Creative as Targeting

Creative as Targeting is the operating principle that on modern AI-driven ad platforms, your creative content is the primary signal the algorithm uses to decide which users see your ad. The creative isn't just persuasion — it's the brief you hand the algorithm to go find your audience.

The shift accelerated with Meta's Andromeda ranking system and the broader rollout of Advantage+ campaigns through 2024–2025. Demographic and interest targeting still exist, but their marginal contribution collapsed as the algorithm's ability to infer buyer fit from creative content improved.

The mechanism

Meta's ad ranking reads the creative — visuals, text, audio, on-screen copy — and classifies it against a model of what converts for which users. A founder-story skincare ad surfaces to users who have previously engaged with founder-story wellness content. A price-comparison electronics ad surfaces to users in active shopping mode. The algorithm is doing audience inference you used to do manually.

The practical consequence: two identical products with identical budgets but different creative strategies reach different audiences at different costs. The brand with a single social-proof angle is only accessible to the audience segment that responds to social proof. The brand running four distinct angles — proof, founder story, before-after, unique mechanism — is accessible to four different audience pools. Volume plus angle diversity is the lever.

This is why tight audience targeting often hurts performance now. You're constraining the algorithm's ability to find the users its models would have found anyway, and you're excluding the users the creative itself would have surfaced.

Broad targeting + diverse angles = the new media buying

Under Andromeda-era logic, the playbook inverts from how DTC brands bought media in 2019–2021:

Old model: Tight audiences → single creative that converts against that audience.

New model: Broad targeting (or Advantage+) → multiple distinct angles → algorithm routes each angle to the users it's most likely to convert.

The testing question also changes. You're no longer asking "does this ad work for this audience." You're asking "does this angle premise unlock a segment the algorithm can find at scale." An angle that finds a niche but converts reliably can still deliver strong ROAS at meaningful volume because the algorithm optimizes delivery toward that niche without you defining it.

This is why Creative Angle is the real unit of testing, and why Concept vs Variation sequencing matters: testing variations of one angle limits your audience surface to one segment. Testing five distinct angles potentially surfaces five separate segments.

How to build for it

1. Go broad on targeting or use Advantage+ campaigns — let the algorithm do the audience inference. Reserve tight targeting for retargeting audiences where intent is already established.
2. Increase angle diversity — audit your active ad set. If every live ad shares the same premise, you're competing for one audience segment. Map the angle families you haven't tested: founder story, unique mechanism, aspirational lifestyle.
3. Maintain angle volume — Andromeda needs creative signal to optimize delivery. Too few ads, or too many variations of the same concept, gives the algorithm less to route. The 3-3-3 modular system is one way to generate angle volume systematically.
4. Track angle-level performance — tag every creative with its angle at upload. Most brands tag by format (UGC, static, video) but not by angle premise. You can't optimize what you don't measure.

DTC example

A supplement brand running broad Advantage+ finds that its social-proof angle (celebrity endorsement) scales efficiently against a 35–55 female segment — but it never intentionally targeted that demographic. The algorithm found them. Meanwhile a founder-story angle the brand had deprioritized surfaces a 28–40 male segment that had been invisible in their historical interest targeting. Two angles, two audience pools, neither explicitly defined by the media buyer.

AdRevila's report names the lead angle of any video ad in a single pass — run it across a brand's full active set to surface angle concentration risk before you've already narrowed your audience to one segment.

When it backfires

• Broad targeting without creative diversity — the mechanism only works if you're giving the algorithm genuinely different signals. One angle at broad targeting is just one angle with inflated reach and no segmentation.
• Volume for its own sake — launching 15 slight variations of the same concept doesn't create angle diversity. The algorithm routes them all to the same audience pool and they cannibalise each other's delivery.
• Ignoring downstream conversion data — the algorithm optimizes for whatever event you're bidding on. If purchase signals are thin, it reverts to proxy signals (clicks, initiates checkout) that may not correlate with real ROAS. Angle diversity matters most when there's enough conversion signal to train delivery.

Related concepts

• Creative Angle — the premise being read by the algorithm; more angles = broader audience access
• Concept vs Variation Testing — how to sequence angle tests to generate clean signal for algorithmic learning
• Modular Creative / 3-3-3 — the production system for generating angle volume at scale