Meta Lookalike Audiences are only as good as the source audience you feed them. A generic customer list gives Meta generic results. A tightly defined audience of your highest-value customers — filtered by persona, LTV, repeat behavior, or signal — gives Meta a precise blueprint to find more people who look exactly like them.
Start with your goal
The right source audience depends on what you're trying to achieve. Before you build anything, answer this question: who is the customer I want more of?
Goal | Recommended source audience |
Acquire a specific buyer type | A single persona that matches your ideal customer |
Maximize revenue | Highest LTV persona or segment |
Improve retention | Persona with highest repeat purchase rate |
Scale efficiently | Largest persona with strong LTV |
Find more VIPs | All ⭐ Signal customers |
There's no single right answer — but being intentional about your goal before building the audience will make your Lookalike significantly more effective.
The audience size question
Meta recommends a source audience of at least 1,000 people to generate a Lookalike, but quality improves significantly with a larger pool. We recommend a minimum of 6,000 customers in your source audience before pushing to Meta.
If your personas are smaller than that, you have two approaches:
Run a backsearch to pull in more history More historical customers means more data for OuterSignal to work with, larger personas, and a stronger source audience for Meta. If you haven't pulled in 12+ months of order history, this is the fastest path to a larger, more accurate audience.
Combine multiple personas into one export If individual personas are too small on their own, combine two or more personas into a single export using OR logic — persona A or persona B — to hit the size threshold Meta needs. This broadens your source audience while keeping it meaningfully defined compared to exporting all customers.
How to choose which persona to use
If you have multiple personas and aren't sure where to start, here's how to think about prioritization:
Highest LTV persona The persona whose customers spend the most over their lifetime. This is usually the right starting point — Meta will find more people who spend like your best customers.
Highest repeat rate persona The persona most likely to come back and buy again. If retention and subscription growth are your goals, this audience will outperform a pure LTV filter.
Largest persona If your top personas are small, the largest persona gives Meta the most signal to work with — even if the LTV isn't the highest. Size and quality both matter.
All ⭐ Signal customers If your goal is to find more high-profile, high-influence customers — rather than just high spenders — your Signal audience is a strong source. These are the customers most likely to have outsized organic reach.
Stacking filters to raise audience quality
Once you've chosen your persona, you can layer additional filters using AND logic to narrow the audience down to the highest-value customers within that group. This gives Meta an even more precise blueprint to work from.
For example:
A specific persona AND LTV over [your threshold] — the highest spenders within your best persona
A specific persona AND 2+ orders — repeat buyers within your best persona
A specific persona AND LTV over [your threshold] AND 2+ orders — your most loyal, highest-value customers within that persona
The tradeoff with AND logic is a smaller audience — so use this approach when your persona is already large enough to comfortably exceed 6,000 customers after filtering. If stacking AND filters brings you below that threshold, either lower the LTV floor, reduce the order count requirement, or go back to a broader persona.
The right LTV threshold will vary by brand depending on your average order value and price point — start with a number that represents your top tier of customers and adjust from there.
Testing multiple audiences
Don't pick just one and stop there. The most effective Meta advertisers run multiple Lookalike Audiences simultaneously — one per persona, or one per goal — and let performance data tell you which is generating the best results at the lowest cost.
Start with your highest-LTV persona, get a campaign running, then build a second Lookalike from your highest repeat-rate persona and run them against each other. OuterSignal makes it easy to export multiple segments, so there's no reason to limit yourself to one.
Getting your audience into Meta
If you're using a single persona If you've set up your Klaviyo stream Playbook, your persona tags are already flowing into Klaviyo automatically — no additional export or Playbook needed. Simply use that persona tag in Klaviyo to build your Meta Custom Audience and create your Lookalike from there.
If you're using a custom stacked segment If you've built a segment using AND logic — for example, a specific persona AND LTV over your threshold — that won't stream to Klaviyo automatically. You have two options:
Klaviyo Playbook — create a Playbook with a Klaviyo Tag action for your custom segment. Once tagged in Klaviyo, use that tag to sync to Meta
CSV export — export your segment directly from OuterSignal in Meta's required format and upload it manually in Meta Ads Manager
Go to Signal Feed → click Filters
Add a rule: Persona → is one of → select your persona(s)
Click the download icon → Export Data → Meta LAL Import
Click Download CSV
Not on Klaviyo? CSV export works as a standalone path regardless of which type of segment you're using.