RoastIQ is most useful when the score changes a decision. The mistake teams make is treating a single number as the answer instead of reading the score in context.
RoastIQ is the scoring engine inside SaliencyLab, a pre-spend creative intelligence platform that gives marketing and agency teams a scored, benchmarked read on ad creative before media budget is committed. This guide is the reference for the five KPI definitions, the score bands, and the benchmark label. For the reading order, the meeting workflow, and the common mistakes, see how to read a RoastIQ report.
Start with the question, not the number
Before reading any number, agree on the business question the report is settling: is the creative strong enough to scale, is the opening losing attention, is the promise clear, is the brand getting credit? If the team does not agree on the decision first, the score will feel more objective than it really is.
What each KPI is doing
Beat the Skip
This is the first-seconds question. Does the opening earn enough attention to stop a fast scroll or passive view? It carries the heaviest composite weight at 25%.
Get Noticed
This score checks whether the creative establishes presence strongly enough to be mentally registered rather than ignored.
Brand Impact
This is the brand linkage question. The ad can be visually strong and still fail if the brand does not get credited clearly enough.
Sell Proposition
This score is about the commercial ask. Is the product reason, benefit, or offer landing clearly enough to support action?
Build Brand
This is the longer-memory layer. Does the creative add something ownable to brand meaning rather than just filling time?
In the composite, Beat the Skip carries 25%, Get Noticed, Brand Impact, and Sell Proposition carry 20% each, and Build Brand carries 15%.
Read the score bands
Individual KPI scores fall into three bands:
- Strong: 70 and above. The dimension is working
- Moderate: 45 to 69. Usable, but worth reading against the benchmark before you rely on it
- Weak: below 45. Structural problem. Two or more KPIs in this band send the whole creative to a Rebuild verdict
The band matters more than the exact digit. A 71 and a 74 are the same conversation. A 44 and a 46 are not, because the first one counts toward a Rebuild trigger.
How to read the benchmark label
The benchmark label matters because the same raw score can mean different things depending on the format and context. A mid-range score on a short-form platform may be more acceptable than the same score on a longer, information-led format.
Always check the sample size shown alongside the percentile. A percentile drawn from a thin category slice is orientation rather than a verdict, which is why the report shows the count instead of hiding it.
Use the benchmark tool when a stakeholder needs more detail on norm context, and the benchmark interpretation guide for creative teams for how to argue about the bar without guesswork.
What the score can and cannot tell you
A KPI score is a model prediction validated against public engagement and click-intent outcomes, not a measurement. In held-out, out-of-sample validation as of May 2026, predicted scores correlated with TikTok engagement (Spearman ρ +0.31, n=700), TikTok CTR (+0.30, n=691, 5-fold cross-validation), and YouTube view counts (+0.32, n=403, 5-fold cross-validation), across more than 1,200 ads with public outcome data. The scores do not predict sales, ROAS, attributed conversion, or brand recall. Meta Feed scores are directional defaults, because outcome data for brand ads in the current cohort is sparse. Method detail is in the methodology.
This matters in the room. "Sell Proposition scored 62" is a statement about how legibly the commercial reason reads, not a prediction of how many people will buy. A practical habit: treat the score and its sample size as facts, the report's suggested next step as a recommendation, and any account of why buyers might resist as a hypothesis until it is tested.
From score to next step
Two habits keep the interpretation honest. Do not average the KPIs into a homemade number, because the composite is already weighted. And do not skip the qualitative explanation attached to each score: the sentence usually tells you what to fix, while the number only tells you how urgent it is.
When a report reveals a buyer tension the team cannot settle from the scores alone, BuyerLens runs structured synthetic buyer interviews anchored to the scored creative, a useful pressure test before committing to human research. Scoring is only the first filter in a full pre-launch process; see how to test ad creative before launch for the end-to-end sequence.