Lookalike audiences on Meta: why your source list decides everything

A lookalike audience is exactly as good as the list you hand Meta. Plenty of companies dump their entire email list or every site visitor in as the source and are then surprised the results disappoint. TL;DR: the algorithm looks for people who resemble whoever you put in, so feed it your best customers and you get more people like them. Feed it noise and you scale that noise. In this article you will read how to build lookalike audiences on Meta, why the quality of your seed list decides everything, and how to use this for B2B leadgen without burning budget.

Measure it yourself: calculate your return on ad spend with our ROAS calculator.

What exactly is a lookalike audience?

A lookalike audience (a similar audience) is a group Meta assembles for you based on a source list you supply. You hand over a set of people you consider valuable, your best customers for example, and the algorithm searches the rest of the population for people who resemble them in behaviour and characteristics.

The difference with a regular interest audience is fundamental. With an interest audience, you are the one guessing at age, job title and interests. Say you sell a planning tool and you aim at managers between 30 and 50. That may be right, but perhaps your real buyers are office managers or freelancers you never expected. A lookalike lets Meta find those hidden similarities itself, instead of you guessing blind. That makes it a strong building block within your social media advertising, and one of the tactics we get the most return from for B2B clients in our social ads work.

Why does your source list decide everything?

This is the point that explains most failed campaigns. The lookalike is a mirror of your seed. Give Meta a list of your most loyal, most profitable customers and the algorithm looks for more people like them. Give it a messy list of everyone who ever filled in a form and you get an audience that resembles people who mostly did not buy.

In B2B that is extra dangerous. You often have far more leads than customers, and most leads never turn into a paying relationship. Put all those leads in your source list and Meta learns that “a good prospect” is someone who downloaded a whitepaper once and then vanished. You then scale precisely the wrong profile.

Our rule of thumb is simple: steer on customers and revenue, not on volume. Feed Meta your paying customers, preferably your high-value customers, and not your entire database. A smaller but clean source list produces a sharper lookalike than a large polluted one. You do not need thousands of contacts to start, by the way; a few hundred real customers is a stronger base than ten thousand unqualified addresses.

Which source do you give Meta for B2B leadgen?

Before you create a lookalike, you first need to build a source audience (a custom audience in Meta). For B2B leadgen these are your main options, in order of quality.

Your customer list (CRM export). This is your strongest source. Export your paying customers from your CRM and upload them as the source. Want it even sharper? Take only your high-value customers. Meta can even work with a value per customer, so the lookalike mainly looks for people who resemble your most profitable relationships. After all, you have a far higher chance of selling something to someone like an existing customer than to a random cold prospect.

Site visitors via the pixel. With the Meta pixel on your site you automatically build an audience of visitors that you can use as a source. Do keep it focused: visitors to your pricing or service pages are a better seed than all your blog readers combined. The deeper in the funnel, the cleaner the source.

People who engaged with your content. Think of who watched your videos or interacted with your page. This is broader and therefore weaker than a customer list, but usable if you still have little customer data.

What we advise against is building a lookalike on your Facebook page fans. A like says little about buying intent; people like a page for all sorts of reasons that have nothing to do with your offer. You then get a fairly random lookalike because Meta has too little signal.

When uploading a customer list, watch the match rate, the percentage of contacts Meta can link to an account. For B2B contacts that is usually lower, often somewhere between 10 and 30 percent, because the business email address in your CRM is often not the address someone uses on social. That is normal. So do not count on your full list being reachable, and make sure your source list is large enough to remain usable after the match.

How big should you make your lookalike audience?

When creating it you choose a country or region and a percentage of that population, from 1 to 10 percent. That percentage sets the trade-off between similarity and reach, a mechanism other platforms apply too with their similar segments. According to Meta, a 1 percent lookalike is the group that resembles your source most closely; the higher the percentage, the bigger the audience but the looser the resemblance.

The temptation to go high for more reach is strong, but it usually turns out expensive. AdEspresso tested various percentages and saw a 10 percent lookalike cost roughly 70 percent more than the 1 percent variant (AdEspresso). The reason is logical: at 10 percent there are simply too many people in there who barely resemble your best customers anymore.

For B2B in a small country like Belgium that is extra relevant. Our guideline:

• Start at 1 to 2 percent. That is where the resemblance to your seed is sharpest.
• Keep your country limited. A lookalike on the Belgian population is more usable than a broad international one.
• Narrow further afterwards. You can still combine the lookalike with job title, industry or interest, so you keep a workable segment instead of millions of people. How to find interest audiences on Meta that actually pay off is covered separately.
• Test one variable at a time. Do not start with six different sizes side by side; that only makes your data unreadable.

For the broader framework of bidding, costs and optimisation it helps to have the basics of PPC under control, because a lookalike is ultimately just an audience within a paid campaign.

How do you use this smartly alongside your other campaigns?

A lookalike works best as cold reach at the top of your funnel, not as the closing piece. You use it to reach new people who resemble your customers, and you then catch whoever responds with retargeting. That way you do not keep paying again for the same audience; instead you build a chain: the lookalike sparks interest, retargeting turns it into an enquiry.

Think of the whole movement as a flywheel. You start with your best customers as the seed, the lookalike brings in comparable cold people, retargeting warms them up to an enquiry, and every new customer makes your seed a little sharper.

On Meta, that is rarely your only channel in B2B. Many B2B companies combine Meta with LinkedIn Ads, where LinkedIn allows sharp targeting on job title and Meta delivers cheaper reach and strong lookalikes. How to weigh those two against each other is covered in our piece on Meta ads alongside LinkedIn for B2B.

One thing you must not forget: a lookalike is not a “set and forget”. The audience refreshes itself periodically as your customer list grows, so the more customers you win, the better your seed gets. That makes it a flywheel, provided you keep steering on the right data. And that means: do not look at click-through rates, but at the number of qualified enquiries and the eventual revenue. A lookalike can look perfectly fine in your dashboard and still attract the wrong people if your source list is off. So steer on your return on ad spend and your pipeline, not on the numbers that always happen to look decent.

Frequently asked questions about lookalike audiences

Do lookalike audiences work for B2B, or only for webshops? They work for B2B, but you have to be more selective. The pitfall is using your entire email list as the source. Feed Meta your real customers or your high-value leads, take into account that the match rate on business contacts is lower, and combine it with retargeting and possibly LinkedIn.

How many people do I need at minimum in my source list? Meta needs at least 100 matched contacts from a single country to be able to create a lookalike, but that really is the floor. For a stable result, Meta recommends a source audience of roughly 1,000 to 50,000 people. More important than the number is quality: a few hundred real customers is a better seed than ten thousand unqualified addresses.

Which percentage should I pick? Start at 1 to 2 percent. According to AdEspresso, a 10 percent lookalike costs considerably more and the audience resembles your source too little. In a small country like Belgium you are better off keeping reach sharp rather than broad.

What is a good match rate when uploading my customer list? For B2B contacts, a lower match rate is normal, often somewhere between 10 and 30 percent, because business email addresses are often not tied to a private account. So do not count on your full list and make sure your source stays large enough after the match.

Ready to point Meta at your best customers?

A lookalike audience is one of the most powerful ways to buy new, relevant reach in B2B, but only if you feed the algorithm your best customers and steer on revenue instead of vanity figures. We are a small team that gives honest advice: no webshop tricks, but campaigns that steer on qualified enquiries. Want to know whether this fits your situation? Plan your free intake.