SignalForge

Inspiration

AI tools have made it easier to build software quickly, but deciding what to build still requires product judgment. Builders often have messy user interviews, support tickets, survey notes, and feature ideas, but no clear way to turn those signals into a scoped MVP, success metrics, and a launch-ready plan.

SignalForge was built to close that gap: make product decisions traceable, evidence-backed, and ready to ship.

What it does

SignalForge turns raw product evidence into a complete decision pack.

Users can start from a sample project or paste their own feedback. SignalForge extracts evidence items, groups them into product opportunity clusters, infers personas and jobs-to-be-done, scores opportunities by impact, confidence, and effort, and highlights supporting evidence, contradictions, and gaps.

The selected decision becomes a Ship Pack with:

• a ranked product recommendation,
• an MVP cutline,
• assumptions and evidence gaps,
• a contextual PRD,
• a product analytics tracking plan,
• a demo script,
• and a shareable decision brief.

The app is designed so judges can reach first value without login.

How we built it

SignalForge is a JavaScript single-page app deployed on Vercel with serverless API routes. The core analysis engine uses a grounded deterministic pipeline so the product remains usable even without external API keys. It also supports an optional OpenAI Responses API strategic review path when configured.

The app includes serverless routes for analysis, artifact generation, health checks, share links, and feedback. Share and feedback can use Supabase persistence when configured, with browser-safe fallbacks for the demo experience.

Novus/Pendo is installed in production and tracks the core product journey: landing, sample start, analysis start, analysis completion, decision selection, Ship Pack view, tracking plan view, artifact copy/download, and share creation.

Challenges we ran into

The hardest part was making the product feel useful rather than just generative. Product artifacts are only valuable if they are grounded in evidence, scoped to a real decision, and connected to measurable outcomes.

We also focused on making the judge flow reliable: no login, fast sample path, safe input handling, deterministic fallback analysis, shareable output, production smoke tests, and Novus proof events.

Accomplishments that we're proud of

We shipped a complete end-to-end product workflow instead of a static demo. SignalForge goes from messy evidence to a concrete product decision, then turns that decision into artifacts a real PM or founder could use: PRD, MVP scope, tracking plan, demo script, and shareable brief.

We are also proud that the app is deployed, test-covered, Novus-instrumented, and usable without requiring private data or account setup.

What we learned

AI product tooling works best when it helps people make better decisions, not just produce more text. The important loop is evidence → decision → scope → measurement.

We also learned that analytics instrumentation should be part of the product design from the beginning. Novus made us think clearly about the activation funnel and which moments prove that users reached value.

What's next for SignalForge

Next, we would add team workspaces, stronger import sources, richer Novus-powered product intelligence, persistent share and feedback storage by default, and deeper LLM review for custom evidence.

The long-term goal is to make SignalForge a product decision cockpit for teams shipping fast: every feature idea linked to evidence, confidence, scope, and measurable outcomes.

Built With

• javascript
• node.js
• novus/pendo
• openai-responses-api
• playwright
• supabase
• vercel