Most accounts have a creative problem disguised as a performance problem. Here's the framework I built to fix that, worked through a roofing lead generation account.
At some point Meta advertising quietly changed, and most accounts didn't adjust.
For years the job was targeting. Find the right audience, build the right interest stack, dial in the lookalike. Creative mattered, but the audience was what you optimized. That's shifted. Meta's move to broader delivery means the algorithm now decides who sees your ads based on what those ads actually signal. Run an ad about storm damage and Meta finds people worried about storm damage. Run one about home value and it surfaces a completely different person. Same campaign settings, different creative, different audience.
So the creative decisions you make before launch are, in practice, audience decisions. Most accounts haven't caught up to that yet.
Here's where it falls apart. When you run similar ads (same angle, same emotional hook, different headline), Meta groups them under the same entity ID. They share delivery, burn through the same audience segment together, and research on ad fatigue points to a roughly 45% drop in conversion likelihood after four exposures to the same ad. Tweaking a headline doesn't reset that. A different concept does.
Every account I've audited that can't scale past a certain point has the same thing going on: lots of ads, very few distinct signals. The framework below is how I solve that before a single ad goes into production.
I call it the creative strategy map. Four layers, worked through in order: Product, Persona, Awareness Level, Pain Point. The output is a matrix of creative directions that don't overlap. Not a list to pick from. A structure that makes running out of angles impossible.
Age, income, homeowner status: that's not a persona. A persona is the specific thing pushing someone toward a decision. The moment that makes them search. Two people with identical demographics can be completely different buyers if the reason they're looking is different. For this roofing account, I mapped three personas that consistently drove volume.
For each persona, I map first-person thoughts across all five awareness levels, from completely unaware through ready to convert. Each cell is a distinct creative direction, each triggering a different entity ID on Meta.
| The anxious procrastinator | The crisis responder | The value-conscious planner | |
|---|---|---|---|
| Unaware | "The shingles look a little rough but it's probably fine." | "That was a bad storm last night." | "We've been thinking about listing the house in the spring." |
| Problem aware | "There's a stain on the ceiling. I keep ignoring it." | "There's water coming in. This is bad." | "The inspector flagged the roof. Buyers will notice." |
| Solution aware | "I know I need a roofer. I just keep putting it off." | "I need someone out here today, not next week." | "A new roof could help us get asking price." |
| Product aware | "I got one quote but haven't called them back." | "I've seen their truck around. Are they legit?" | "I've seen 3 quotes. Who gives me the best value?" |
| Most aware | "Free inspection, no pressure? I'll do that." | "Same-day response, licensed. I'm calling." | "Financing available, warranty transferable. Easy decision." |
Three personas, five awareness levels, one primary angle each: that's 15 creative directions before format is even a consideration. Layer in format variation: UGC video, static, carousel, testimonial. You have more combinations than any campaign exhausts in a quarter.
What matters is that each combination is structurally different. Not a different headline on the same concept. A different emotional entry point, a different assumed starting point, a different person. Meta sees them that way too: separate entity IDs, separate delivery, separate audience segments.
In practice I focus on two personas and their strongest angles rather than testing all 15 at once. Around 12 concepts with three hook variations each gives a serious testing pool without creating a production problem.
Building this map manually used to take the better part of a week: pulling reviews, going through competitor creative, mapping audience psychology by hand. I now run most of the research phase through an AI workflow that cuts that down dramatically. The judgment calls are still mine. The legwork doesn't have to be.
I apply this framework to every account I work with, lead gen or eCommerce.
If your creative is running but not really testing, let's talk.