Skip to content

Instantly share code, notes, and snippets.

@bigsnarfdude
Last active May 26, 2025 16:56
Show Gist options
  • Save bigsnarfdude/3f5c7020586bc82b6aaebd3e803eccc6 to your computer and use it in GitHub Desktop.
Save bigsnarfdude/3f5c7020586bc82b6aaebd3e803eccc6 to your computer and use it in GitHub Desktop.
wikiPRD.md

PRD: Wiki Search That Actually Works

Fix the damn search box


Document Info

Owner: Alex Chen
Status: Ready to build
Timeline: 4 weeks
Last updated: May 26, 2025


The Problem

Our wiki search is broken. When someone searches "deploy backend," they get 47 results including a 2019 holiday party planning doc. Real examples from last week:

  • Sarah spent 45 minutes finding the API rate limiting guide (it was buried in "Infrastructure > Legacy > Archive")
  • Marcus gave up looking for the new hire onboarding checklist and just asked in #general again
  • The deployment runbook has 3 different versions and nobody knows which one is current

Data that matters:

  • 67% of #help-general questions are "where is the doc for X?"
  • Average time to find a doc: 12 minutes
  • People create duplicate docs because they can't find existing ones

What We're Building

Phase 1: Fix search basics (2 weeks)

  • Replace current search with Elasticsearch
  • Index all content properly (title, body, tags, recent edits)
  • Show actual relevant results instead of random garbage

Phase 2: Add helpful features (2 weeks)

  • Recent/popular docs widget on homepage
  • "Related docs" suggestions
  • Basic tagging system

That's it. No AI, no fancy graphs, no revolutionary knowledge management. Just search that works.

User Stories

As a developer, I want to:

  • Search "deploy" and see the actual deployment guide first
  • Find docs my team created without remembering exact titles
  • See when docs were last updated so I know if they're stale

As someone writing docs, I want to:

  • Tag docs so people can find them
  • See what people are searching for but not finding
  • Know if my doc is actually being used

Success Metrics

We'll know this worked if:

  • Search success rate >80% (currently ~30%)
  • #help-general wiki questions drop by half
  • People stop creating duplicate "how to deploy" docs

We'll measure:

  • Click-through rate on search results
  • Time from search to finding the right doc
  • Number of "doc not found, created new one" incidents

Technical Details

What we're using:

  • Elasticsearch for search indexing
  • Current auth system (no changes needed)
  • Existing wiki database (just better indexing)

Performance requirements:

  • Search results in <1 second
  • Reindex new docs within 5 minutes
  • Works on mobile (responsive design)

Risks & Mitigation

Risk: Elasticsearch is new to our team
Mitigation: Jordan has used it at previous job, will pair with backend team

Risk: Content migration breaks existing links
Mitigation: We're not moving content, just improving search indexing

Risk: People still won't find docs because docs are actually bad
Mitigation: Phase 2 includes analytics to identify which searches fail most

Dependencies

  • DevOps: Elasticsearch cluster setup (1 week)
  • Design: New search UI mockups (3 days)
  • Content team: Clean up obviously outdated docs before launch

What We're NOT Doing

  • AI-powered anything
  • Changing how people write or organize docs
  • New permissions or access controls
  • Integration with external tools
  • Mobile app

Timeline

Week 1: Elasticsearch setup + data migration
Week 2: Basic search implementation
Week 3: UI improvements + tagging
Week 4: Polish + launch

Launch plan: Soft launch to engineering team first, then company-wide if no major issues.


Open questions:

  • Should we archive docs that haven't been viewed in 2+ years?
  • Do we need to notify doc owners when their content isn't being found?

Next steps:

  • Jordan starts Elasticsearch spike Monday
  • Design review with Sarah Wednesday
  • Engineering kick-off Friday
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment