Feature Funeral

A structured way to kill features intentionally.

Comment

What I built

Feature Funeral is a tool that helps product managers decide whether to keep, kill, or investigate a feature — and if it's time to kill it, gives that feature a proper, shareable send-off instead of a quiet deprecation nobody acknowledges. A PM describes the feature and either uploads real usage event data (CSV) or enters rough estimates. An AI verdict engine reasons across four dimensions — usage trend, opportunity cost, risk if removed, and strategic fit — and returns a confident Keep, Kill, or Investigate verdict with a reasoning breakdown. If the verdict is Kill, the feature gets a tombstone-style page: an AI-written epitaph, cause of death, eulogy, and a note on what happens to its users — built to be screenshotted into Slack and shared. Every kill is added to a public Graveyard, a running record of what a team has deliberately chosen not to maintain.

Who it's for

PMs and product leads at companies with feature sprawl — five-plus legacy features nobody owns, that everyone privately agrees should die but no one wants to be the one to kill. The tool gives them a structured, defensible case instead of a gut-feel argument in a planning meeting.

What I used

Next.js (App Router) + TypeScript + Tailwind, deployed on Vercel. The verdict engine uses the claude-haiku-4.5 model via the Vercel AI SDK, and stores the response in a Prisma Postgres DB. Novus.ai installed for usage analytics.

What I learned

  • How AI can be used to appraise the utility of app features, and helps to decide what to sunset, keep or investigate, by weighing four named dimensions explicitly; usage trend, opportunity cost, risk if removed, and strategic fit. Specificity in the prompt produced specificity in the product.

Built With

  • aisdk
  • baseui
  • claude-haiku-4.5
  • next-themes
  • nextauth
  • nextjs
  • prisma
  • tailwindcss
  • typescript
  • zod
Share this project:

Updates