Core Audience targeting: demographics, interests, and behaviors

Core Audiences are Meta's most basic targeting layer — demographics, location, interests, and behaviors, built directly at ad set creation. This part builds a real one for Bright Leaf Coffee's "Coffee Enthusiasts" ad set from part 4.

The building blocks

Location: United States
Age: 25–54
Interests: Specialty coffee, Coffee roasting, Home brewing, Third wave coffee

Interests are based on what people engage with on the platform — pages liked, posts interacted with, content genuinely consumed — not a search someone actively typed. This is the fundamental difference from Google's keyword targeting worth internalizing fully: interest targeting reaches people Meta's system associates with a topic based on real behavior, but with meaningfully less certainty about active, present-moment intent than someone who just typed "coffee subscription" into a search bar.

Why interest targeting is broader, and less precise, than a keyword

Google keyword "coffee subscription":  someone actively searching, right now, for this
Meta interest "Specialty coffee":       someone who has, at some point, shown interest
                                          in this general topic

Someone matching Bright Leaf Coffee's "Specialty coffee" interest might be an enthusiast who already has three subscriptions, or might have engaged with one coffee-related post eighteen months ago and never again. This uncertainty is real and unavoidable — it's exactly why part 1's delivery-prediction system matters so much on Meta specifically, and why creative and offer quality (parts 9 and 10) carry more of the actual persuasion burden here than they do in a search context where intent is already established.

Detailed targeting expansion: a real tradeoff

Meta offers to automatically expand targeting slightly beyond the specified interests if it predicts better results — a genuine tradeoff between precision and delivery efficiency. Leaving it on lets the algorithm find adjacent, similar audiences it judges likely to perform well; turning it off keeps targeting strictly confined to what was explicitly specified. For a new campaign with no data yet, leaving it on gives the system more room to find what actually works — worth narrowing later once real performance data suggests where to focus.

Audience size: too narrow is a real problem, not just theoretical

Too narrow:  a stacked combination of five specific interests, narrowed further by
             a small age range — estimated reach: 4,000 people
Reasonable:  broader, related interests, wider age range — estimated reach: 800,000 people

Meta's ad set creation shows an estimated audience size directly, and a too-narrow audience causes a real, specific problem: too little inventory for the delivery system's machine learning to actually learn from, which routinely produces both a worse "learning phase" outcome (part 13) and a higher effective cost — the opposite of the intuition that narrower targeting should be cheaper and more efficient.

Excluding an audience deliberately

Exclude: People who already purchased in the last 90 days

Excluding existing customers from a cold-audience acquisition ad set is a genuine, common best practice — showing an acquisition-focused ad to someone who already subscribed wastes budget on someone who was never the intended target for that specific ad set to begin with, and it's a different job entirely from the retargeting ad sets covered in part 6.

Next: Custom Audiences — targeting people based on their own real, direct relationship with the business already, rather than a general interest category.