How custom audiences capture demand

Website sources that work#

Pixel-based audiences capture visitors, viewers, adders-to-cart and purchasers with URL and event rules up to 180 days back. Define by intent depth always: all visitors for reach, product viewers for consideration, cart abandoners for closing.

Verify Pixel fires on every template before trusting the pools: theme changes silently shrink audiences. Meta own custom audiences guide lists the source options. Start at the Meta Ads hub.

Segment pools by recency: 7-day visitors convert differently from 180-day ghosts. Pixel setup keeps the plumbing honest.

Name pools for humans: source, window and intent in the title, so future editors inherit clarity. Mystery pools get deleted in cleanups. Naming discipline saves audiences.

Engagement sources#

Video viewers, page engagers, form openers and shoppers build pools without a Pixel: 365-day windows catch slow burners that site rules miss. Stack engagement with site pools for proper full-funnel warmth.

Prioritise high-intent engagement always: 75 per cent video viewers and form openers out-convert page likers manyfold. Video engagement compounds with every single view.

Refresh creative for warm pools often: familiar faces need new reasons, not repeated reminders. Fatigue curves run faster on small warm pools.

Customer lists that match#

Upload emails and phones for matching against accounts: purchasers, subscribers, trialists, churned users. Match rates reward clean data, so normalise formats and dedupe before upload.

Lawful basis and GDPR documentation are mandatory for UK uploads: consent or legitimate interest assessed, recorded and honoured. Meta own customer-list terms set the rules. Lead routing feeds the lists.

Refresh lists on schedule: weekly for buyers, monthly for subscribers. Stale lists target ghosts. Lookalikes multiply good lists.

Windows that work#

Short windows convert, long windows prospect: 7 to 30 days for closing, 90 to 180 for nurturing, 365 for seeding. Match window to sales cycle, not to defaults.

Layer windows in separate ad sets with distinct creative and bids: one pool, one message, one price. Blended windows blend results into soup. Window splits structure the layers.

Watch pool sizes weekly: shrinking pools signal tracking breakage, not audience apathy. CAPI coverage protects the pools.

Exclude converters from long-window pools where the message assumes prospecting: veterans spotting newbie creative smell desperation. Message must match window warmth. Warm creative differs fundamentally.

Exclusions that protect margin#

Exclude converters from prospecting with customer lists refreshed weekly: prospecting the converted burns goodwill and corrupts CAC. Exclude recent purchasers from acquisition always.

Custom exclusions replaced detailed-targeting exclusions in 2025, so rebuild old structures on the new control. Exclusion builds show the method.

Audit exclusions quarterly: dead lists exclude nobody while looking protective. Ask our team for an exclusion audit.

Sync exclusion refreshes with list refreshes: buyer lists update weekly, so exclusions must too. Stale exclusions protect nobody while looking diligent. Calendar the refresh beside the upload, never separately. Refresh rhythms compound quietly.

Lists as lookalike seeds#

Best customers make best seeds: purchasers beat visitors, repeat buyers beat one-timers, high-value cohorts beat averages. Seed quality decides lookalike quality absolutely.

Keep seeds between 1,000 and 5,000 quality profiles: too few starves the model, too many dilutes it. Lookalike sizing sets the bands.

Refresh seeds as the customer base evolves: last year's buyers mislead this year's model. B2B seeding runs on qualified lists.

Score seeds annually against challengers: new segments keep champions honest. Stale champions decay quietly while reporting stays green. Seed challengers prevent quiet rot.

Frequently asked questions#