Why Facebook ads fail: messaging, creative volume, and attribution

Executive overview

Most Facebook ads fail before the budget is spent — because the message never triggers curiosity. Businesses skip the hard work of understanding their customers and jump straight into ad spend, testing unproven messages against theoretical audiences.

The fix is a three-part foundation: nail your survival sound bite first, use AI-assisted research to generate insights fast, and measure new customer acquisition — not attribution models — as your north star.

The brands that win are the ones that stay closest to their customers, not the ones with the biggest budgets.

The house model: front steps, porch, and door

• Survival sound bites (front steps) — the hook that gets someone to want to know more; e.g. "Lose 30 pounds in 3 months"
• Enlightenment content (front porch) — lead magnets, YouTube videos, white papers, social content
• Commitment collateral (front door) — proposals, pitch decks, sales scripts, webinar CTAs
• Most businesses score D or F on front steps, C- on the porch and door — fixing steps alone moves people onto the porch
• Competitors with inferior products win because their sound bites are stronger
• Ads that skip survival sound bites assume the customer is already almost inside the house

Why creative volume is misunderstood

• Meta's own message: "your creative is your targeting" — the algorithm now does audience targeting; your message does the work of finding the right people
• For every $1M in ad spend, data shows 300–500 creatives are needed; most brands have 3–5
• Of those 300–500 ads, only 20–30 drive 90% of positive performance — a sub-10% hit rate
• High creative volume has traditionally required large teams: copywriters, designers, video editors, media buyers, data analysts
• The real cause of needing 500 ads: brands were testing random messages because they hadn't done customer research — they were ignoring the "nothing part" of marketing
• With proper upfront research, the required creative volume drops significantly

Using AI to compress the research cycle

• Traditional research — scraping competitor sites, reviews, Reddit, YouTube comments, Amazon — took weeks and many people
• With AI deep research and prompt engineering, the same research now takes 15–30 minutes for one person
• Key questions to extract from AI research: What are the most common problems customers describe? What emotional language do they use? When do these problems come up in their lives? What have they tried that didn't work?
• After gathering data, feed AI your results: what ad worked, what didn't, and why — better examples with explanations outperform abstract prompts
• The agency process: audit what worked before → deep research in two days → present insights and proposed campaign head-to-heads to client
• AI levels the playing field: a three-person team can now compete with brands spending millions

On Meta vs Google: who you're really competing against

• On Google, you compete against other brands in your category
• On Meta, your ad competes against Four Seasons, Ford, and Michelin-starred restaurants in the same auction for the same audience
• You don't need to beat your category competitors — you need to be on par with the best advertisers on the platform
• Scrappy e-commerce brands (meal prep, organic goods, massage guns) with tight margins are often the best practitioners because they're forced to find shortcuts

Attribution: stop chasing the wrong metric

• John Wanamaker's "half my budget is wasted, I just don't know which half" has driven companies into increasingly complex attribution stacks — with no improvement in clarity
• Last-click attribution shows the final click before purchase, not what created the customer or demand
• First-touch attribution shows what originally created the customer, but misses all the steps that moved them down the funnel
• Multi-touch attribution assumes you can track customers across the entire internet — you can't, especially as cookies are phased out
• Human purchase decisions are non-linear: customers talk to friends, read reviews, consult colleagues — none of that is trackable
• Use attribution tools for insight, not final decisions
• The practical alternative: focus strategy entirely on new customer acquisition; measure whether CAC relative to average order value is acceptable; test one variable at a time

What's not changing in marketing

• "Today in marketing, everything and nothing is changing" — tools, AI, placements, campaign types all change; understanding customers does not
• Getting close to your customers and staying close is the "nothing part" that never changes — it's also the part most brands skip because it's hard
• Meta's push for fully AI-generated ads will produce copycat, look-alike marketing — it cannot differentiate between shoe brand A, B, and C
• Dove, P&G, and Unilever have survived every marketing transformation by knowing and staying close to their customers
• Use AI to get to customer insights faster — not to replace the human work of understanding them