Retargeting has a reputation for being the "easy win" of Meta Ads. Show your ad again to people who already visited your site, and conversions should follow. In practice, most retargeting campaigns underperform because they treat every past visitor the same way. Someone who bounced off the homepage in three seconds is not the same buyer as someone who added a product to their cart and left.
Reach Is Not the Same as Intent
A retargeting audience built purely on "all website visitors, last 30 days" is a reach audience wearing a retargeting label. It'll rack up impressions, but the cost per result usually looks a lot like a cold campaign, because most of the people in it were never close to buying.
Building Intent Tiers That Actually Work
Instead of one broad retargeting audience, I break visitors into tiers based on how close they got to converting:
- Tier 1: Add to Cart / Initiate Checkout. The closest to buying. Short, direct messaging, often with urgency or a small nudge like free shipping.
- Tier 2: Viewed Content (product pages). Interested but not ready. Social proof, reviews, and use-case content work better here than hard offers.
- Tier 3: Engaged but never visited the site. Page likers, video viewers, Messenger openers. Needs a reason to visit in the first place, not a checkout nudge.
The people closest to buying and the people who barely noticed you should never see the same ad.
Frequency Is a Budget, Not a Bonus
Retargeting audiences are small, which means frequency climbs fast. Left unchecked, the same warm audience sees an ad a dozen times in a week and starts tuning it out (or worse, gets annoyed by it). I cap frequency per tier and rotate creative on a schedule instead of letting one ad run until it dies.
The Takeaway
If your retargeting results have plateaued, the fix usually isn't a new offer, it's better segmentation. Split your audience by how close they actually got to converting, message each tier differently, and watch your cost per purchase move in a direction a single broad audience never could.
