Level Marketing / Articles / Paid ads
Lookalike audiences: how to find people like your customers
Where can you find people who look like your best customers? Meta's answer is lookalike audiences: you give the algorithm a sample, and it finds new people similar to it. Here is how lookalikes work, which source to build them from and which mistakes to avoid.
How lookalikes work
You provide a source audience — for example, your list of buyers. Meta analyses what these people have in common — interests, behaviour, demographics — and gathers the people in your chosen country who resemble them most. You end up advertising to an audience close to your real customers without guessing interests by hand.
Which source to use
| Source | Quality | Notes |
|---|---|---|
| Buyer list (phones, emails) | Highest | Export it from your CRM |
| Pixel: Purchase or Lead | High | If your site has tracking |
| People who messaged you in Direct | Medium–high | Good for message campaigns |
| People engaged with your Instagram profile | Medium | Large but "warmer" |
| People who watched 75% of a video | Medium | If your content is strong |
The rule is simple: the closer the source is to a sale, the better the lookalike works. The source needs at least 100 people from one country, but results are more stable with hundreds or thousands.
Which percentage to choose
A lookalike covers 1% to 10% of a country's audience. 1% is the most similar but small; 10% is broad but less similar. Meta's audience in Uzbekistan is smaller than in large markets, so 1% is sometimes too narrow. A practical approach: start with 1–3%, then test 3–5% and broader separately.
Step by step
- In Audiences, create a source audience or upload a customer list.
- Choose "Create lookalike audience" and set Uzbekistan as the country.
- Pick the percentage — for example, 1%, 3% and 5% as three separate audiences.
- Test each in its own ad set with the same creative.
- Exclude existing customers — there's no need to advertise to them as if they were new.
Common mistakes
- A weak source. Build from "buyers", not "all leads".
- An outdated list. Refresh the source every 1–2 months.
- Overlapping audiences. Several ad sets compete for the same people and push up costs.
- Ignoring the creative. Even the best audience won't save a weak creative.
Meta now often recommends Advantage+ audience, which gives the algorithm more freedom. In those campaigns you can add a lookalike as a "suggestion" the algorithm uses as a starting point. Only an A/B test will show which approach works better for you.
Conclusion
Lookalikes are a fast way to turn your existing customer base into a flow of new customers. A quality source, testing several percentages and excluding current customers are the three main rules. At Level Marketing we build audiences from the buyers in your CRM to bring cost per lead down.
