Inspiration

A star rating tells a company whether someone liked a product. It rarely explains what they needed, why they considered buying it, or what would make the product more useful.

We built SignalScope around a different question: What product decision should this feedback help us make?

What it does

SignalScope is a Voice-of-Customer research workspace that connects customer feedback, unmet needs, competitor evidence, and product improvement hypotheses.

Its guided workflow helps researchers:

  • Select a product and inspect its documented capabilities.
  • Collect permitted feedback from reviews, questions, discussions, and direct responses.
  • Organize evidence across ten customer-need categories, including price, reliability, convenience, and compatibility.
  • Discover relevant competing offerings through an automated research pipeline.
  • Compare supported claims while keeping missing information visible.
  • Build a source-backed roadmap of improvements and experiments.

Our Kearney portfolio demonstration includes Velocity AI, Tariff Simulator, Ingredient Compass, Ignite, and PERLab Design-to-Value services.

How we built it

We built the application with TypeScript, React, TanStack Start, Tailwind CSS, and Recharts. An original Three.js visualization introduces the connection between customer signals and product decisions.

The backend uses Google’s Gemini SDK behind a configurable provider interface. Source collection, feedback extraction, competitor research, and roadmap synthesis are separate stages. Validation checks quotations against captured source text before accepting an analysis.

What makes it different

SignalScope keeps customer feedback, provider claims, and research hypotheses separate.

A product page is not a customer review. A missing review corpus is not evidence of satisfied customers. A recommendation is a hypothesis to test, not a proven diagnosis.

Source references and explicit evidence gaps help researchers inspect the reasoning behind a proposed improvement.

Challenges and lessons

We encountered model-access errors, source-capture failures, and deployment constraints. We added actionable diagnostics and product-description fallbacks rather than substituting invented findings.

We learned that useful consumer insights require more than sentiment analysis. They require context, traceability, and a clear next decision.

What’s next

We plan to validate the workflow with researchers, connect authorized customer datasets, evaluate insight quality against human coding, and improve persistent research storage. Live research coverage still depends on provider access and source availability.

Built With

  • bun
  • cheerio
  • google-gemini
  • google-genai-sdk
  • playwright
  • react
  • recharts
  • tailwind-css
  • tanstack
  • three.js
  • typescript
  • vercel
  • vite
  • vitest
  • zod
Share this project:

Updates

Submission history