Loading tutorials…
Loading tutorials…
Lookalike audiences are easy to create and hard to make work. The difference between a 1% LAL that scales and one that wastes budget is almost entirely in source audience hygiene — a step most owners skip.
Who this is forMeta Ads operators with a pixel that has been firing Purchase or Lead events for 60+ days. If your source data is younger or thinner than that, your Lookalikes will underperform regardless of how you build them.
What you'll need
Step 1
Lookalikes are only as good as the source. Choose the highest-quality audience available — Purchasers > High-value site visitors > AddToCart > PageView.
Best to worst sources, in order: (1) Top 10% LTV customers (Custom Audience from CRM), (2) All Purchasers from pixel (last 180 days), (3) High-intent site behaviors (e.g., users who viewed >5 product pages), (4) AddToCart users, (5) Newsletter subscribers, (6) Engaged Page followers.
Avoid: Lookalikes of "Page Likes" or "Video Viewers" — these are top-of-funnel signals and produce LALs that look like "people who like Facebook content," not "people who buy your product."
Minimum source size: 100 users. Recommended: 1,000+. Meta technically allows 100 but quality degrades sharply below 500.
Critical: if you can build a Custom Audience from your CRM filtered to "top 25% of customers by LTV," that source produces 2-3x better Lookalikes than a raw "all Purchasers" pixel audience. The data hygiene matters more than the model.
Step 2
Ads Manager → Audiences → Create Audience → Lookalike Audience. Select your source and configure.
Meta Ads Manager → top-left menu → Audiences (or Business Settings → Asset Library → Audiences).
Click 'Create Audience' → 'Lookalike Audience.'
Select source: dropdown shows all available Custom Audiences (CRM imports, pixel audiences, engagement audiences). Pick your best-quality source per Step 1.
Choose location: typically your country (e.g., US). For multi-country campaigns, build separate LALs per country — see Step 5.
Audience size: this is the critical knob. 1% = top 1% of similar users in the country (~2M users in US). 5% = top 5% (~10M). 10% = top 10% (~20M). More on the tradeoff next.
Step 3
1% LAL = highest similarity, smallest. 10% = lowest similarity, largest. Start at 1% and widen only when you've saturated.
1% LAL: ~2M users in US. Tightest similarity. Best initial performance. Smallest scale ceiling — campaigns saturate within $5-10K/mo on the audience.
2-3% LAL: still strong similarity. Larger ceiling. Good for accounts scaling past $15K/mo.
5% LAL: wider, lower similarity per user but more scale. Useful for prospecting at high spend levels.
10% LAL: widest. Performance per user is lower but scale is much larger. Often used as a complement to tighter LALs in campaign budget optimization.
Strategy: launch 1% and 3% in parallel as separate ad sets. Let Meta CBO/Advantage+ allocate budget. When 1% saturates, widen to 5%.
Avoid creating only 10%+ LALs early — the wider audience dilutes signal and slows learning phase exit.
Step 4
After configuring source, location, and size, click "Create Audience." Meta calculates the LAL (takes 6-24 hours for first build).
After all parameters set, click "Create Audience" at the bottom right.
Meta begins calculating the Lookalike. Status shows as "Populating" — this typically takes 6-24 hours for first builds, faster for refreshes.
Do NOT use the Lookalike in campaigns until status reads 'Ready.' Using a Populating audience means Meta is still calculating who's in it — ad delivery is unstable.
Once Ready, the audience appears in Audiences list with size estimate. Verify the size matches expectations (1% in US ≈ 2M users; if you see 50K, the source audience is too small or too narrow).
Step 5
Lookalikes are country-specific. For multi-country campaigns, build separate LALs per country — never one global LAL.
Meta calculates Lookalikes based on country-specific behavioral patterns. A 1% LAL of 'US Purchasers' targeted at 'US + UK + AU' actually expands to 1% in each country separately, not 1% globally — and source patterns may not translate.
Best practice: if you sell in multiple countries, build a separate Lookalike per country (e.g., 'US 1% LAL Purchasers,' 'UK 1% LAL Purchasers,' 'AU 1% LAL Purchasers').
Each requires sufficient source data IN THAT COUNTRY. If your UK Purchaser count is 50 and your US is 5,000, the UK LAL will be unreliable.
For new markets where you have no source: use your highest-performing country's LAL as a starting point, but expect performance to degrade by 30-50% compared to the source country.
Step 6
Don't mix the new LAL with existing audiences in one ad set. Create a separate ad set with only the LAL audience, run for 14 days, evaluate.
Campaigns → your campaign → ad set level → Duplicate an existing ad set → change the audience to your new LAL only.
Keep all other variables constant (placements, creatives, bid strategy) to isolate the LAL's contribution.
Run for 14 days minimum. The first 7 days are typically learning phase — performance is volatile.
Evaluate CPA against your existing best ad set. The new LAL should perform within 20-30% of your top performer to be worth scaling.
If LAL CPA is 50%+ higher than your top ad set after 14 days, the source audience is likely the issue (refresh data) or the % similarity is wrong (try a tighter 1% or wider 5%).
Step 7
Lookalikes degrade as the source audience grows stale. Refresh by rebuilding from updated source data on a monthly cadence.
Meta does NOT auto-refresh Lookalikes from pixel-based sources — the source is a snapshot at LAL creation time.
For pixel-based source (e.g., Purchasers last 180 days): the underlying source audience auto-updates daily, but the LAL itself uses the snapshot from creation.
Best practice: every 30-60 days, create a NEW Lookalike from the (updated) source. Compare CPA against the old LAL. Migrate ad sets to the fresher one when performance improves.
For CRM-based source: re-upload the CRM list monthly with fresh data (new customers, removed churned customers, updated LTV tiers). Rebuild the LAL from the refreshed list.
Naming convention: include the date in the LAL name (e.g., "US 1% LAL Purchasers 2026-05") so you can track refresh history at a glance.
Common mistakes
Building LALs from too-small source audiences
What goes wrong: Lookalike of a 50-user source is essentially noise — Meta's model has too little signal to extrapolate. The resulting LAL behaves like a random audience. CPA is 100-200% higher than your existing best ad sets.
How to avoid: Wait until your source has 500+ users (1,000+ ideal) before building a Lookalike. In the meantime, run broad targeting or Advantage+ Audiences to accumulate signal.
Never refreshing the Lookalike
What goes wrong: You build a 1% LAL of Purchasers in January, set it and forget it. By August, the source data is 7 months old and patterns have drifted. CPA has crept up 30-50% but you don't realize the LAL is the cause.
How to avoid: Calendar reminder: rebuild every LAL on the 1st of each month. Compare new vs old in a 14-day test. Promote winner.
Stacking 5+ Lookalikes in one ad set
What goes wrong: You add 1%, 3%, 5%, 7%, 10% LALs to one ad set 'to be safe.' Meta combines them into one massive audience and the wider tier (10%) dominates spend because it has more inventory. Effective % similarity is closer to 10% than 1%, and CPA reflects that.
How to avoid: Run 1-2 LALs per ad set max. Use campaign budget optimization (CBO) or Advantage+ to allocate between ad sets if you want to test multiple similarity tiers.
Using engagement-based source audiences for purchase intent
What goes wrong: You build a LAL of 'Video Viewers' or 'Page Likers' and target it for Purchase optimization. The audience is biased toward content engagers, not buyers. CPA is 2-3x your purchaser-based LAL.
How to avoid: Match source intent to campaign objective. Purchase campaigns = Purchaser LAL. Lead campaigns = Lead LAL. Engagement source audiences are fine for top-of-funnel awareness only.
Building LALs for tiny markets
What goes wrong: A 1% LAL in a country with 5M Facebook users is ~50K people — too small for Meta to optimize against. Learning phase never completes. Performance is erratic for the campaign's life.
How to avoid: For small markets (under 20M FB users), use 3-5% LALs minimum. For very small markets (under 5M), skip LALs entirely and use interest + behavioral targeting.
Not naming Lookalikes consistently
What goes wrong: Three months in, you have 'LAL 1%,' 'New LAL,' 'Best LAL 2,' and 'Purchaser Lookalike Final' in your audience list. You can't tell which is current, which is stale, or which is performing.
How to avoid: Naming convention: [Country] [%] LAL [Source] [Date]. Example: "US 1% LAL Purchasers 2026-05." Use this for every LAL.
Recap
Done — what's next
How to lower Meta Ads CPM without breaking your performance
Read the next tutorial
Hand it off
Lookalike strategy is one of those areas where the marginal lift from expertise compounds fast. A specialist on EverestX will build and rotate LALs systematically, A/B test similarity tiers, and refresh source data on cadence. Most ongoing engagements include LAL maintenance as part of campaign management — typical cost: $400-1,200/mo at $14-16/hr depending on account size.
See ongoing management rates
Meta technically allows 100 but performance degrades sharply below 500. Aim for 1,000+ events in the source window. If you're below 500 conversions, broader targeting (Advantage+ Audiences or interest-based) typically outperforms a thin LAL.
Increasingly both, in different ad sets. Advantage+ Audiences let Meta autonomously expand beyond your defined audience. Lookalikes are a tightly-controlled audience. Best practice in 2026: run both in parallel under campaign budget optimization, let Meta allocate. Advantage+ tends to win at higher spend; LALs at smaller scale.
Three usual causes: (1) source audience is too small or too narrow; (2) you set the country wrong (e.g., US but the source is mostly UK users — Meta filters out non-matching countries); (3) the LAL is still Populating. Open Audiences → click the LAL → check status and source details.
Yes — upload the email list as a Custom Audience first (Audiences → Create Audience → Custom Audience → Customer List → upload CSV). Then create a Lookalike from that Custom Audience. Hashed emails match to ~50-70% of users on Meta. Higher LTV-tier lists make better LALs.
Pixel-based LALs degrade noticeably after 60-90 days. CRM-based LALs degrade based on how often you re-upload the source (so refresh source + rebuild LAL monthly for best results). Set a calendar reminder — Meta won't auto-refresh.
Custom Audience = the actual users (your top 1% customers by LTV from CRM). Lookalike = Meta's model of who looks LIKE that group. You retarget the Custom Audience for repeat purchases. You prospect to the Lookalike for new customers. Different tools, different jobs.
Meta Ads
CPM creeping past $30 with no clear cause? The lever isn't usually a budget cut — it's structural: campaign objective mismatch, narrow audiences, stale creatives, or wrong placement mix. Here's the systematic approach.
Meta Ads
The browser pixel alone loses 20-30% of conversions on iOS. CAPI sends events from your server to Meta, recovering most of it. Three install paths: Shopify-native, server-side GTM, and direct API. We cover all three.
Meta Ads
Meta's response to iOS 14.5 — Aggregated Event Measurement — caps you at 8 prioritized events per domain and forces you to rank them. Get the order wrong and your most important conversion (Purchase) silently drops out of iOS reporting.
Meta Ads
DIY Meta Ads is the right call until it isn't. This is the honest framework: when the cost of self-managing exceeds the cost of hiring, and how to tell which side you're on right now.