PopNet Venture Evaluation & PRD Framework

The problem

A solo operator with more ideas than build hours needs a consistent way to kill weak ideas early, before time goes into code. Ad hoc evaluation led to repeated false premises. Twice, market-gap claims like "no software exists for X" turned out to be wrong once they were researched. New projects also kept starting before open decisions on active ones were resolved.

What I built

• A 12-section evaluation pipeline covering validation, competitive landscape, business models, MVP definition, go-to-market, financials, roadmap, risk, resources, Taste & Story, verdict, and PRD generation
• An "infrastructure lens" applied to every idea (product vs. tollbooth, control points, network effects), plus a separate 1–10 infrastructure score alongside viability
• A Taste & Story layer that evaluates founder-market narrative and aesthetic fit for the industry, plus a counterpoint audit covering market, execution, competition, and business-model failure points
• A score gate: PRDs are generated only for software ideas scoring 7 or higher on viability
• A PRD template (v2) with lettered clarifying questions, atomic user stories with verifiable acceptance criteria, dependency maps, and a Claude Code starter prompt
• A mandatory Agent Opportunities section, with an agent inventory, full system prompts, idempotency requirements, an agent_runs logging schema, and quality gates
• Research requirements: web verification of competitors and pricing, and API cost checks before numbers are presented
• Workflow rules: separate planning and build projects to limit context bloat, kill criteria written before validation calls, and a build-vs-buy check before any build
• A default build stack written into generated PRDs: Next.js, TypeScript, Tailwind CSS, Supabase, Clerk, Stripe, Resend, Vercel, GA4, Magic UI

Stack

Outcomes

Invalid market-gap premises caught before building

2

Source: Recorded in project history. Competitive research disproved two "no software exists" claims.

API cost estimate error caught

10x

Source: Stated in framework research requirements. A realtime voice API cost 10x the initial estimate, so cost verification became a required step.

What is next

• Validation gates on two franchise-industry tools, with kill criteria defined before any build
• Confirming license-key format before Week 1 build of an offline-first lifetime-deal productivity suite
• A creator membership platform, blocked on starter data and final brand/domain

More from PopNet Media

• North Alabama Local Lead-Gen Site PortfolioA portfolio of search-driven lead-gen and affiliate sites for high-ticket local services in North Alabama, built on a versioned playbook that gates every page on facts competitors don't publish.
• Local RV Rental City SitesA growing network of city-level RV rental sites, each built on one codebase and localized to the rules, routes, and venues of its own market.
• Long-Term RV Rental NetworkA network for monthly and long-term RV stays, aimed at the people who need somewhere to live for months rather than a weekend trip.