A customer who buys once during a clearance sale and a customer who returns every month both appear in your purchase data. They may tell you very different things about whom to acquire next.
Facebook Lookalike Audiences let you use those differences. You choose a group of customers or people who have engaged with your business, and Meta identifies others with similar characteristics. The useful work begins before you select an audience percentage: deciding which behavior you want more of.
In 2026, there is another decision to make. When you add a lookalike as an Advantage+ audience suggestion, Meta can deliver beyond that audience. Building a lookalike and restricting delivery to it are different things.
This guide covers the fundamentals, the setup decisions that matter, and how to test whether a lookalike contributes to your acquisition results.
Key Takeaways
- Choose a source audience that reflects your campaign goal. Repeat purchasers, qualified leads, and engaged visitors represent different outcomes.
- A smaller lookalike percentage means greater similarity to the source. It does not guarantee a lower acquisition cost.
- Under Advantage+ Audience, a lookalike can be a suggestion that Meta expands beyond. Review the controls in your actual campaign setup.
- Test a lookalike suggestion against a comparable setup without that suggestion. Keep creative, conversion goals, and measurement consistent.
- Evaluate customer quality alongside acquisition cost. A cheaper conversion is useful only if it contributes to the business.
What Is a Facebook Lookalike Audience?
A Facebook Lookalike Audience is an audience Meta builds from people who share characteristics with a source group you provide. That source might include customers, website visitors, app users, or people who engaged with your business on Meta.
Think of a homeware brand with several customer groups: occasional gift buyers, repeat home decorators, and subscribers to a replenishment product. A lookalike built from the whole database blends those groups. One built from subscribers starts with a more specific commercial question: can we find more people like the customers who keep renewing?
Three choices shape the audience: the source group, the market you want to reach, and the audience size or similarity percentage.
The distinctions between targeting options are useful here:
- Custom Audiences use existing relationships or interactions, such as a customer list or website visits. They can support retargeting, exclusions, and lookalike creation.
- Lookalike Audiences identify people similar to a source group. They are commonly used for prospecting.
- Interest targeting uses available interest and behavior categories.
- Broad targeting gives Meta room to optimize within the selected controls without adding a specific interest or lookalike suggestion.
Advantage+ Audience is a delivery approach that can use audience suggestions, including lookalikes. It should not be treated as a separate customer segment.
How Lookalikes Fit Into Meta Advertising in 2026
The distinction to understand is between an audience you create and the people your campaign is allowed to reach.
With Advantage+ Audience, a lookalike can guide delivery without limiting it. Meta can look beyond the suggested audience for people it predicts are likely to produce the selected outcome.
Review the controls shown in your account, including location, minimum age, and Custom Audience exclusions. Do not assume that a suggested age range or an included audience creates a firm boundary. Available settings can vary by campaign configuration and eligibility.
This changes how to interpret results. If you add a 1% lookalike as a suggestion and the campaign performs well, you have evidence about that campaign configuration. You have not established that every conversion came from inside the original 1% audience.
Where Andromeda fits
Meta's Andromeda retrieval system selects relevant ads from a much larger pool before subsequent ranking stages decide what to show. Meta reports improvements to retrieval recall and ad quality, but those system-level results are not lookalike campaign benchmarks.
For an advertiser, the practical implication is to evaluate audience inputs alongside the ads, offer, and conversion signals. A well-defined seed cannot compensate for an unclear product proposition or a checkout that loses customers.
Choose a Source Audience That Matches the Outcome
Before opening Ads Manager, write down the result you want to repeat. “More customers” is a starting point. “More customers who reorder within our normal buying cycle” gives you a clearer basis for selecting a source.
For an established ecommerce business, useful candidates include:
- Repeat purchasers.
- Customers with higher recorded purchase value.
- Buyers of a relevant product category.
- Subscribers or members who remain active.
- Recent purchasers whose behavior reflects the current offer.
For a lead-generation business, consider whether qualified leads or converted customers would represent your goal more accurately than all form submissions. Build only from data you are permitted to use and that meets Meta's eligibility requirements.
Website visitors, video viewers, and social engagers can also provide starting points. Their behavior shows interest, but it does not establish the same buying intent as a purchase. If that is the data available, make the uncertainty part of the test.
This is where using first-party data strategically becomes practical. The job is to select and maintain a group whose behavior is relevant to the next campaign.
Balance relevance with enough usable data
Meta's lookalike creation guidance generally recommends a source of 1,000 to 5,000 people. Treat that as guidance rather than a reason to pad a focused source with unrelated contacts. Eligibility also depends on matching and the requirements shown for your chosen source.
Avoid dividing a modest customer base into so many segments that none supports a useful test. Start with one commercially meaningful source, then test another if the budget and available data allow it.
Consider value-based lookalikes
Where available, a value-based source lets you include customer value information. This can help represent differences between customers who spend different amounts.
Use a consistent definition of value. Recorded spending over a defined period and predicted lifetime value are different measures. Keep currencies, date ranges, and treatment of refunds consistent so the uploaded values remain interpretable.
How to Create a Facebook Lookalike Audience
The exact interface may vary. Use the following sequence alongside the options available in your account.
1. Prepare the source
Check that the audience reflects the behavior you selected. For a customer list, review duplicates, outdated records, and the identifiers available for matching. For an event-based source, verify that the event represents the intended action.
Write a short definition your team can reuse, such as “customers with a completed, non-refunded purchase during the selected period.” This makes future comparisons easier.
2. Open Audiences in Meta Ads Manager
Find the Audiences area, choose the creation option, and select Lookalike Audience where available. Select an eligible source that your account has permission to use.
3. Choose the market
Select the target geography supported by the setup. Geographic relevance is worth testing: pricing, language, shipping, and buying habits can differ between markets.
When entering a new market, adapt the offer and landing page as well as the audience. Similarity to existing customers does not establish local demand.
4. Select the audience size
Choose an available percentage. A 1% audience contains the closest matches within the selected market; larger percentages broaden the group.
Start with a clear hypothesis. You might compare a narrower audience with a broader option to see whether the additional reach supports your acquisition goal. Neither should be declared the winner before the test.
5. Name and create the audience
Use a name that identifies the source, geography, and percentage, such as “Repeat Buyers | US | 1%.” Keep a record of the source criteria and update date.
Check processing status and eligibility in Ads Manager rather than planning around a guaranteed population time.
6. Review how the campaign will use it
At the ad set level, confirm whether your lookalike is being used as an audience suggestion and whether expansion applies. Check location settings and any customer exclusions needed for the acquisition goal.
Record the actual delivery configuration. An audience name alone is not enough to explain how the campaign ran.
What Lookalike Percentages Tell You
The percentage describes similarity and audience size. It does not tell you what CPA, ROAS, or lead quality to expect.
A narrower audience can be a useful starting hypothesis when the source represents a specific customer type. A broader option may offer more room to find conversions. Actual results depend on the market, offer, creative, budget, and optimization setup.
Use the audience estimate in Ads Manager rather than relying on a fixed US population figure from an article. For tests involving expansion, remember that the suggested lookalike's size does not describe the campaign's entire possible reach.
Also check how percentage ranges are configured. A 0–3% audience and a 0–1% audience overlap by definition. Do not treat them as independent groups simply because they have different names.
How to Test and Optimize Lookalike Audiences
Ask one question at a time
A useful first question is: does adding this lookalike suggestion improve results compared with leaving the suggestion out?
Compare otherwise similar setups. Keep the conversion event, creative, offer, geography, exclusions, bid approach, and attribution settings consistent. Use an experiment setup that separates test groups where available, rather than assuming two ordinary campaigns form a controlled test.
If you change the audience, landing page, and creative together, the result can still inform campaign decisions. It cannot isolate the contribution of the audience change.
Define success before spending
Choose a primary metric that matches the business goal. For ecommerce, that might be acquisition cost or new-customer acquisition cost, supported by order value and margin. For lead generation, it might be cost per qualified lead, supported by the rate at which leads become customers.
Keep the distinctions explicit. Platform-reported ROAS measures attributed revenue relative to ad spend; it does not establish profit or incremental revenue by itself.
Decide the evaluation period and acceptable cost before launching. Allow time for the normal buying cycle and conversion reporting delays. If the test produces too few outcomes to distinguish the options, record it as inconclusive.
Keep source data current
Review whether the source still reflects the customers you want to acquire. A list built around last year's product mix may be less useful after the business changes direction.
Separate source maintenance from audience rebuilding. Customer-list sources need an upload or integration process to reflect changes; event-based sources depend on their configured rules and incoming events. Choose a review schedule that fits how quickly your customer base changes.
Refreshing the source is also different from refreshing creative. If performance declines, inspect the offer, landing page, delivery, and signs of ad fatigue before concluding that the seed is the cause.
Check conversion data quality
If you use both Meta Pixel and Conversions API, make sure the implementation represents customer actions accurately. When the same action is sent through both channels, Meta's event-deduplication guidance explains how matching event information helps prevent duplicate counting.
Ask the implementation owner to verify events, purchase values, currency, and diagnostics. A campaign comparison becomes harder to trust when the underlying conversion data changes halfway through it.
Use exclusions deliberately
For a new-customer acquisition campaign, review the Custom Audience exclusions available in the setup and keep customer records current. An audience built from one customer segment should not be assumed to exclude every existing customer.
Decide separately how to handle recent website visitors. Excluding them can help separate prospecting from retargeting, but the choice should follow your campaign design and measurement plan.
Scale based on observed results
Once a setup supports the acquisition goal, test additional budget or reach in deliberate steps. Record the change and assess its effect after an appropriate period.
There is no useful universal rule that every campaign should expand after the same frequency threshold. Monitor whether additional spend brings acceptable acquisition costs and customer quality, then decide whether to continue, adjust the creative, or test another source.
What Brand Examples Can Teach You
Historical examples are useful for understanding decisions. Their results should not become targets for an unrelated campaign.
Seltzer Goods: build from the data available
Inflow's Seltzer Goods campaign began with interest targeting and added lookalikes based on key page visitors, cart additions, and purchasers as sufficient source data became available.
The agency reported a 785% increase in monthly revenue and 9.68x ROAS for the broader campaign. Originally documented in 2020, it also involved creative, messaging, budget changes, and unusual market conditions. Those results cannot be attributed to lookalikes alone.
The useful lesson is the sequence: establish measurement, use the relevant data available, and add stronger sources as they develop.
Epitaph: examine what the source represents
In Skai's Epitaph case study, an agency used Skai's API-only Conversion Lookalike Audiences, built from selected ad set conversion data. Skai reported 68% more registrations than the Facebook lookalike audience within that campaign.
That is a comparison between a specialized Skai audience approach and a native Facebook lookalike, not evidence that all lookalikes outperform interest or broad targeting.
Its practical relevance is source selection: how an audience is built matters when interpreting its results.
Common Mistakes to Avoid
- Selecting a source for convenience. Use a group that represents the intended outcome, rather than whichever list is easiest to export.
- Assuming similarity guarantees performance. Evaluate the customers and costs the campaign actually produces.
- Ignoring expansion. Document how the selected audience affects delivery in the campaign configuration.
- Testing too many variables. Separate audience tests from creative tests when you need to identify what drove a difference.
- Over-segmenting a limited budget. Give a small number of useful hypotheses enough support to produce an interpretable result.
- Treating every decline as audience fatigue. Check conversion tracking, creative, the offer, and the website before rebuilding audiences.
- Copying historical benchmarks. Use case studies to understand an approach, then establish your own baseline.
How Pixis Supports the Work Around Audience Testing
An audience test creates follow-up work: reviewing performance, investigating changes, allocating spend, and producing the next set of ads.
Pixis Prism brings campaign monitoring, performance analysis, and optimization recommendations into a conversational interface. Teams can use it to investigate campaign results and identify inefficiencies as part of their ongoing review.
Adroom supports the creative side with asset variations, copy generation, and tools for applying brand guidelines. Keep creative consistent during an audience comparison, then use a separate creative test to explore messaging or visual changes.
Together, these capabilities support a repeatable workflow: define the question, run the comparison, review the outcome, and prepare the next action. Pixis's guide to AI-powered user acquisition on Meta and Google explores the wider acquisition workflow.
Frequently Asked Questions
Are Facebook Lookalike Audiences still useful in 2026?
They remain an available approach to testing source-based prospecting. Their usefulness depends on the source and campaign setup. Compare a lookalike suggestion with a comparable broad setup to assess whether it helps your business.
What is the best source for a lookalike?
Start with behavior that represents your desired outcome. Repeat purchasers may be appropriate for a repeat-purchase business; qualified leads may be appropriate for a sales-led business. Test relevance rather than assuming one source wins everywhere.
Should I start with a 1% lookalike?
It is a reasonable test candidate because it prioritizes similarity. It is not a universal best-performing setting. Compare it with another relevant option using consistent measurement.
Does Advantage+ stay within my lookalike?
When the lookalike is an audience suggestion, Meta can expand beyond it. Review the actual controls and expansion behavior in your ad set before interpreting performance.
Can I use lookalikes for retargeting?
Lookalikes are generally used to find similar prospects. For people who have already visited, purchased, or engaged, use an appropriately configured Custom Audience. Check delivery settings when that audience is used as a suggestion rather than a restriction.
Final Thoughts
A useful lookalike strategy starts with a specific customer behavior and ends with an evidence-based decision about acquisition.
Choose the source deliberately. Understand how the campaign will use it. Keep the comparison fair, and judge the customers it brings in alongside the cost of acquiring them.
Once those decisions are clear, use Pixis Prism to support performance review and Adroom to develop the next creative test.