Lookalike from high-LTV customers

Quick Answer

Upload a custom audience containing your top 10% of customers by lifetime value, then build a 1% lookalike from it. This is the single highest-quality seed you can give Meta. CPAs are typically 20% to 40% lower than lookalikes built from all purchasers, because Meta is modelling against your most profitable segment instead of every buyer.

Why high-LTV lookalikes are the gold standard

Not all customers are equal. The customer who buys once and never returns has very different signals to the one who has spent £800 across nine orders. If you train Meta on the average buyer, you get average prospects. If you train it on your best buyers, you attract more of them.

The maths is simple. If your top 10% of customers drive 50% of revenue (which is common), then a lookalike of those buyers is 5x more concentrated on the people who actually move your business.

Step-by-step setup

1. Pull customer data from your store or CRM. You need email, total spend, order count, and signup date for the last 12 to 24 months.
2. Sort by lifetime value descending. Cut the list at the top 10%. Some teams cut at top 20% if the absolute number is too small.
3. Set a minimum threshold. Aim for at least 1,000 customers in the seed. If 10% is below 1,000, widen to 15% or 20%.
4. Hash and upload via Customer File. Audiences > Create Audience > Customer List > Upload File. Include first name, last name, email, phone, country.
5. Build a 1% lookalike from the high-LTV custom audience. Name it LAL_1%_TopLTV_v1 and pick the country.
6. Run it as a separate ad set with creative tailored to your ideal customer (use language, offers, and visuals that resonate with high-LTV buyers — usually quality, longevity, premium positioning).
7. Track CPA and Day 30 LTV of the customers it brings in to verify the audience is delivering high-value buyers, not just cheap conversions.

Real examples

Subscription wine club, £55/month. Built a top-LTV lookalike from members with 6+ months tenure. CPA on the top-LTV lookalike was £42 vs £36 on the all-purchaser lookalike. But the top-LTV cohort had 3.2 month average tenure vs 1.8 months — meaning revenue per acquired customer was 60% higher.

B2B software, $129/month. Top 10% of accounts by ARR went into a custom audience. Lookalike CPA was $58, vs $41 for the all-customer lookalike. Worse on paper. But MRR per converted account was $284 vs $147. Net profit was double.

Beauty brand, £35 AOV. Top 20% by total spend (most had 4+ orders). Lookalike delivered 28% lower CPA and 41% higher 90-day repeat rate than the all-purchaser lookalike. A rare double win.

What to use as your value signal

• Total spend over 12 months — easiest, works for most ecom
• Order count — good for repeat purchase categories
• Subscription tenure — best for SaaS and subscription products
• Average order value — good when one-off purchase size matters
• Profit margin — best for businesses with mixed product margins

You can also use Meta's value-based lookalikes (next post in this series) which automate this with purchase value events, but a manual upload of your true top customers gives you tighter control.

FAQs

How many customers do I need for a high-LTV seed?
Minimum 1,000 for stable lookalike performance. Below that, build the lookalike but expect noisier results.

Should I exclude refunds and chargebacks?
Yes. Filter out anyone who has refunded a significant order or charged back. They are not customers you want more of.

Can I use a top-LTV lookalike for cold prospecting?
Yes — that is the best use case. Run it as a top-of-funnel ad set with creative that speaks to your ideal buyer.

How often should I refresh the seed?
Every 60 to 90 days. As new high-LTV customers emerge, you want them in the seed.

What if my product is too new to have LTV data?
Use the next-best signal: customers who placed a second order, or customers who upgraded a plan. Anything that signals commitment beyond a single transaction.

Find more high-value customers, faster

A high-LTV lookalike attracts the right people, but it cannot fix a confusing product page or weak hero image. Pix-Vu tests your landing pages and creative so the prospects you work hard to attract actually convert. Stop sending great audiences to mediocre experiences. Try Pix-Vu free at https://pix-vu.com.