Targeting

Targeting is the most misunderstood part of paid media. Intuition says that narrowing the audience improves accuracy. In practice the opposite holds: narrowing shrinks the model's search space, raises cost, and slows learning.

Why broad targeting wins

Interest layers are a very coarse summary of what the platform knows about a user. When you say "ages 30–45, interested in fitness", you constrain a model that could already work from far finer signals. Narrow audiences also mean:

• Higher CPM, because auction competition is denser
• Frequency climbing fast, so creative fatigues sooner
• Lower weekly conversion volume, so the learning phase never ends

The practical approach: leave targeting as broad as possible (country, language and legally required restrictions aside) and let the creative do the sorting. If you want to reach a specific segment, produce an ad that depicts that segment rather than narrowing the audience box.

Audience types and what they really do

Type	Source	Real function
Broad	Platform model	The main source of new-customer volume
Lookalike	Your customer list	A bridge when broad is not performing
Retargeting	Site visitors, cart abandoners	Bringing back the undecided
Customer list	CRM data	Cross-sell, retention, exclusion
Exclusion	Existing customers	Protecting the acquisition budget

The exclusion list is the most-skipped and cheapest win. If you do not exclude customers who purchased in the last 60–90 days from your acquisition campaign, your new-customer cost will look better than it is — while real growth slows.

Sizing retargeting correctly

Retargeting always shows the highest ROAS in the dashboard, because it advertises to people already close to buying. That is not a measure of success; it is selection bias.

In a healthy account, retargeting should not exceed 10–20% of total budget. A growing share is usually a sign that the top of the funnel has dried up: with no new visitors arriving, you are showing more ads to the same people.

How to check: pause retargeting for two weeks. If total revenue does not drop meaningfully, move the budget to acquisition.

First-party data

As third-party cookies narrow, the data you own has become an asset that converts directly into media performance.

Data	How it is collected	Where it helps
Email and phone	Checkout, accounts, newsletter	Match quality, lookalikes
Purchase history	Order system	Value-based segments, cross-sell
Site behavior	Server-side events	Optimization signal
Survey response	"How did you hear about us?"	Cross-checking channel contribution

A one-question post-purchase survey is the cheapest measurement tool you can set up. It does not replace attribution, but it exposes the systematic errors in platform reporting.

Audience–creative fit

Showing the same ad at every stage of the funnel is a common loss. The message has to match what the user already knows:

Stage	User's state	Message that works
Unaware	Has not named the problem	A scene or story showing the problem
Problem-aware	Looking for a solution	Positioning the solution category
Solution-aware	Comparing alternatives	Comparison, proof, demo
Product-aware	Knows you, still hesitant	Reviews, guarantee, offer

The first two rows work in broad acquisition; the last two work in retargeting. Brand Brain builds this mapping from your brand's own audience data.