Build Better Facebook Lookalike Audiences

Your Facebook lookalike audiences are probably built on email addresses and nothing else. You export a CSV from your CRM, upload it to Ads Manager, and let Meta's algorithm figure out who resembles your customers. The problem is that Meta has almost nothing to work with. An email address tells the algorithm that a person exists. It does not tell the algorithm that this person pays you $1,200/year, uses three product features daily, and has never filed a support ticket. That gap between what you know about your customers and what Facebook knows is where ad spend goes to waste.

We covered the theory behind this in our recommendation systems overview. Every prediction system, including Facebook's lookalike model, is bounded by the quality of its inputs. Our lookalike audiences explainer walks through seed list quality in detail. This guide is the practical companion: how to actually build enriched seed lists by syncing customer data from your billing, product, and support tools into your CRM, then pushing those lists to Facebook.

Why most Facebook lookalike audiences underperform

The standard workflow when you create a lookalike audience on Facebook looks like this: open your CRM, filter for "paying customers," export the list, upload to Facebook. The exported file contains names and email addresses. Maybe a lifecycle stage field if someone filled it in manually.

Facebook's algorithm receives this list, matches as many emails as it can against its user base (typically 40-70% match rate), and builds a statistical profile of those matched users. But the profile is based almost entirely on what Meta already knows from its own platform data. Your contribution was a list of email addresses. You've told the algorithm who your customers are, but not what makes them valuable.

Compare that to a seed list that includes lifetime revenue, subscription tier, product feature adoption depth, login frequency, and support ticket count per customer. Now Facebook has six behavioral dimensions to pattern-match on. The algorithm builds a profile that incorporates both its platform data and your business data. The resulting lookalike audience targets people who resemble your best customers across multiple axes, not just people who share demographic traits with anyone who ever gave you an email address.

The reason most teams can only produce the bare-email version is architectural. Revenue data lives in Stripe. Product usage sits in a separate database or analytics tool, and support history is locked inside Zendesk or Intercom. The CRM holds contact information and maybe a deal stage. Nobody has connected these systems, so the CRM never has the complete customer picture that would make a seed list actually useful.

What customer data to sync for better Facebook lookalike audiences

Not every field matters equally. Start with the fields that distinguish your top-20% customers from everyone else.

A seed list built from these synced fields gives Facebook 8-10 data points per customer instead of 1. The algorithm's matching accuracy scales with input dimensionality. More signals per customer means a tighter definition of "similar."

One field worth calling out specifically: lifetime_revenue. Facebook's value-based lookalike audiences let you weight seed list members by a numeric value. If your seed list includes a revenue column, the algorithm gives more weight to patterns shared by your highest-revenue customers. This single field often produces a measurable ROAS improvement over unweighted lookalike audiences.

Step-by-step: build a high-quality seed audience for Facebook lookalike targeting

The entire process takes roughly 45 minutes. Most of that is deciding which fields to sync and what threshold defines "high value" for your business.

1. Sync billing data into your CRM. Connect your billing tool as a source and your CRM as a destination. In Oneprofile, authenticate both tools with API keys, select Customers as the record type, and map billing fields to CRM contact properties. Use email as the matching key. For fields that don't exist yet in your CRM (like lifetime_revenue or subscription_status), Oneprofile creates custom contact properties automatically with the correct field type. Set a 15-minute sync schedule so the data stays current.

2. Sync product usage data into your CRM. Connect your product database as a second source pointing to the same CRM destination. Map features_activated, last_login, and session_count to CRM contact properties. Same matching key (email), same schedule. After both syncs run, each CRM contact has billing data from Stripe and product data from your database layered onto the same record.

3. (Optional) Sync support data. If you want support health in your seed list, add your helpdesk as a third source. Map ticket_count and last_contacted_date. This step is optional but useful for excluding high-churn-risk accounts from your seed.

4. Build a high-value segment in your CRM. Filter contacts by the synced fields. A reasonable starting point:

• subscription_status = active

• lifetime_revenue > your top-20% threshold (check your actual distribution)

• features_activated >= 3

• ticket_count < 5

This segment represents your healthiest, highest-value customers. These are the people you want Facebook to find more of.

5. Export the segment and upload to Facebook. Export the filtered list as a CSV. Include every synced field, not just email. Upload it to Facebook Ads Manager as a Custom Audience. Facebook hashes the data and matches records against its user base. With richer data, the match quality improves because Facebook can cross-reference more identifiers and behavioral signals.

6. Create a lookalike audience on Facebook from the seed. Select your new Custom Audience as the source. Choose 1% similarity for the tightest match. Facebook builds the lookalike from the multi-dimensional profile of your best customers rather than from a flat list of emails.

Beyond Facebook lookalike audiences: suppression and retargeting with synced data

Once billing and product data flow into your CRM, the enriched contacts enable workflows beyond lookalike marketing.

Customer suppression. Export a segment of all active paying customers and upload it as a Facebook exclusion audience. This prevents you from spending ad dollars on people who already converted. The value compounds with sync freshness. If someone upgrades today and your suppression list refreshes tonight, tomorrow's ad spend no longer targets them. Without synced billing data, the suppression list goes stale within days.

Funnel-based retargeting. Build segments by subscription status. Customers on a free plan with high feature adoption are strong upgrade candidates. Export that segment, upload as a Custom Audience, and run retargeting ads with upgrade messaging. This is more precise than retargeting everyone who visited your pricing page, because you know their actual product behavior.

Churn win-back. Filter for subscription_status = canceled in the last 90 days with lifetime_revenue above your median. Upload as a Custom Audience and run win-back ads. The audience is specific enough that you can reference their past usage level in the ad creative, which tends to outperform generic re-engagement campaigns.

All three workflows depend on the same infrastructure: billing and product data synced into the CRM. The lookalike audience gets most of the attention, but suppression and retargeting probably deliver more immediate ROI for teams already running Facebook ads. We've seen teams recover 15-20% of wasted ad spend just from proper suppression, before they even touch lookalike targeting.

Facebook lookalike audience best practices

Start at 1% similarity and expand only if volume is insufficient. A 1% Facebook look alike audience in the US is roughly 2.6 million people. That's more than enough for most budgets. Expanding to 3% or 5% dilutes the match quality. Only go wider if you're spending enough to exhaust the 1% audience.

Refresh your seed list on a schedule. Stale seed lists include churned customers and miss people who recently became high-value. If your CRM syncs every 15 minutes, your seed data is always current. Re-export and re-upload the Custom Audience at least weekly. Some teams automate this with Oneprofile's segment activation, which pushes CRM segments directly to ad platform Custom Audiences without the CSV step.

Use value-based lookalikes when possible. If your seed list includes lifetime_revenue, select "Value-Based Lookalike" in Ads Manager. This tells the algorithm to weight the profile toward your highest-revenue customers rather than treating all seed members equally.

Test overlapping audiences against each other. Run your enriched-seed lookalike against your old email-only lookalike in a split test. Measure CPA and ROAS over 14 days. In our experience, the enriched version outperforms by 30-50% on CPA, but your numbers will vary with audience size and industry.

Don't neglect match rate. If your match rate is below 50%, check data quality. Inconsistent email formatting and outdated addresses reduce matches. Synced data tends to improve match rate because fields come directly from systems of record rather than manual entry.