Most funnels lose the majority of clicks without converting, and retargeting is often set up as an afterthought — one broad audience, one generic creative. Five common mistakes in that setup.
Retargeting almost always gets added to a campaign, but rarely gets actually designed — most often it is one broad "everyone who visited the site" audience and one creative copied straight from cold prospecting. As a result, retargeting budget gets spent almost blindly, even though this segment should, in theory, produce the cheapest conversions in the entire funnel.
One retargeting audience for everyone
Treating every site visitor as one uniform audience wastes budget on people who bounced in a second and on people who reached checkout alike — segment by depth of engagement: viewed a product, added to cart, started checkout.
Showing the same ad they already ignored
Reusing the cold-prospecting creative for retargeting wastes the format — the ad needs to answer a specific objection (price, trust, shipping), not repeat the first-touch message.
No frequency cap on retargeting
Retargeting audiences are small, and ad fatigue hits faster than on the same account's cold audience — cap frequency around 3–4 per week to avoid banner blindness and negative sentiment.
Ignoring the time-decay of intent
Someone who visited 24 hours ago converts differently than someone who visited 21 days ago — split by window (1–3 days, 4–14 days, 15–30 days) with a progressively stronger offer or urgency in the later windows.
Excluding converters too late
Without immediate exclusion of already-converted users from the audience, budget keeps chasing people who already bought or signed up, quietly inflating CPA.
In short
- Segment retargeting by depth of engagement, not as one blanket audience.
- Match the creative to the specific objection at that funnel stage, not to the original cold-prospecting ad.
- Cap frequency deliberately — small audiences fatigue fast.
- Exclude converters immediately, don't wait for a scheduled audience refresh.
The point isn't to copy the whole setup — it's to understand which part of it actually drives the result.