Inspiration

I've been in the marketing industry for over 5 years, and I found every marketer asks the same questions. Brand Manager — ‘Which ad maximizes consideration for Android users? Creative Director — ‘What exact shot should I change? But pre-testing today is broken. Surveys tell you what. Customer feedback tells you why. Videos tell you what’s on screen. But here is the thing - they are not connected — so marketers wasted up to $245 billion per year globally on creative that doesn’t move the needle. Like Swayable, that’s the gap we’re closing — by turning creative data into something marketers can actually act on Monday morning.

What We Built

We built Creative Compass — an AI-powered system that turns raw pre-test data into actionable creative intelligence with qualitative feedback and video content cues.

It then synthesizes everything — by segment: device, age, income, region — and delivers a creative scorecard showing: Favorability, Consideration, Conversion

What’s working and why — integrating data, sentiment, and visuals.

What to change next — giving concrete creative recommendations (“Keep humor in the intro; shorten product shots; test social proof overlay”).

It produces marketer-friendly reports that explain

How We Built It

We built a structured insight framework prompting an LLM to synthesize multi-modal data (quantitative + qualitative + visual). Our pipeline includes:

Data Parsing: Clean and align pre-testing metrics, verbatim responses, and scene-level metadata.

Insight Generation: Custom LLM prompt templates that output recommendations grouped by:

Keep / Amplify

Change / Shorten / Clarify

Add / Test

Challenges

Integrating unstructured qualitative data with structured metrics without losing nuance.

Teaching the model to speak like a strategist.

Ensuring the system scales from one ad test to hundreds—without manual review.

Built With

Share this project:

Updates