Loading benchmark context
Pulling the current norm set so the score can be read in the right frame.
What Are Ad Benchmarks and Why Do They Matter?
An ad benchmark is a reference point that turns a raw creative score into a meaningful signal. Without benchmark context, a RoastIQ score of 68 could mean the ad is average for Instagram Reels in DTC skincare or exceptional for YouTube pre-rolls in automotive. Benchmarks anchor the interpretation to the right norm frame so the team makes the correct call: scale, sharpen, or rebuild.
What Benchmarks Are Available
SaliencyLab benchmarks are segmented across four dimensions: category (e.g., DTC skincare, fintech, FMCG), market (e.g., UAE, Morocco, Global), platform (Instagram, TikTok, YouTube, Facebook), and language. Each benchmark segment shows the average score, percentile distribution, and sample size so you know how reliable the norm is. Where SaliencyLab shows percentile context, the support behind each number is labelled as Observed (the exact slice has at least 30 real ads), Modeled (the exact slice is below that observed floor and the percentile is borrowed from an adjacent slice), or Default (no exact or adjacent slice is usable yet — the percentile is shown as an em-dash).
How Benchmarks Are Built
Every ad that runs through the RoastIQ pipeline contributes to the benchmark pool after anonymization. Ads are tagged by category, platform, market, and language during upload. The benchmark engine calculates rolling averages, percentile ranks, and confidence intervals per segment. As the pool grows, more slices clear the observed floor and Modeled estimates settle into Observed ones. SaliencyLab also ingests publicly available ad data from Meta Ad Library, TikTok Creative Center, and Google Ads Transparency Center to seed coverage in underserved categories and regions, including MENA and Africa markets. Public-data seeds appear as Observed (public) when the slice clears the observed floor, and as Modeled when it does not.
Driver Explorer: What Moves Each KPI
Beyond static averages, the driver explorer shows which creative features tend to move each KPI for a given segment. For example, in DTC skincare on Instagram, early product proof may correlate with higher Sell Proposition scores. These are directional snapshots, not causal claims. Use them when the team needs a sharper sense of what may be doing the work, not as definitive creative rules.
Benchmarks vs Predictions
Benchmarks help teams interpret scores. They are not a published validation coefficient. A high percentile rank means the ad scores above most ads in the same segment according to the model — directly, when the slice is Observed, and through an adjacent slice when it is Modeled. Either way, a percentile does not guarantee in-market success. Use benchmarks alongside RoastIQ diagnostic scores and BuyerLens buyer interviews for the most complete pre-spend creative read. See the glossary for definitions of all benchmark terminology.