Blindspot

When a launch flops, the sales drop is loud and the cause is quiet. Blindspot uses Qloo's taste graph to find the likely hidden cause: a gap between what a brand sells and what its crowd loves

Comment

Inspiration

A doctor sees the symptom first. The disease is quiet.

Business works the same way. When a launch flops or a venue's regulars stop coming, the effect is loud: sales fall. The cause is quiet, and teams usually blame the loudest suspects, like price and ad spend.

We wanted a tool that works backward from the symptom and asks a different question: does what we are selling match what our crowd actually loves?

What it does

Blindspot is a taste diagnostic. You enter a brand, venue or launch, and the audience it was meant for.

It then:

  1. Compares the brand's audience with the audience of the partner or event it chose, using Qloo's taste graph.
  2. Shows where the two overlap and where they do not.
  3. Suggests alternatives with better overlap, and drafts a short outreach note.

It reports a likely cause with evidence, not proof. Taste data shows alignment. It does not show sales.

The app includes three illustrative scenarios so it can be explored without a key. They are examples, not recorded Qloo output. [KEEP the next sentence only if your live key works, otherwise delete it:] With a Qloo hackathon key connected, you can query any brand or entity live.

How we built it

Plain Node.js with no runtime dependencies, a small HTTP server, and a front end in HTML, CSS and JavaScript, deployed on Vercel.

Requests to Qloo go through a validated client. It checks every request against an allowlist before sending it and logs the endpoint and response time.

The client is built around Qloo's search, compare and insights endpoints. The repo also includes MCP configuration and workflow files for the Qloo harness.

Challenges we ran into

  • Access. Hackathon keys are issued manually, so we built the app and its checks first and connected live calls as access allowed. [Edit this to match what really happened.]
  • Staying honest. It was tempting to show polished numbers. We labeled the example scenarios as illustrative and kept the "likely cause" wording, because taste is not sales.
  • Shipping in a short window.

Accomplishments that we're proud of

  • A clear flow from symptom, to likely cause, to fix.
  • A request validator that blocks unsupported parameters and old routes.
  • A test suite you can run with npm test. [Run it first and write the real number.]
  • A public repo, MIT license, and a one-command local start.
  • Limits written into the product, not hidden.

What we learned

  • Fame and fit are different things.
  • Starting from the symptom changes the question you ask.
  • Honest wording earns more trust than big claims.
  • Qloo's data describes groups, not individuals, and the product has to respect that.

What's next for Blindspot

  • Live audits on real entities, with the results recorded.
  • Let users enter their own budget so the stakes appear next to the result.
  • Track a brand over time to see whether the gap is closing.
  • Offer the diagnostic as an MCP tool other agents can call.

Built With

Share this project:

Updates

Submission history