Attention collapses early
A weak opening loses the viewer before recall, lift, or sentiment studies ever get the chance to measure an outcome. The skip happens in roughly 1.7 seconds, the survey arrives weeks later.
Research philosophy Upstream of validation, not in place of it
Most ads fail before traditional research can measure the failure.
SaliencyLab exists to explain the early perceptual risk while the asset is still movable, before media spend turns a weak opening into a six-figure mistake, and before lift studies measure what was already lost in the first second.
The first-second problem
The gap before validation
By the time recall, lift, and sentiment studies report back, the decision has already cost money. Three things go wrong in the first second, and none of them are what downstream research is built to catch.
Traditional research measures downstream
Lift, recall, and brand-tracking studies can tell a team whether a message landed. They are not designed to explain the first-second perceptual filtering that decided whether the viewer ever got there.
Creative teams still need an upstream read
That is the gap SaliencyLab is built to serve: the moment before spend, before validation, when the asset is still cheap to change and a single edit can move the entire campaign trajectory.
Complementary, not competing
Three roles in the same workflow
SaliencyLab does not replace lift studies, brand trackers, or human panels. It runs earlier, and makes the deeper research more valuable by routing only the surviving routes into it.
Traditional research
Validates outcomes
Confirms memory, lift, and recall after exposure. Tells the team what happened.
- Lift studies, brand trackers, surveys
- Two to six weeks to report
- $30k–$150k per study
- Best on the final asset
SaliencyLab
Explains the creative
Diagnoses the asset before or alongside validation so the team can adjust earlier.
- Attention, brand impact, proposition
- Under 90 seconds per ad
- Included in your Pro plan limits
- Best on early cuts and rough edits
Together
Sharper decisions, less waste
Use SaliencyLab upstream to filter. Escalate into deeper research only when the route deserves it.
- Triage early, validate late
- Kill weak routes before research spend
- Bring stronger candidates to lift
- Closer to a real pre-spend decision
| Dimension | Traditional research | SaliencyLab |
|---|---|---|
| Position in workflow | After final cut, before launch, validation | Upstream Before or during creative development |
| What it answers | Did this ad work? (outcome) | Why is it likely to work, or not? (diagnosis) |
| Time per read | 2–6 weeks | 90 seconds (images) · ~3 min (video) |
| Cost per asset | $30k–$150k per study | Included within Pro plan limits |
| Validation target | Sales lift, recall, attributed conversion | Engagement & click-intent on public outcomes |
| Held-out signal | Reported per study | Spearman ρ +0.30–0.32 (TikTok engagement, TikTok CTR, YouTube views) |
How it changes the decision
From taste & tabs to a first signal
The creative review meeting today is mostly subjective: gut calls, opinions in a Figma comment, an executive overriding a brand planner. SaliencyLab adds a defensible upstream read, without locking the team out of judgment.
i
Step 01 · Concept
Creative concept
The team has an asset, a debate, and a launch timeline. Today the decision is made on taste, with two or three voices dominating the room.
ii
Step 02 · Diagnose
SaliencyLab analysis
Score the asset, read the benchmark frame, and pressure-test the route. The score is a model prediction, not a verdict, but it removes the floor of the debate.
iii
Step 03 · Validate
Validation research
Escalate into lift, recall, or human-panel work only when the question genuinely requires that depth, on stronger routes, not weak ones.
Truth-first
What we claim, and what we do not
The validation envelope is the product. Overclaiming kills it. Here is the line, drawn in public.
✓ What SaliencyLab predicts
Engagement and click-intent on public outcomes
- ✓Public engagement (likes, shares, comments) on TikTok
- ✓Click-intent (CTR percentile band) on TikTok, calibrated separately
- ✓YouTube view-count percentile
- ✓Cross-validated on 1,200+ ads with public outcome data
- ✓Held-out OOS Spearman ρ of +0.30–0.32 (2026-05-05)
- ✓Pool-wide quintile lift 6.5× (top vs bottom quintile)
✗ What we do not claim
Downstream business outcomes
- ✗Sales lift or attributed conversion
- ✗ROAS or media-mix attribution
- ✗Brand recall or purchase intent (those are survey constructs)
- ✗A replacement for human panels or lift studies
- ✗Meta-Feed scores are directional defaults, not held-out validated yet
- ✗A measurement of eye-tracking, heatmaps are predicted attention
1,200+
Ads with public outcome data in the validation set
+0.31
Held-out Spearman ρ · TikTok engagement (n=700)
6.5×
Pool-wide quintile lift, top vs bottom by predicted score
90s
Median time per image analysis on the platform
Validation snapshot 2026-05-05. Meta-Feed cohort data is currently sparse for brand ads, so SaliencyLab returns directional defaults there, not held-out scores. Full methodology in /methodology.
“
We are not trying to measure every possible outcome.
We are trying to explain the creative before the wrong decision gets locked in.
Oussama Nakhil
Founder · SaliencyLab
Common objections, answered
Questions teams ask in the first meeting
If a question is not here, write us, we keep this list honest as the validation envelope grows.
Does SaliencyLab replace traditional ad research?
No. SaliencyLab sits upstream of validation. We diagnose perception risk before spend, attention, brand impact, proposition clarity, so teams can fix the creative before lift, recall, or sales studies are commissioned. The right workflow is: SaliencyLab first to triage, then escalate into deeper research only on the surviving routes.
What does SaliencyLab actually predict?
Public engagement (likes, shares, comments) and click-intent (CTR percentile) on TikTok, with separate calibrated scores. Held-out OOS Spearman ρ of +0.30–0.32 across TikTok engagement, TikTok CTR, and YouTube view counts, cross-validated on 1,200+ ads with public outcome data. Meta-Feed scoring is directional defaults today, we will only graduate it to held-out validated once we have enough cohort data.
Why don't you predict sales or ROAS?
Because those are downstream outcomes governed by media mix, pricing, distribution, seasonality, and brand equity, almost none of which the creative file can explain on its own. Anyone who claims a single-creative ROAS prediction is overclaiming. We predict the leading indicators that come before those, engagement and click-intent, and stay honest about the boundary.
How is this different from human panels or eye-tracking studies?
Human panels and eye-tracking measure real people. SaliencyLab predicts what is likely to happen based on a model trained against public outcome data. It is faster and cheaper by 2–3 orders of magnitude, but it is a prediction, not a measurement. Use SaliencyLab to filter; use panels to validate the surviving routes when the budget and decision weight justify it.
Why 90 seconds? What is the constraint?
Because anything slower falls outside the creative review loop. A 4-hour analysis becomes a thing you commission. A 90-second analysis becomes a thing you do in the meeting. The whole product is shaped around fitting inside the moment when the asset is still movable.
How do you keep the benchmark credible?
Manually curated. Every ad in the benchmark pool is tagged across 43 attributes (format, category, brand role, hook type, CTA, etc.), and we publish the validation methodology in /methodology. We do not scrape, we use official public transparency tools (Meta Ad Library, TikTok Creative Center, Google Ads Transparency). When you see a benchmark score under 100 ads, we show "Building, N ads" rather than pretending the sample is final.