Ad Creative That Converts: What We've Learned from 400+ Campaigns

Why creative is the variable that matters most

In mature paid social markets, most advertisers are competing for the same audience segments with broadly similar targeting. The audience is roughly the same. The bid strategy is roughly the same. The landing page is often similar. The creative is the primary differentiator.

Meta's own data consistently shows that creative quality accounts for more than 50% of campaign performance variance. In our experience managing campaigns across sectors, that figure is if anything conservative — we've seen the same audience and the same offer perform two to three times better with improved creative.

What high-performing creative looks like on Meta / Instagram

• Stops the scroll in the first 1.5 seconds — usually through motion, a striking visual, or a direct question
• Leads with the problem, not the solution — 'Tired of chasing invoices?' outperforms 'Introducing InvoiceEasy'
• Uses real faces — particularly for service businesses, authentic-looking people significantly outperform graphics
• Has a clear, specific CTA — 'Book a 20-minute demo' outperforms 'Learn more'
• Looks native to the platform — creative produced for a billboard performs poorly in a feed

What consistently underperforms

• Stock photography — audiences have become adept at pattern-matching and ignoring it
• Long copy that buries the offer — make the value clear in the first two lines
• Horizontal video in vertical placements — still shockingly common, still penalised heavily by the algorithm
• Offers without specificity — '10% off' performs worse than '£150 off your first month, this week only'
• Testimonials without specifics — 'I love this product!' performs worse than 'We reduced invoice chasing from 8 hours to 45 minutes'

"The best-performing ad we ran last quarter looked like it was filmed on someone's phone. The second best was a slide with three bullet points. The worst was a beautifully produced video that cost three times either of them."

Testing methodology that actually works

• Test one variable at a time when possible — creative format, hook, offer, or CTA
• Run tests until statistical significance, not until 48 hours
• Segment test results by audience — what works for warm audiences often doesn't work for cold
• Maintain a creative log — note what you tested, when, for which audience, and what the result was
• Kill underperformers quickly but not too quickly — most platforms need 3–5 days to exit the learning phase

The creative production trap

Many advertisers produce a small number of high-effort pieces and run them until they fatigue. The advertisers with consistently strong performance do the opposite: they produce many low-to-medium effort variants, identify the two or three that work, and then invest production value in polished versions of those proven concepts.

This approach requires a different relationship with creative quality — accepting that most things you produce won't work, and that the goal of most content isn't to be good, it's to generate signal about what good looks like for your specific audience.