BrandMapGPT

Home Page
Result Page

🚀 Inspiration

We were inspired by the universal challenge that brands face when entering new markets: a lack of cultural context and insight. Traditional market research is often slow, expensive, and doesn't capture the nuanced tastes of a region.

When we discovered the Qloo Taste Graph API, a cultural intelligence engine, we realized we could bridge this gap with a generative AI assistant that provides contextual brand advice based on real cultural data.

🧠 What it does

BrandMapGPT helps brands and marketers explore and plan entry into new geographic markets by combining cultural insight with generative intelligence.

Given a target country or region, BrandMapGPT:

Uses the Qloo API to retrieve culturally relevant entities (e.g., local musicians, destinations, fashion brands, movies)
Analyzes cross-domain taste patterns of people in the region
Uses Gemini 2.5 Flash to:
- Generate a strategic market summary
- Suggest brand tone, marketing approaches, and positioning ideas
- Identify local collaborations or avoidances based on cultural sensitivity
- Provide competitive context using Qloo's taste graph data
Presents the output in a conversational and visually organized interface

🛠️ How we built it

Frontend: React + Tailwind UI
Form-based interface to input target countries, and a dynamic card UI to display insight summaries
Backend: Django + REST Framework
Endpoint to query Qloo's search, insights, and recommendation endpoints, with middleware to sanitize, validate, and enrich user input. GPT-4 prompt orchestration generates actionable recommendations.
AI Stack:
- GPT-4 (OpenAI API) for language generation
- Qloo API for cultural data
- Gemini (used as a fallback agent for alternate insights when needed)
Database: PostgreSQL for session logs and entity caches
Testing & Debugging: Postman, curl, and unit tests

🧱 Challenges we ran into

Qloo’s API returned errors for invalid or low-density entities. It took careful parsing of valid entity types (like urn:entity:locality) and proper URL encoding to get consistent results.
Signal filtering using Qloo’s signal.interests.entities required entity IDs with the correct structure and subtype. We had to reverse-engineer from sample IDs.
Gemini flash model occasionally hallucinated when Qloo data was weak, so we added fallback clarifications and entity fallback sequences.
Rate limits from LLM APIs made debugging iterative requests tricky.

🌟 Accomplishments that we're proud of

Built a fully working pipeline from query → insight → GPT reasoning → strategic output within 48 hours
Successfully combined cultural intelligence (Qloo) with strategic narrative generation (GPT)
Delivered real-time insights for over 50+ global locations during testing

📚 What we learned

Gained proficiency in working with Qloo’s cultural graph, including handling UUID-based signal filtering and decoding entity types
Learned how to extract insights from cross-domain recommendations
Understood how to balance factual data from APIs with generative reasoning from LLMs

🔮 What’s next for BrandMapGPT

Add multi-region comparison, such as "How does Gen Z in Brazil compare to Spain for skincare brands?"
Allow users to upload their brand deck for more personalized recommendations
Introduce taste-based product recommendation modules
Explore fine-tuning LLMs on historical marketing playbooks for deeper personalization
Launch as a B2B product for agencies and global brands