Retargeting audience architecture: match the ad to the intent.

A founder asked me this on a call. They were about to run two Google campaigns to the site, one for a high-intent bottom-funnel term and one for broad awareness, and they wanted to retarget each campaign's visitors separately on Meta. It is a smart instinct. It also does not work the way most people assume, and understanding why is the whole of good retargeting architecture.

The short version: the ad platform does not retarget by campaign. It retargets by page. If you want your retargeting to follow intent, you have to build that intent into your pages and URLs first, before you spend a rupee on traffic.

The problem: one page, one blended pool

Here is what most accounts actually run. Every campaign, from every channel, drives traffic to the same one or two pages. A single retargeting audience is built from "everyone who visited the site in the last 30 days." One ad chases all of them.

Inside that pool sit two completely different people. One clicked a "book a demo" ad, read the pricing page twice and left to check with their team. The other clicked a broad "what is X" ad, bounced in nine seconds, and will never buy. They are worth wildly different amounts, and they are getting the identical retargeting ad. The high-intent visitor gets a soft brand nudge when they were ready to close; the tire-kicker gets a hard closing offer they are nowhere near acting on. Both convert worse than they should, and the pool as a whole looks like "retargeting doesn't work for us."

What the platform actually keys the audience on

This is the mechanic the founder's question ran into. A Meta website custom audience is built from three things: your pixel, the events it records, and the URL the visitor landed on. That is it. Meta never receives "this person arrived from Google campaign #4." The referring campaign is invisible to it. When you tell Meta to build an audience of "people who visited specific web pages," you are handing it URLs, not campaign names.

Google Ads works the same way for remarketing. Its lists are built from on-site behavior, the pages and events a visitor triggers, not from which upstream campaign or channel sent them. Membership can last a long time (Google keeps a remarketing list member for up to 540 days), but the thing that defines the list is still the page and the action, not the source.

So the unit of segmentation is the URL. Two campaigns pointing at the same page produce one indistinguishable audience. Two campaigns pointing at two different URLs produce two audiences you can target and message separately. The segmentation you want does not happen at the campaign level. It happens at the page level.

Design the pages so intent segments itself

Once you accept that the URL is the unit, the fix becomes an architecture decision you make up front, not a setting you flip later.

• Give each intent its own URL. A dedicated landing page per campaign is the cleanest version: /demo, /pricing, /guide-download. Each one becomes a clean signal of what that visitor wanted.
• If dedicated pages are too heavy, use URL parameters. A website custom audience rule can match "URL contains," so a consistent utm_campaign or a custom query string in the landing URL lets you split the same page into separate audiences. It is still URL-level segmentation, just done with parameters instead of paths.
• Decide the taxonomy before you launch. The audience only starts collecting from the moment you create it, and it does not backfill. If you launch traffic first and design audiences later, you have thrown away the first weeks of intent data. Set the pages and the rules up before the spend starts.

Build the audience ladder, not a single bucket

With intent living in the URLs, you can build a ladder of audiences that maps to how close someone is to buying, and message each rung differently.

• Top of the ladder: all visitors. Useful for reach and for seeding a lookalike, but the weakest intent. Do not spend your closing budget here.
• Middle: engaged and specific-page visitors. People who hit the pricing page, the case studies, or a product page. This is where retargeting earns its keep.
• Top rung: high-intent actions. Started a demo form, viewed pricing more than once, added to cart. These get the direct, closing message.
• Suppress the people who already converted. Layer an audience exclusion of existing customers and recent purchasers so you stop paying to retarget people who are already in. This one change quietly recovers budget in almost every account we audit.

The best seed for a lookalike is the top rung, your actual converters, not "all visitors." A lookalike only copies the seed you feed it, so a pool of browsers produces a lookalike of browsers.

Match the message to the rung, and watch frequency

Architecture is only half the job. The point of segmenting is to say different things to different intent. Demo-page visitors get the closing ad: the offer, the proof, the booking link. Top-of-funnel visitors get the nurture: the case study, the comparison, the reason to come back. Sending the closing ad to the cold rung wastes it; sending the nurture to the hot rung wastes the moment.

Keep an eye on frequency as your fatigue gauge. A small retargeting pool saturates fast. When the same people are seeing the same ad five, six, seven times, you are burning goodwill, and it is a sign to refresh the creative or widen the rung, not to raise the budget.

The measurement trap that hides all of this

Retargeting audiences are warm, so they convert well, and both platforms are eager to take credit. Meta's default window includes view-through attribution, which will claim a conversion from someone who merely saw the retargeting ad and did not click. On a warm pool that inflates the numbers and makes a lazy one-bucket setup look better than it is.

Judge retargeting the same way you judge everything else: against one source of truth, tied through to pipeline, not on the platform's self-reported total. The full logic is in the paid media attribution guide. Good architecture and honest measurement are the two halves of retargeting that actually pays back.