The essentials
- Objective: reduce the cost per result of prospecting by targeting strangers who resemble your customers, rather than guessed interests.
- A six-step method: source, bands, exclusions, adapted creatives, a 14-day test, trade-off with Advantage+.
- Expected result: a cost per result 20 to 40% lower than interest-based audiences, on accounts that have a good source.
- The trap: a vague source (all visitors, all followers) that produces an unremarkable lookalike.
Lookalike audiences turn your customers into a treasure map: the platform looks for the people who resemble them. Used well, they bring down the cost per result of prospecting; used badly, they do no better than random targeting. The difference lies in the source, the bands, the exclusions and the test. This article gives the operational method; the definition and how they work are in lookalike audiences.
Step 1: prepare a quality source
| Recommended source | How to build it | Target size |
|---|---|---|
| Best customers | CRM export of the highest-value customers over 24 months (email, phone, city), imported as a hashed list | 2,000 to 5,000 contacts for 1,000 to 2,500 matches |
| Recent converters | Custom audience on the Purchase or Lead event from the last 90 to 180 days (pixel + Conversions API) | 500 to 2,000 events |
| Qualified leads | Audience on the filtered Lead event, or a CRM list of leads turned into opportunities | 500 and above |
| Value-based source | Purchase event with value: the platform weights by amount | Same |
A source must be homogeneous (one type of customer), recent (less than 6 to 12 months) and large enough (1,000 matches). If you do not have 1,000 customers, use add to basket events or 75% video views as a fallback source, which is less precise. The data must be collected with marketing consent; see GA4 and GDPR for the framework.
Step 2: create exclusive bands
- Create the 1% lookalike on the source, country France.
- Create the 1 to 3% band (the platform automatically excludes the previous band if you choose “1 to 3%”), then 3 to 5%.
- Create one ad set per band, with the same creative and the same budget, to compare.
- On LinkedIn (predictive audiences) and TikTok (narrow, balanced, broad), the principle is the same with the sizes offered.
Step 3: exclude
- The source itself and all customers (the full CRM list): paying to reach customers in prospecting is a waste.
- Recent converters and the people already in the remarketing audiences, which have their own campaign.
- Employees, and identified competitors where possible.
- Areas outside the perimeter if the business is local.
Step 4: adapt the creatives
A lookalike audience does not know you: prospecting creatives must present the problem, the promise and the proof, not assume familiarity. Three to five variants (benefit, figures as proof, testimonial) in three ratios; see high-performing visuals and videos for ads. The creatives that converted on the remarketing audiences are not necessarily the right ones here.
Step 5: test over 14 days
| Rule | Why |
|---|---|
| Same budget per ad set, or CBO with minimum budgets | A fair comparison |
| 14 days minimum, with no changes | Exit the learning phase, smooth out the week |
| Decide on the cost per final result (qualified lead, purchase), not on CPM or CTR | A broad lookalike has a lower CPM and a higher cost per purchase |
| Also compare with an “interests” ad set and an Advantage+ ad set | Know whether the lookalike brings anything |
| Check frequency | A 1% lookalike wears out quickly: above 3 a week, widen or refresh |
Step 6: weigh it against automatic targeting
On Meta, compare the manual lookalike ad set with the best Advantage+ Audience ad set (with the same source as a suggestion). If Advantage+ does better or as well, let it run and keep the lookalike as a signal; if the lookalike does clearly better (often on small accounts or precise targets), keep it and widen gradually. Review the trade-off every quarter: the algorithms improve and your conversion volume changes. The overall logic of audiences is in audience segmentation in SMA and the account structure in structuring a Meta Ads account.
Maintain
- Dynamic sources (pixel events, connected CRM) rather than fixed lists.
- Refresh the creatives every 3 to 6 weeks.
- Refresh the customer list every quarter.
- Cut the bands whose cost per result exceeds the target for two weeks.
- Read the results in GA4 and the CRM, not only in the platform; see the SMA KPIs.
Expected results
On accounts that have a quality source, 1 to 3% lookalike audiences achieve a cost per result 20 to 40% lower than interest-based audiences, and comparable to Advantage+ on small accounts. Without a good source, the gap disappears. The gain is measured over a full month, on cost per qualified lead or per purchase, set against your margin; see ROAS in SMA.
How GreenRed helps
Rather than juggling several tools, GreenRed's social media tracking brings these metrics together in a single dashboard, compares them over time and tells you which actions come first. You can try it free, with no card, from the Pricing.
Frequently asked questions
How many customers do you need to create an effective lookalike audience?
Around 1,000 matched people in the source, that is 2,000 to 3,000 imported contacts given a match rate of 40 to 70%. Below that, use a fallback source (add to basket, 75% video views) or wait until you have the volume; a lookalike built on 100 people is unreliable.
Which lookalike band gives the best results?
The 1% band has the lowest cost per result but exhausts quickly; 1 to 3% is the usual compromise for a small business; 3 to 5% and above serve volume. Test the bands in parallel with the same budget for 14 days and cut those that exceed your target cost.
Manual lookalike or Advantage+?
Test both: on accounts with a high conversion volume, Advantage+ Audience often does as well or better; on small accounts or precise targets, a manual lookalike on a good source keeps the edge. In every case, provide the source as a suggestion to Advantage+ and review the trade-off every quarter.
Why is my lookalike no better than interests?
Almost always because of the source: too broad (all visitors, all followers), too old, or too small. Rebuild it on your best customers or your recent converters. Also check the exclusions (customers, converters) and the creatives, which must speak to strangers.