Insights

How interest targeting on Facebook and Instagram actually works

When this post first went up, choosing the right interests was most of the work in a campaign. Since then two things changed: the available interests were cut back hard, and the delivery system got much better at finding buyers on its own. Both matter for how you should set up today.

What an interest actually is

It is not a statement of identity. It is a signal Meta derived from behaviour: pages engaged with, things clicked, apps used, content watched. Someone with "yoga" attached to them may teach it, may have watched one video about it, or may have a partner who does. That is why a precise-looking interest so often delivers an imprecise audience.

This also explains the disappearances. Anything that could be read as sensitive, such as health, beliefs, or political leaning, was removed, and narrow interests get retired when they get too small. If you built a campaign on one interest in 2016, that campaign probably cannot be rebuilt today.

Why broad targeting often beats a careful list now

The system optimises against the event you ask for. Give it a large audience, a clear conversion event and enough signal, and it will find the people who take that action faster than your interest list will. Narrowing the audience first takes options away from the thing that is better at this than you are.

That is not a slogan, it is a testable claim: run your usual interest stack against a deliberately broad audience with the same budget and creative, and look at cost per result after a week. In most accounts I see, broad wins or ties. When it loses, it loses for a reason worth knowing.

When interests still earn their place

Small budgets. With very little spend the system never gets enough signal to learn. A sensible interest layer gives it a head start.

Genuinely niche offers. If your buyer is defined by owning a specific thing or working in a specific trade, an interest can be a real shortcut.

Exclusions. Often more useful than inclusions. Keeping existing customers or an obviously wrong segment out costs nothing and cleans up the numbers.

Reading the result. Even on broad delivery, look at which segments actually converted. That is market research you are paying for anyway.

Where the work moved

It moved to the creative and to the signal. The ad itself now does most of the targeting: the first second decides who keeps watching, and who keeps watching is what the system learns from. Clean conversion tracking decides whether it learns the right thing at all.

Put differently: in 2016 you told the platform who to show the ad to. Today you show the platform who you want by who responds. Which is why five honest variations of an ad are worth more than five hours spent picking interests.