them most. That was pure profit in a time when default delivery was still broad and dumb: the lookalike gave the algorithm a head start it did not yet have on its own. Accounts ran entire structures on stacks of lookalikes at increasing percentages, and it worked.

What has changed since then?

Two things, and they reinforce each other. First, the algorithm itself became dramatically better. Delivery systems now predict per auction who is likely to buy, based on behavior far richer than anything a seed list contains. Advantage+ options go a step further and treat your audience setting as a suggestion at best. The head start a lookalike provided is now largely built into the system itself.

Second, the seed signal has thinned. Since iOS14, Meta sees a smaller and more skewed share of your conversions, and a lookalike is never better than the list it is built on. A seed missing half your buyers produces an audience that resembles the wrong half. The instrument weakened at exactly the moment its alternative got stronger.

The lookalike was not beaten by a better trick, but by the system that made its trick the default.

When do lookalikes still work?

There are three situations in which we still seriously consider lookalikes. None of the three is a growth strategy; they are tools for specific moments.

• Smaller markets: where broad lacks the volume to learn quickly, a lookalike can meaningfully narrow the search space.
• New accounts: without conversion history the algorithm has little to hold on to, and a lookalike on a strong customer list provides a starting point.
• Segment tests: if you want to know whether an angle resonates with a specific type of customer, a lookalike on that segment is a readable experiment.

Notice what is not on that list: scaling a proven account in a large market. There we see, across nearly the board, that broad with strong creatives wins, because every manual restriction cuts away buyers the system would have found on its own. The accounts we grow from €15-20K to €150-200K per month lean on creative volume and clean structure, not on audience lists. A lookalike can be a starting point, but it never sets the ceiling of your growth.

How do you test it yourself instead of debating it?

You do not need to settle the lookalike debate on forums; your account can settle it. Set up an honest test: the same creatives, the same budget, one campaign broad and one on your best lookalike. Let it run for weeks rather than days, because daily results are noise. And judge on what lands at the bottom line: cost per new customer and contribution to total revenue, not the ROAS of the best day.

Watch out for overlap while you do. When broad and lookalike run side by side, they partly fish in the same pond and drive up each other's prices. So look not only at the test winner, but also at what consolidation would yield: all of the budget in one broad campaign often outperforms the same budget split across two halves. Fragmentation is the quietest performance killer in today's Meta.

What does this mean for your account structure?

The conclusion for most B2C companies advertising at a serious level: build your structure around broad, let creatives do the targeting work, and use lookalikes only where they solve a specific problem. That makes your account simpler, your signal cleaner and your test results more readable. The energy that used to go into building audiences can go to the only thing that scales structurally: testing more and better creatives.

Conclusion

Lookalikes are a tool from a previous era that still serves in a few situations, but growth today comes from broad, clean structure and creative volume. What works in your situation is not discovered in a debate but in a disciplined test. Designing those tests, reading them correctly and translating them into an account structure that can scale is exactly what we do daily within paid social. Curious what would win in your account? Book a call, we are happy to take a look with you.