Why benchmarks matter
A RoastIQ score without context is just a number. Is 71 good? Compared to what? The benchmark pool answers this: your ad scored in the 68th percentile for DTC Skincare on Instagram Reels.
This is also why we never leave a score permanently unbenchmarked. If a category slice is still thin, the report says so and shows the sample size it is working from, rather than hiding the uncertainty behind a clean-looking percentile.
How we build the pool
Every ad in the benchmark pool comes from publicly available sources:
- Meta Ad Library. The official free API
- TikTok Creative Center and TikTok Ad Library. Public top-ads and running-ads data
- Google Ads Transparency Center. Public transparency data
- Manual curation. Real ads collected and scored through the pipeline by hand
Each ad is processed through the full RoastIQ pipeline: frame extraction, visual analysis, transcript scoring, and the 5-KPI diagnostic. The result is a scored ad carrying the same metrics as your upload, which is the only way a percentile comparison is meaningful.
Current coverage
As of April 2026, the pool covers 500+ ads across:
- Instagram Reels and Feed
- TikTok
- YouTube Shorts and pre-roll
- Facebook Feed
Categories include FMCG, DTC Skincare, DTC Fashion, Tech and SaaS, Finance, and Food and Beverage.
One caveat worth stating plainly: coverage is not uniform. TikTok and YouTube slices carry held-out validation against public outcomes, while Meta Feed scores remain directional defaults, because outcome data for brand ads in the current cohort is sparse. Treat Meta percentiles as orientation rather than as validated ranking.
What comes next
The pool grows continuously as we ingest more ads. Validation work has since expanded well past this milestone, to 1,200+ ads carrying public outcome data.
Every new ad that enters the pool updates the percentile context for all future analyses. Your score gets more meaningful over time, not less.