Interest Targeting

While ads were narrowed only by age and gender, very different people sat inside one condition. Bracket 25 to 34 and the pool comes to around 300,000, containing both people looking for a collaborative document tool and people who will never need one in their lives. That's why cost per click comes out at $0.90.

Attributes of a person don't reveal what that person is looking for right now. Two people the same age respond differently to the same ad depending on what they searched for last week.

Once a customer list exists, you can use a custom audience, built by uploading a list you already hold, or a lookalike audience, where the platform finds people resembling it. The problem is the early period before that list exists. Interest targeting is one of the few things available then.

It runs in five steps.

1. Pick the labels to aim at. Searching the platform's catalog turns up entries like productivity, project management, and remote work, each with a pool size attached. Topics absent from the catalog can't be used, so the first task is restating what your product is in the platform's vocabulary rather than your own.
2. Intersect or union the labels. Requiring two at once shrinks the pool sharply; accepting either one widens it. Intersecting productivity with startups takes 300,000 down to 60,000.
3. Add another layer of criteria. Pair it with demographic targeting, which narrows by age and gender, or with a location condition.
4. Split the ad sets by label. Run one set across three labels and there's no way to learn afterward which label worked.
5. Clean up the labels on performance after a week or two. Judge by click rate and signup conversion, and switch off the weak ones.

Use it with the premise that these labels are estimates you cannot inspect. A productivity label holds people looking for a meeting-notes tool and people who clicked one related article. When the pool comes back at several hundred thousand, you have attached a name rather than narrowed anything.

"You put three interests in one ad set. Split them or we won't know which one works."

The setup makes per-label performance impossible to read.

"The productivity pool is 400,000. That isn't narrowing."

A condition was applied and the audience barely shrank.

"No list yet, so start with interests and move to lookalikes once we pass 2,000 signups."

A plan that states outright this method is for the early stage.

At Pagedeck, Ravi started with six labels split across six ad sets. Two weeks later he switched four off, leaving project management and remote work. Cost per click fell from $0.90 to $0.60, but impressions fell with the smaller pool, so the location condition was widened back out nationally.