Separate similarity signals from the business objective
In an acquisition plan, lookalikes are one approach to test for new customers. A custom audience describes source people or relevant interactions; a lookalike seeks people resembling patterns in that source. The concepts do not describe an identical group.
Meta’s definition describes using demographic, interest and behavioural source information. This does not mean you know every personal signal in the model or that a new audience member has never encountered the brand. Modelling existing customers and delivering exclusively to new customers require separate controls.
Past buyers of one product may not represent customers for a new offer. Examine category, period and buying reasons. A lookalike does not replace market research, pricing or fulfilment suitability when entering another country. Avoid promising a risk-free way to locate customer “twins” in a new region.
Choose the source through outcomes and usage rights
The event and data plan should explain what the source contains. Completed buyers, suitable leads, page visitors and video viewers represent different actions. A larger source is not automatically a better commercial sample.
For purchasers, inspect cancellations, returns, repeat customers and product mix. For leads, distinguish spam, suitability and actual sales. If using values, verify their meaning and consistency. The largest order amount is not automatically the largest contribution or lasting customer value.
Current creation guidance requires a source of at least 100 people and generally recommends 1,000–5,000. This is not an ideal-results guarantee. Raw rows may differ from matched, usable people. Customer lists also need appropriate usage rights, platform compliance and opt-out removal; avoid segments based on sensitive health or financial criteria.
FROM READING TO A NEXT STEP
Connect the audience test to actual customer evidence
Share the source definition, current controls and new-customer records to define a comparison scope.
Check source permissions and campaign restrictions
The data and campaign owners should verify the source-use decision together. Being a customer does not make every subsequent use appropriate. Hashing is neither a permission document nor proof of anonymity; avoid adding unnecessary fields.
Required account, Page, Pixel or application permissions depend on the source type. Official creation guidance states that saved audiences cannot be sources. If creation fails, first inspect eligibility, usable source size and access instead of attributing the error solely to the percentage selection.
Some advertising for financial products and services, employment and housing has limited or unavailable lookalikes; audiences including young people also have restricted options. Check category, country and the current campaign experience separately. A general lookalike guide does not authorise special-category or sensitive-data targeting.
Read percentages as a similarity and reach trade-off
In the targeting decision, smaller percentages describe closer source similarity and larger percentages broader possible reach. The percentage is not a conversion rate. Assess commercial outcomes instead of declaring 1% inherently highest quality or 10% inherently more expensive.
Verify the available size options in the interface. Nested ranges such as 0–1 and 0–3 are not independent groups. Meta’s audience-set example separates 0–1, 1–3 and 3–6; active expansion means the whole campaign’s delivery need not retain that separation.
| Choice | Meaning | Not established |
|---|---|---|
| Smaller range | Closer source patterns | Certain purchases or profit |
| Larger range | Broader possible reach | An inevitable quality decline |
| Nested ranges | Potentially shared audience members | Independent experimental groups |
| Separate audience-set ranges | Defined range separation | Zero overlap with expanded delivery |
| Value-based source | A defined value signal | Certain lifetime customer profit |
Review the current creation flow for new audiences
When preparing the store campaign, match the visible flow to current official guidance. New lookalikes no longer select location during creation; they use the location settings of the ad set that runs them. The guidance separately says existing lookalikes are unaffected.
Define the source
Choose lookalike creation in Audiences and select or prepare an eligible source with appropriate permissions.
Choose the size
Use available sliders to define size or ranges; name the source, period and scope clearly.
Connect it in the ad set
Inspect location, usable controls and the estimate where the new audience runs; a new estimate may not appear during creation.
Read status and expansion
Check Populating, Ready or Audience not created. Meta says population can take up to 3 days; this is not a result or ad-approval deadline.
Record Advantage+ expansion in the comparison plan
Outcome evaluation should record the actual delivery approach alongside the percentage. Advantage+ lookalike guidance explains that delivery may expand beyond the selected audience when the system identifies performance opportunities. This does not guarantee an actual improvement.
A lookalike included as a suggestion and one with stricter limitations are not identical treatments. Review suggestions, further reach limitations and location expansion in the interface. Do not assume identical controls for every account and objective. Excluding source people from the constructed lookalike does not establish automatic exclusion of every existing customer from wider campaign delivery.
Compare a lookalike with broader targeting or a different source. Keep the offer, creative, outcome event and evaluation period comparable. Equal budgets do not guarantee equal reach or random assignment. Track new-customer costs separately from existing-customer and returned orders; avoid fixed daily 15–20% scaling rules.
Distinguish acquisition messaging from retargeting
New-customer creative should explain the brand and offer. A reminder prepared for someone who already inspected the page may be unsuitable unchanged. Similarity does not tell you what somebody knows about the product or whether they are ready to buy.
Building a retargeting audience from later visits depends on data use and platform conditions. Do not assume every visitor can be retargeted. Consider appropriate purchaser exclusions, current stock and prices, and the follow-up purpose together. A discount is not the answer to every objection.
When results disappoint, inspect source events, product representation, expansion settings and offer/destination fit. Many leads but few sales raise a question about whether the source models lead submitters or suitable customers. Tie refresh periods to the actual business cycle; a fixed 90- or 180-day window does not suit every source.
BEFORE YOU DECIDE
Frequently asked questions
Are lookalikes and custom audiences the same?
No. Custom audiences are based on source people or interactions; lookalikes seek similar people. Campaign expansion and exclusion settings separately affect delivery boundaries.
Is a 1% lookalike always better?
No. Greater source similarity does not guarantee better sales or profit. Compare business outcomes alongside the offer, source representation, budget and usable reach.
How many people should the source contain?
Current Meta guidance requires at least 100 and generally recommends 1,000–5,000. Check technical eligibility and matching; the recommendation is not a quality or results guarantee.
Why can I no longer select a country during creation?
Current guidance says new lookalikes use ad-set locations. The estimate appears where the audience runs. Review existing audiences and the account experience separately.
Will source customers never receive the ad?
Excluding them from the constructed lookalike differs from excluding them from all campaign delivery. Verify expansion and purchaser-exclusion settings in the actual setup.
Is a lookalike sufficient for a new market?
No. Behaviour may differ from the source market. Review price, language, delivery, rules and product demand separately; similarity does not establish market suitability.
LET’S DEFINE THE SCOPE
Connect the audience test to actual customer evidence
Share the source definition, current controls and new-customer records to define a comparison scope.
Discuss the audience test