Why it matters
Prospecting needs scale; retargeting exhausts. Lookalikes are the middle layer: find users statistically similar to people who already converted. The platform's similarity model uses whatever the seed contains. If the seed is "anyone who purchased once," the lookalike optimizes for easy converters, not profitable repeat buyers or retained subscribers.
As privacy limits custom audience uploads and match rates fall, seed construction from data warehouse exports (hashed emails, high-LTV deciles, retained subscribers) becomes a strategic asset. Pairing lookalikes with value-based bidding on the same campaigns aligns who you target with what you bid.
Lookalikes are not a substitute for conversion signal quality. Weak Meta CAPI or Google Ads Conversion API value delivery undermines performance even with a perfect seed.
Lookalike audience
pLTV improves seeds and the conversion signals lookalike campaigns learn from:
- First-party data in your data warehouse ranks customers by realized or predicted LTV.
- Export top-decile or payback-positive users as hashed seed lists (policy-compliant size minimums apply per platform).
- Churney sends user-level pLTV on conversion events directly to ad networks so prospecting campaigns optimize value, not only lookalike reach.
- Refresh seeds on a schedule as model drift and feedback loop shift customer mix.
- Measure incrementality and cohort LTV at cohort maturity; retire seeds that scaled low-margin converters.
High-LTV seeds plus value signals beat broad purchaser seeds plus conversion optimization alone.
Category variants
| Model | How lookalike audiences show up |
|---|---|
| Ecommerce / DTC | Seeds from repeat buyers, high margin cohorts, or pLTV top decile; 1–3% lookalike tiers for prospecting tests. |
| Subscription app | Seeds from paid subscribers after trial, not all installers; iOS seeds affected by ATT match limits. |
| SaaS / PLG | Seeds from activated accounts or closed-won CRM lists uploaded via hashed customer match. |
Common mistakes
- Tiny or stale seeds. Platforms need minimum viable seed size and fresh exports.
- Lookalike without value optimization. Reach improves but bids still chase cheap conversions.
- Judging lookalikes on CPA only. Low CPA lookalikes may underperform on LTV at maturity.
Advertiser lens
| Role | What they ask | What good looks like |
|---|---|---|
| Head of Performance / UA | Which seed should we use? | Documented seed definition (LTV decile, repeat buyer, paid sub), size, and refresh cadence. |
| VP Growth / CMO | Are lookalikes scaling profit? | Cohort LTV by lookalike seed version, not just prospecting CPA. |
| Marketing Analytics / Data Science | Did the seed improve incrementality? | Holdout or geo test on seed swap; match rate on uploaded lists tracked. |
| Data Engineering | Can we export compliant hashed lists? | Automated pipeline from data warehouse with PII hashing and minimum count checks. |
| Finance / Procurement | What defines a high-value seed? | Finance-aligned LTV or margin threshold in seed spec. |
FAQ
What is a lookalike audience?
A lookalike audience is a platform-generated prospecting audience designed to resemble users in a seed list you provide, such as customers, purchasers, or high-value segments.
How do you build a high-quality seed list?
Use first-party data from your data warehouse: repeat purchasers, subscribers past trial, or top pLTV deciles. Exclude known low-margin or refunded cohorts when possible.
Do lookalikes replace value-based bidding?
No. Lookalikes shape targeting; value-based bidding shapes how much you pay per user. Best results often combine high-LTV seeds with pLTV value signals on conversions.
What seed size do platforms require?
Minimums vary by network and region (often hundreds to thousands of matched users). Check current Meta, Google, and TikTok documentation before upload.
How often should seeds refresh?
After meaningful shifts in product, offer, or pLTV model; many teams refresh monthly or quarterly. Stale seeds encode outdated customer mix.
How does ATT affect lookalikes for iOS apps?
Lower match rates shrink effective seed size on iOS. Supplement with web purchaser seeds or platform-native engagement seeds where policy allows.
Can Churney export lookalike seeds?
Churney focuses on pLTV modeling and value signal delivery. Seed exports typically come from your data warehouse or CDP using pLTV scores Churney produces.
Not the same as
| Term | Difference |
|---|---|
| Custom audience | Your uploaded or pixel-built list; lookalike is modeled expansion from a seed. |
| Broad targeting | No seed similarity constraint; lookalikes narrow prospecting toward seed-like users. |
| Retargeting | Known visitors or customers; lookalikes are cold prospecting. |
| pLTV | Per-user value score; lookalike is an audience construct built from seeds. |
