Lookalike from website visitors

Quick Answer

Building a lookalike from all website visitors is a mistake — too much noise from accidental clicks, bots, and bounces. Instead, build a lookalike from a segmented visitor audience: people who visited a high-intent page (pricing, demo, product), spent more than 60 seconds on site, or visited 3+ pages. These segments give Meta a much stronger signal than raw traffic.

Why raw visitor lookalikes underperform

If you have ever built a lookalike from "all website visitors, last 180 days" you have probably been disappointed. The seed includes everyone who bounced from a Facebook ad in three seconds, every lost click, every scraper, every accidental visit. Meta cannot find a meaningful pattern because there isn't one.

A segmented visitor lookalike is different. By filtering for people who did something on your site, you get a seed that actually represents intent.

Step-by-step setup

1. Confirm your Pixel is firing properly. Use the Meta Pixel Helper Chrome extension. You should see PageView events on every page and custom events on your conversion actions.
2. Create a segmented custom audience. Audiences > Create > Custom Audience > Website. Pick "People who visited specific web pages" and enter the URL of a high-intent page (e.g., /pricing, /demo, /checkout).
3. Add time and frequency filters. Use "Visitors by time spent" and choose Top 25%. This filters out bouncers automatically.
4. Set the time window. 90 to 180 days is the sweet spot. Shorter windows give fresher data but smaller audiences. Longer windows give scale but stale signal.
5. Build a 1% lookalike from this segmented audience. Name it LAL_1%_PricingVisitors_25%_TOP_v1 so you remember the segmentation.
6. Compare to a control. Run a separate ad set with a lookalike of all visitors and let them compete. The segmented version should win on CPA.

Real examples

SaaS reporting tool. Lookalike from "all visitors L180D" → CPA $38. Lookalike from "visited /pricing or /features L180D" → CPA $24. Lookalike from "visited /pricing AND spent >120s on site" → CPA $19. Same creative, same budget. Segmentation alone halved CPA.

DTC apparel brand. Lookalike of all visitors gave a £19 CPA. Lookalike of visitors who viewed 3+ product pages dropped to £13. The brand combined this with a value-based lookalike for top customers and spend efficiency improved 40% across the account.

Online learning platform. Built a lookalike from people who visited the course curriculum page (a deep-funnel page). CPA was 31% lower than the catch-all visitor lookalike, and trial-to-paid conversion was higher because the seed was more committed.

Best segments to use

• High-intent page visitors: /pricing, /demo, /checkout, /contact, /book-a-call
• Top 25% by time on site: Filters out bouncers automatically
• Multi-page visitors: People who hit 3+ pages signal genuine interest
• Returning visitors: Anyone who came back within 30 days
• Specific event firers: AddToCart, ViewContent, Lead, but not yet Purchase

What to avoid

• All website visitors with no filters
• Time windows under 30 days (audience too small)
• Time windows over 365 days (signal too stale)
• Pages with high accidental traffic (homepage from organic search, viral blog posts)

FAQs

How big does my segmented visitor audience need to be?
At least 1,000 people in the segment. Below that, the lookalike will be unstable. If you cannot reach 1,000, widen the time window or relax the filters.

Should I exclude existing customers from the seed?
Yes. Add an exclusion for people who fired a Purchase event. You want lookalikes of high-intent prospects, not your existing customer base.

Can I combine multiple visitor segments into one custom audience?
Yes. Use OR logic — "visited /pricing OR visited /demo OR spent >60s on site". This grows the seed without adding noise.

Does iOS 14 and limited tracking ruin this?
It reduces signal but does not break it. Pixel data is still useful for lookalike modelling, especially when combined with the Conversions API for server-side tracking.

Can I use this for retargeting too?
The custom audience yes, but a lookalike is for prospecting not retargeting. The lookalike contains people who have never visited your site.

Test before you scale

A tightly segmented website visitor lookalike still needs creative that earns the click. Pix-Vu tests your ad creative against real audiences before launch so you do not waste budget on assets that will not perform. Find your winners before Meta finds them for you. Try Pix-Vu free at https://pix-vu.com.