What it means
A business may know who buys from it but not where to find comparable prospects, so it gives an advertising platform an eligible source group, often called a seed, and the platform models a broader group. The result can support acquisition testing.
Meta describes lookalikes as audiences built from customer or conversion seeds, and its documentation allows different similarity and reach settings for a selected location, which are platform-specific rather than a general mathematical rule. Source quality matters: a list of all email subscribers may include people who never bought, while a list of repeat customers may better reflect the intended goal, although that does not guarantee the latter will be cheaper to reach.
Include enough valid records for the platform to model and obey its minimum requirements, remove bad data and define the event consistently. Mixing purchasers, support enquiries and giveaway entrants can blur the signal.
Privacy comes first, because the business must have a lawful basis to use customer data for this advertising purpose and comply with the platform's policies, and a customer list is not free to upload simply because it exists in a CRM. Sensitive categories may face tighter restrictions, as Meta's developer guidance warns about custom or lookalike audiences that suggest certain health or financial conditions, and housing, employment and financial ads can also carry special platform rules.
The audience is not a copy of named customers but seeks other people whose modelled traits resemble the source group, so do not imply the advertiser sees a person's private characteristics or exact matching logic. Some systems treat a lookalike as a targeting boundary while others use it as a suggestion, and Google Ads says its Demand Gen lookalike segments are moving toward a suggestion mode that can reach beyond earlier similarity thresholds.
Check the live campaign settings before describing the result. Location changes the pool too, since a narrow model in one country is not the same size or composition as a similar percentage in another, so compare campaigns only after noting geography and objective.
Exclude existing customers when acquisition is the goal, where the platform and data permit it, because otherwise a campaign can claim success by selling again to people already in the source or customer base, and retention campaigns require a different test. Test against a sensible alternative by comparing a lookalike campaign with broad or interest-based targeting under similar budgets, creatives and measurement windows, since differences in attribution can masquerade as audience performance.
Watch customer acquisition cost and quality, because a low cost per click is not enough if leads do not buy or return. Avoid overconfident claims, since similarity is a prediction, not proof that someone wants the product, and platform models and delivery can change, so yesterday's result is not a guaranteed benchmark.
Refreshing the seed can change the model, so revisit the source definition and document when it was updated, and respect scale because a tiny number of conversions can make acquisition cost swing sharply. For an owner, a lookalike is a way to test reach beyond known customers, and its value comes from a suitable, permitted seed and measured incremental results, not from the label alone.
In practice
Real-world examples.
Example
A retailer builds a lookalike from repeat purchasers in a chosen market, then compares new-customer cost with a broad campaign. It excludes existing customers from both tests and uses the same budget and creative. The comparison covers repeat purchases as well as first orders.
Example
A clinic discovers its proposed seed reveals sensitive health conditions. It checks platform rules and lawful use rather than uploading the list. The marketing team chooses a different, non-sensitive seed such as newsletter sign-ups who requested general information.
Example
A Demand Gen marketer checks whether a lookalike is a strict audience constraint or a suggestion under current campaign settings. The settings show the platform may reach beyond the seed's similarity range. The report describes the audience as a signal and not as a fixed segment.
Formula
Calculation
Lookalikes are generated by platform models, not a universal public formula. For evaluation, customer acquisition cost = spend on the test / attributable new customers, with clear attribution.
Spending $10,000 for 200 customers implies $10,000 / 200 = $50 per customer, but suppose 40 of the 200 were already customers. The genuinely new customers are 200 - 40 = 160, so the true cost is $10,000 / 160 = $62.50. A broad-targeting test with the same $10,000 that won 150 new customers costs $10,000 / 150, about $66.67 each, so the lookalike looks modestly better, although quality and repeat purchases still need checking.Case study
Seen in the real world.
This entirely fictional case follows Marble Goods, an invented online retailer testing a lookalike of repeat buyers. It used a permitted seed and separated existing customers from the acquisition report. After a controlled test, the team compared new-buyer cost and repeat purchases with broad targeting. The brand and outcomes are invented; a lookalike alone was not assumed to outperform.
In the invented test, each arm received $10,000. The lookalike arm produced 160 new buyers at $62.50 each and the broad arm 150 at about $66.67, a difference small enough that the team repeated the test over a longer window before changing its budget. It recorded the seed date and campaign settings so that the next test could be compared like for like.
Watch out
Common mistakes.
- Uploading customer data without checking permission and platform restrictions.
- Assuming "similar" people are certain to convert.
- Comparing acquisition costs while counting existing customers as new.
Questions
People also ask.
Does a lookalike reveal a person's private profile?
No. The platform models an audience; advertisers should not assume they can see individual matching logic.
Is it always a strict targeting limit?
No. Some current campaign types use the seed as a suggestion, so inspect the platform's settings.
What makes a good seed?
A relevant, lawful and sufficiently sized source group aligned with the campaign goal.
From the founder's library
Take it further with the book.
Build your financial confidence beyond this definition. Shihan's full-length guide, Accounting Fundamentals, takes the same plain-English approach and turns it into a complete, practical playbook for non-finance managers, business owners and students - with chapter-end quiz answers and presentation slides included.
25% off with code MMHQ25, applied at checkout. Priced in USD - checkout may show the equivalent in your local currency.
View the book and save 25%