Hub · 6 guides + case gallery · updated 18 May 2026

Creative analysis,
written down.

Six guides. One methodology. Built from 2,000+ scored ads on TikTok, Meta and YouTube. Read them in order to learn the system, or jump to the chapter you need before your next edit.

2,047 ads analyzed 1,200+ with public outcomes ρ +0.31 held-out OOS 6.5× quintile lift

Open the field guide → Score an ad first

Oussama Nakhil, Founder

Multiple years buyer-side: NielsenIQ insights, then L'Oréal Groupe in global marketing insights, where I managed Kantar as a vendor. Built SaliencyLab to replace the slow, expensive consumer research loop with something pre-spend, public-data, and honest about its ceiling.

KPIs frozen · never renamed

90s

to a verdict, per ad

platforms validated (TikTok · YouTube · Meta-directional)

// Read this first

"Most creative advice is rumour. This field guide is the opposite, every rule is anchored to a number, every number is anchored to an ad in a benchmark we keep growing. Read it once and you will leave with a methodology you can defend in a room of skeptical buyers."

The field guide

Six guides. One methodology.

Featured · Case gallery

Twelve real ads, twelve verdicts.

Twelve real ads across nine global brands, Nike, Pepsi, Coca-Cola, BMW, L'Oréal Paris, Netflix, CeraVe, Nivea, Booking.com, pulled from public transparency tools and scored through the same RoastIQ pipeline. SCALE, SHARPEN, REBUILD on the record.

12 ads 9 brands 3 platforms 3 verdict bands

Open the gallery →

01 / 06

Beat the Skip

First two seconds: a frame-by-frame teardown

Twenty openings, ten that won and ten that lost, scored against the same rubric. The patterns to copy and the anti-patterns to avoid before you finish the edit.

14 min · CraftRead →

02 / 06

Hook timing

Hook rate vs thumb-stop: what the platforms actually measure

TikTok's "watched 6s" and Meta's "thumb-stop" are not the same number. A clean breakdown of which metric should drive your edit decisions, and which one to ignore.

8 min · Beat the SkipRead →

03 / 06

Brand impact

Brand cue placement: the 1.5-second rule

The single best predictor of Brand Impact in our 2,000+ sample is not how often the logo appears, it is when the first distinctive asset hits the frame. The curve, the audit, and the category exceptions.

6 min · Brand ImpactRead →

04 / 06

Scoring

Reading a RoastIQ verdict end to end

A worked example. An ad scored 62, three reviewers disagreed, and here is how the decision trace settled it. Composite math, KPI weights, confidence labels and the verdict ladder, walked one step at a time.

12 min · ScoringRead →

05 / 06

Methodology

Synthetic panels vs live audience tests: a decision tree

When 36 simulated buyers tell you what you need, and when you should still pay for a real audience. The decision tree, the Kantar comparison, and the honest limits of LLM-based personas.

10 min · ValidationRead →

06 / 06

Methodology

What "benchmarked" really means

A score with no peer is just a number. How big a benchmark pool needs to be by use case, where our 2,000+ ads come from (all official public APIs), and how we defend the methodology in a room full of buyers.

9 min · ValidationRead →

// The methodology, in one page

If you only read this section.

SaliencyLab decomposes every ad into the same five KPIs, weights them into one composite, and lands the result on a three-step decision ladder. The names are frozen. The weights are calibrated against held-out outcome data. Nothing here is opinion.

The five KPIs (frozen).

Composite is a weighted sum, not an average. Beat the Skip carries the most weight because losing the first 2 seconds nullifies everything downstream.

25%

Beat the Skip

20%

Get Noticed

20%

Brand Impact

20%

Sell Proposition

15%

Build Brand

The verdict ladder (no fourth).

Three verdicts force a decision. Five would let you defer one. We resisted the temptation to add nuance bands.

≥70

SCALEComposite ≥ 70 and no KPI < 55. Buy media.

55-69

SHARPENMid-band. Fix the weak KPI; do not start over.

<55

REBUILDComposite < 55 or two+ KPIs < 45. Brief is wrong.

// Honesty pact

What we can claim. What we will not.

Half of being trustworthy is being clear about your ceiling. The full version of this is in "What benchmarked really means". The short version is here.

// What SaliencyLab claims

Predicts engagement (likes, shares, comments) and click intent (CTR percentile) on TikTok, with separate calibrated scores.
Predicts view counts on YouTube, Shorts and In-Stream.
Held-out OOS Spearman ρ between +0.30 and +0.32 across all three validated platforms (2026-05-05).
Pool-wide quintile lift of 6.5× between top and bottom quintile by predicted score.
Returns a verdict in 90 seconds at €0.005 per ad in scoring cost. €0 data acquisition (public APIs only).

// What SaliencyLab will not claim

Does not predict sales lift, ROAS, attributed conversion, or in-market brand recall.
Does not reproduce survey-based recall methodologies, we use behavioral proxies, not consumer interviews.
Meta-Feed scores are directional defaults, brand-ad outcome data is sparse, no held-out validation yet.
Heatmaps are predicted visual attention, not measured eye tracking.
Attribute detection is ~85% accurate, not 100%. Formal inter-rater paper still in pipeline.