SaliencyLab documentation
Learn how to research competitor ads, diagnose creative with RoastIQ, interpret benchmarks, and explore buyer reactions with BuyerLens.
SaliencyLab is a creative intelligence workspace for teams making paid-social advertising decisions. It connects three jobs that are usually split across different tools: learning from competitor ads, diagnosing your own creative, and understanding likely buyer reactions.
Use this documentation to move from a creative question to an evidence-backed next step. Each guide explains what the product shows, how to interpret it, and where its evidence stops.
Start with the decision you need to make
- Browse the market: Study competitor ads and public outcome signals before choosing what to make or test.
- Diagnose a creative: See where an ad sits across five creative KPIs and what evidence drives the result.
- Explore the buyer why: Use BuyerLens when the diagnosis leaves an audience question unanswered.
- Review API access: Inspect the asynchronous analysis contract. General REST access is disabled; current access is restricted to an approved QA allowlist.
The recommended workflow
- Create or open a project so brand context and analysis history stay together.
- Browse Market for relevant category, platform, format, and market evidence.
- Run RoastIQ on the creative you are deciding about.
- Read the verdict, evidence, and recommended edit in context—not as isolated scores.
- Open BuyerLens only when you need a deeper explanation of buyer reactions.
What SaliencyLab does not claim: SaliencyLab gives modeled creative diagnostics and benchmark context. It does not guarantee campaign performance or replace live campaign measurement, brand-lift research, or representative consumer research.
Choose the right starting point
| If you need to… | Start here | You should leave with… |
|---|---|---|
| Understand what competitors are running | Market intelligence | A relevant pattern or comparison set |
| Review an image or video before launch | RoastIQ | A prioritized diagnosis and edit hypothesis |
| Explain a score or verdict | Understanding results | The evidence, benchmark, and confidence behind it |
| Explore an audience-specific objection | BuyerLens | Buyer hypotheses to investigate or test |
| Automate analysis from another system | API reference | The contract and current access restrictions |
Core concepts
- Project: the durable container for a brand, campaign, context, and related analyses.
- Creative diagnosis: a modeled assessment of what the ad communicates and how its execution may work.
- Benchmark: the comparison context used to interpret a score. It is most useful when platform, market, category, format, and length are closely matched.
- Confidence: an indication of how strongly the available evidence supports an interpretation. It is not the probability that an ad will succeed.
- Observed evidence: a signal originating from an external platform or measured source.
- Synthetic perspective: a simulated audience response used to generate hypotheses, not representative research.
A practical decision loop
- Write down the decision before opening a tool: approve, revise, choose between variants, or investigate an audience question.
- Select the closest available context instead of the broadest benchmark.
- Read the evidence before reacting to the headline score.
- Turn the diagnosis into one material edit with a clear reason.
- Re-test the variant or validate it with live campaign data.
This loop keeps the output connected to an actual creative decision and makes later comparison more meaningful.
Quick links
What you provide and what you receive
| Workflow | Inputs to prepare | Output to inspect | What still needs verification |
|---|---|---|---|
| RoastIQ | Image or video, platform, market, objective, audience and brand context | Five modeled KPIs, source evidence and proposed edits | Whether the edit improves an actual campaign outcome |
| BuyerLens | A saved RoastIQ result, eligible viewer profiles, study purpose and reviewed questions | Recorded synthetic answers, resistance themes and contradictions | Whether real buyers share those views |
| Market | A category, platform and market question | Available public references and their provenance | Coverage, comparability and permission to reuse assets |
A useful output preserves the chain from the creative to the evidence to the decision. Keep observed platform signals, model interpretations and synthetic answers separate when sharing a report.
Example: the benefit is clear, but the proof is weak
Use RoastIQ to locate the exact claim and the evidence presented for it. If the report points to an unsupported product promise, ask BuyerLens which information a simulated buyer would need to assess it. Read the original answers before interpreting a theme. Then write a revision brief that adds verifiable proof and states what a real campaign test or human interview should measure.
This example describes a workflow, not a measured customer result. An improved score or repeated synthetic answer does not establish conversion lift.
Evidence and access checks
- Read methodology and confidence before interpreting validation figures or cohort comparisons.
- Use the troubleshooting guide when media evidence is incomplete or audience coverage is insufficient.
- Review API access restrictions and supported MCP clients before planning an integration.
- Review workspace security before sharing creative or reports outside your team.