SponsorFit

Turn audience cultural taste into brand sponsorships

Comment

Inspiration

Independent creators—newsletter writers, boutique podcasters, and essayists—face significant hurdles when pitching prospective sponsors. When seeking brand partnerships, creators are commonly restricted to blunt demographic summaries: "We reach 35,000 monthly listeners aged 25–40" (as illustrated in our demo sample preset).

This commoditizes creators, forcing them to compete directly against programmatic ad networks on CPM price. Meanwhile, brand sponsorship outreach often demands authentic cultural resonance, yet creators lack transparent, data-driven tools to demonstrate shared audience affinities.

We built SponsorFit to explore replacing flat demographics with cultural taste. Powered by Qloo's Taste Graph, SponsorFit enables creators to evaluate empirical cultural alignment using structured entity queries.

What It Does

SponsorFit is a cultural intelligence agent that turns audience taste into quantifiable, pitch-ready sponsor partnerships:

  1. Audience Seed Grounding: Creators enter 3–5 cultural anchors their audience loves (e.g., A24, Phoebe Bridgers, Severance, Tartine Bakery).
  2. Canonical Entity Resolution: Queries Qloo /search to resolve input keywords into canonical cultural URNs across cinema, music, dining, and brands.
  3. Taste Graph Traversal: Ingests high-affinity candidate brands and explainability signals via Qloo /v2/insights.
  4. Deterministic Ranking Engine: Ranks sponsor candidates by affinity scores and computes relative multipliers against candidate sample averages ($A_i / \bar{A}$).
  5. Grounded Cultural Media Kit: Synthesizes a one-page media kit and tailored cold outreach pitch letters where every single statistic links directly to an underlying Qloo response field—eliminating hallucination.
  6. One-Click Export: Download the generated Media Kit as Markdown and copy personalized pitch emails with a single click.

How We Built It

We designed SponsorFit from scratch with a focus on mathematical grounding, speed, and zero hosting costs:

  • Frontend: Built with React 19, TypeScript, and Vite. Features an editorial typographic design (Newsreader & Space Grotesk) optimized for demo clarity at 1440x900 and 1280x720, with a compact summary bar keeping rankings above the fold.
  • Agent Core & API: Framework-neutral ES module architecture (api/agent.mjs, api/qloo-client.mjs, api/handler.mjs) compatible with Web Standard APIs (Request/Response). Runs seamlessly on Node.js, Vercel Functions (api/[...path].mjs), and Cloudflare Workers.
  • Transparent Mock & Live Layer: Because Qloo hackathon keys are provisioned manually, the client implements a transparent mock layer conforming to official OpenAPI schemas, transitioning to live queries on https://hackathon.api.qloo.com upon setting QLOO_API_KEY.
  • Honesty Verification: Built an automated mathematical honesty test suite (api/honesty-audit.test.mjs) asserting that every number displayed traces to an official API response path.

Challenges We Ran Into

  1. Provenance Enforcement: LLMs tend to embellish partnership pitches with unverified statistics (for example, inventing unbacked comparative claims or unsubstantiated percentages). We solved this by implementing a deterministic template synthesizer and strict provenance tracking: every pitch explicitly cites underlying Qloo fields (results.entities[i].affinity and query.explainability.signals).
  2. Above-the-Fold Demo Polish: Initial testing showed results starting below the fold on 1440x900 screens. We engineered a responsive view switcher with a collapsible summary bar so judges see ranked sponsors immediately without excessive scrolling.
  3. Free Tier Hosting Compatibility: We architected the API to be framework-neutral using pure standard fetch and Web Standard Request/Response, allowing zero-cost deployment on Vercel Hobby or Cloudflare Workers with $0 budget.

Accomplishments That We're Proud Of

  • 16 of 16 Tests Passing: Complete automated test suite covering live request building, mock fallback, ranking math, honesty provenance, and Vercel serverless adapters (computed by npm test).
  • Deterministic Field Provenance: Pitch letters don't just recommend sponsors—they prove why with exact field citations (results.entities[0].affinity = 0.942). In mock mode, candidate entities (such as sample entities Aesop and Oatly) are evaluated strictly against sample mock fixtures; in live mode, candidates reflect live Qloo Taste Graph API query responses.
  • Demo-Quality UI: Editorial, warm aesthetic crafted specifically for human judges and creators, verified headless via Edge CDP automation with zero console errors.

What We Learned

  • Qloo's cross-domain taste graph uncovers fascinating, non-obvious affinities that demographics miss entirely (e.g., an indie film fan base indexing strongly on specialty espresso hardware).
  • Explainability signals (query.explainability.signals) provide concrete, verifiable context for outreach proposals: rather than relying on abstract claims, proposals link directly to specific cultural drivers identified by the taste graph (e.g., sample signals linking candidate affinity to audience affinity for A24).

What's Next for SponsorFit

  • Multi-Platform Audience Ingest: Direct sync with Substack subscriber surveys and Spotify for Podcasters data.
  • Dynamic Live LLM Pitch Variations: Activating live Google Gemini 2.5 Flash synthesis when GEMINI_API_KEY is provided to generate multiple tone variants (concise, creative, executive).
  • Outreach Tracker: Lightweight CRM tracking response rates from pitched brand marketing teams.

Built With

Share this project:

Updates

Submission history