What actually performs in paid social right now — hook rates, hold rates, CTR, CPM and ROAS, broken down by format and industry. So you know whether your creative is winning or just spending.
Median performance across prospecting campaigns. Hook rate is 3-second views over impressions; hold rate is thruplays over 3-second views. Use these as a bar to clear, not a target to copy.
| Format | Hook rate | Hold rate | CTR | CPM | ROAS |
|---|---|---|---|---|---|
| UGC video | 34% | 38% | 1.42% | $14.20 | 2.10 |
| Talking-head / founder | 22% | 44% | 0.92% | $12.80 | 1.15 |
| Motion graphic | 29% | 26% | 1.18% | $16.40 | 1.36 |
| Static image | — | — | 0.94% | $11.10 | 1.02 |
| Carousel | — | — | 1.05% | $13.60 | 1.30 |
UGC's edge isn't the hook — talking-head actually holds attention longer once it starts. UGC wins because more people stop for it in the first place. If your founder videos convert but underspend, the fix is almost always a stronger opening 3 seconds, not a new script.
A 1.4% CTR is elite for considered purchases and mediocre for impulse-buy categories. Median blended ROAS (Meta-reported) and CTR by vertical.
High-consideration categories (home, furniture) will never post supplement-tier ROAS on last-click — their creative job is assisted conversions and retargeting efficiency. Judge them on blended MER and time-to-purchase, not the platform number, or you'll kill winners.
The single most expensive mistake in the data isn't a bad creative — it's a good one run too long. Across sample accounts, CTR holds roughly steady until frequency crosses ~2.5, then declines gently. Past 3.1, the decline steepens sharply while CPM rises, and CPA drift follows within days.
The teams who caught this early weren't watching a threshold — they were watching a slope. A creative losing 15% of its CTR week-over-week at rising frequency is fatiguing whether or not it's "hit" any single red-line number yet. By the time a static threshold fires, you've usually overspent the tail by a week.
They treated their top performer as a depreciating asset, not a permanent win — briefing the next iteration while the current one was still working, so the refresh was ready before the fatigue bit. And they read patterns at the angle level, not the ad level: knowing "testimonial UGC" was their engine let them replace a fatigued execution without losing the winning idea.
We're compiling the 2026 edition from real, aggregated, anonymized data. Leave your email and we'll send it the day it's published.