Quickstart
Create a SaliencyLab project, add brand context, run a first RoastIQ analysis, and turn the result into a concrete edit.
1. Create a project
Open Projects and create a project for the brand or campaign you are working on. A project keeps brand context, creative history, analyses, and follow-up work connected.
Use a name another teammate will recognize later. A brand plus campaign, market, or quarter is usually more useful than “Test 1.” If the audience, proposition, or market changes materially, create a separate project rather than mixing unlike decisions.
2. Add useful brand context
Add the information that should remain stable across analyses: brand, category, market, positioning, and distinctive assets. Specific context makes the output more useful than a standalone upload.
Prioritize facts that change how the creative should be judged:
- The audience and buying situation
- The intended platform and placement
- The proposition or offer the ad must communicate
- Mandatory brand assets, product cues, or legal copy
- The role of the creative in the campaign
Avoid pasting personal data, credentials, customer lists, or information that is not necessary for the analysis.
3. Start a new analysis
Choose New analysis, upload an image or video ad, and confirm the platform, market, format, and category. These fields determine which context and benchmark slice is relevant.
Before submitting, check that you uploaded the intended export—not a thumbnail, rough file, or version without final audio. For video, confirm that the opening frame, sound, captions, and end card match the version your team is considering.
4. Wait for the result
Video analyses may take longer than static images because the pipeline processes scenes, audio, text, and visual evidence. You can leave the processing page and reopen the result from the project or results library.
Do not create duplicate jobs simply because a video is still processing. First return to the project and check whether the original analysis is continuing or has moved to a failed state.
5. Make one decision
Start with the verdict and the highest-priority recommendation. Then inspect the evidence behind it. Use the five KPI scores as diagnostic dimensions, not as five independent targets to maximize.
A good first test: Use a real creative that your team is currently debating. The product is most useful when its evidence can change a concrete edit, approval, or media decision.
6. Read the result in the right order
- Start with the action verdict: Rebuild, Sharpen, or Scale.
- Read the summary explaining why that action is recommended.
- Inspect the evidence behind the strongest and weakest KPI dimensions.
- Check benchmark fit, fallback language, and confidence.
- Select one edit that can be expressed as a testable hypothesis.
For example: “Move the brand cue into the opening two seconds because attention is currently attached to the scene, not clearly to the brand.” That is more actionable than “increase Brand Impact.”
7. Compare responsibly
Keep variants in the same project and hold the analysis context steady where possible. A score change is easier to interpret when the platform, market, category, format, and intended role remain comparable. If those inputs change, treat the result as a new context rather than a clean before-and-after test.
Before sharing the result
- Confirm the recipient understands that scores are modeled diagnostics.
- Include the evidence and recommended edit, not only a screenshot of the score.
- Preserve benchmark and confidence labels.
- Separate SaliencyLab output from observed campaign performance.