Work with us
We review the evidence with your team and decide which creative to prioritize together.
The decision system
Two tools. One ladder. Score first, then explain.
RoastIQ runs first on every creative. If the verdict is clear, ship. If the score reveals buyer resistance, BuyerLens opens from the same project, same context, same evidence trace.
1
You are here · RoastIQ
The pre-spend gate
Inspect five modeled KPIs, the supporting creative evidence and predicted attention where available. Use the findings to prioritize an edit and a controlled test.
Use when: you have a cut and need a verdict before the media plan goes live.
→↓ if Sharpen or Rebuild
2
Step 2 · BuyerLens
The buyer-resistance layer
Synthetic users score the same creative against your real buyer segments. Surfaces the objection patterns the KPI score can't, language, framing, cultural fit. Never runs without a RoastIQ result first.
Use when: the verdict is borderline and you need to know why a segment is bouncing.
The shift
The verdict used to take three weeks. Now it takes minutes.
Most creative testing happens after launch, by the time data comes back, the cut is locked and the budget is spent. RoastIQ moves the verdict to the moment the file is exported.
What you get back · authored example
Your creative. The evidence. The next edit.
A 60-second Coca-Cola creative, five diagnostics and an edit plan. Explore how the recommendation connects to the evidence in this authored example.
Keep
The holiday atmosphere and distinctive brand cues.
Change
Make the closing message and next step clearer.
Test
Compare the edit with the original, using the same audience and brief.
Validation
Calibrated against real ads, not vibes.
Separate platform outcome models are evaluated against public TikTok CTR and YouTube view counts. These figures do not validate the five diagnostic KPIs. Reports show the available benchmark scope; customer campaign outcomes require in-market measurement.
1,200+
Ads in the historical calibration pool
ρ +0.33
TikTok click-through rate · held-out Spearman, n=691
ρ +0.40
YouTube view counts · held-out Spearman, n=397
6.5×
Top-quintile vs bottom-quintile lift
Last refresh 2026-08-25 · Read the full methodology →
Before you send an asset
Formats, scope and what you receive.
- Images
- JPEG, PNG and WebP, up to 50 MB per file. Image analysis excludes audio and motion-specific stages.
- Video software intake
- MP4 and MOV, up to 180 seconds and 100 MB per file. Missing audio, transcript or scenes limit what a report can conclude.
- Expert review packages
- One image or video up to 60 seconds per creative. Each edit, language version or placement adaptation counts separately. Longer videos need a separate quote.
- Comparisons
- Use the same objective, market, platform and audience. A creative with incomplete evidence cannot support a confident winner. Scores from different providers are not directly comparable.
- Reports and exports
- RoastIQ diagnostics and revision priorities; separate BuyerLens synthetic studies. PDF/JSON report export is a paid software feature. Service deliverables are included in the agreed scope.
GIF, HEIC, SVG, TIFF, AVI, MKV and WebM are not supported by this intake. AI diagnostics and simulated interviews do not measure consumer responses or campaign lift.
How we work together
You bring the ads.
We help you decide.
Bring the creative and the decision at stake. We help your team choose which direction deserves the budget.
- 01 / BRIEF
Show us the ad.
Share the creative, audience, market and what you need to decide.
A focused review brief - 02 / REVIEW
We find the friction.
Creative diagnostics first. Category references and simulated buyer interviews where they help answer your question.
Evidence, reviewed by our team - 03 / DELIVER
Know where to double down.
A clear recommendation on which creative to back, what needs improving, and what to leave behind—with the reasoning for each choice.
Scope and timing agreed before we begin.
Campaign outcomes are measured in market.
Questions
The honest answers.
What is RoastIQ by SaliencyLab?
RoastIQ analyzes an image or video ad with its brand and platform context. Five modeled creative KPIs organize the diagnosis, source evidence and next-edit hypotheses. Use it to prepare a creative decision before testing the result in market.
Why is RoastIQ “Step 1”?
RoastIQ establishes the creative diagnostic and saved context. Start with the Keep / Change / Test recommendation, then inspect its evidence and human review status. BuyerLens opens from that result when a specific messaging question needs simulated buyer exploration.
Is the score predicting sales or ROAS?
No. The five KPIs are modeled creative diagnostics. A separate platform-native attribute model is evaluated against public TikTok CTR and YouTube views. Neither a diagnostic score nor simulated interviews establish sales lift, ROAS or attributed conversion.
How is this different from a generic AI critic?
RoastIQ keeps the creative, brand context, platform and evidence in the same project. Reports distinguish modeled diagnostics from benchmark references, and show the scope and limitations of available evidence.
What about my own data and IP?
Your creative is private to your workspace. We never use customer ads to train scoring models, calibration runs on the public benchmark cohort only. Storage is purged on request.
How accurate is the heatmap?
The heatmap is a predicted attention map from a saliency model, not measured eye tracking. Use the available regions and sampled video frames to form an edit hypothesis. It does not prove what a viewer actually saw.
Does it work for static images and short videos?
Yes. Supported formats: MP4, MOV, JPG, PNG, WebP. Software accepts MP4/MOV videos up to 180 seconds and 100 MB, and JPEG/PNG/WebP images up to 50 MB. GIF and HEIC are not supported. Static images skip audio and video-specific stages. Fixed-price expert review packages cover videos up to 60 seconds; longer assets need a separate quote.