ThinQ - Taste Target

TasteTarget: Predict what your audience feels — not who they are. Sell smarter with AI-powered cultural insight. No cookies, just cravings. Precision marketing for a post-demographic world.

Comment

A. 🎨 Inspiration

Modern marketing is fundamentally flawed. Brands still rely on outdated demographic categories like “18–25-year-old females” or “urban millennials”, broad labels that fail to capture what people actually care about. These generic groupings ignore the emotional and cultural drivers behind real behavior.

With TasteTarget, we flipped the script: Instead of asking who your customers are, we ask what they love, feel, and value. That’s what inspired us to build a privacy-first cultural intelligence engine that helps brands understand audiences based on Taste, not traits.

B. 📚 What We Learned

Throughout the build, we explored how AI can match, and even enhance, the way marketers think:

  • Learned to simulate Qloo’s API using GPT-4 + structured prompt engineering to create realistic taste-based audience clusters.
  • Experimented with tone-matching to generate emotionally resonant content tailored to each persona’s “vibe.”
  • Studied traditional persona-building workflows to understand how AI could make them faster, cheaper, and culturally deeper.

C. 🛠 How We Built It

We developed a full-stack MVP that integrates multiple AI services into a single seamless experience:

  1. Frontend:
  • Built with Streamlit for a clean, fast interface that supports dynamic input and real-time previews.
  1. Backend:
  • Powered by FastAPI, handling API interactions, form submissions, and persona generation logic.
  1. LM Integration:
  • Used OpenAI's GPT-4 to replicate Qloo-style taste clustering and auto-generate all campaign assets.
  1. Visual Generation:

Integrated Hugging Face models to create culturally tuned logos and posters based on audience vibes.

D. 🔑 Key Features

  1. Enter brand/product info and target mood.
  2. Instantly receive 2–3 culturally rich, taste-based personas.

  3. Auto-generate:

    • Slogans
    • Social captions
    • Blog intros
    • Full campaign messaging

  4. Generate matching visuals (posters/logos).

  5. Export full reports in JSON or PDF formats.

  6. Optional: Influencer and content strategy suggestions based on persona profiles.

E. 🚧 Challenges We Faced

  1. Simulating Qloo without API access We reverse-engineered its clustering logic using GPT-4 and prompt patterns, a creative challenge in precision and structure.

  2. Tone generation complexity Capturing the right emotional tone (bold, poetic, humorous, etc.) meant chaining prompts and dynamically adjusting based on persona traits.

  3. Balancing creativity and UX clarity Designing a flow that feels creative while staying data-informed required careful UI/UX decisions and a flexible system architecture.

Built With

Share this project:

Updates