Thoughts about Managing Search Teams

Source: https://queryunderstanding.com/thoughts-about-managing-search-teams-7b3e9f1c4d2a Author: Daniel Tunkelang

Summary

Practical advice for leading search engineering teams. Tunkelang distills his experience into four core recommendations for search team management.

Recommendation 1: Combine Product and Engineering

Search is a product discipline, not just an engineering one. Teams that separate “search product” from “search engineering” create friction:

  • PMs without technical depth make poor trade-off decisions
  • Engineers without product intuition build the wrong things
  • Best teams have hybrid roles or extremely tight collaboration

Recommendation 2: Be Data-Driven

Search quality is measurable. Teams that operate without measurement are flying blind:

  • Define metrics before building features
  • Run A/B tests for every significant change
  • Build offline evaluation infrastructure early — it pays for itself
  • Don’t rely solely on anecdotes (“users told us search is bad”)

Recommendation 3: Incremental Execution

Avoid big-bang rewrites. Search systems are complex enough that:

  • Large changes have unpredictable interactions
  • Incremental improvements compound over time
  • Smaller experiments are easier to measure and roll back
  • “Ship, measure, iterate” beats “design, build, launch”

Recommendation 4: Holistic Approach

Search quality is end-to-end, not just the ranking model:

  • Indexing quality matters (what’s indexed, freshness)
  • Query understanding matters (how queries are parsed)
  • UI/UX matters (how results are presented)
  • Feedback loops matter (how signals are collected)
  • Teams that optimize one component in isolation miss cross-component leverage

Search PM Skills

A good search PM needs to understand:

  • How indexing and query parsing work
  • What the ranking features are and their trade-offs
  • How to design and interpret A/B tests
  • How to prioritize across relevance, latency, and UX

Key Concepts

  • Combined product-engineering — avoid org silos in search teams
  • Data-driven search — measurement and A/B testing as core practice
  • Incremental execution — prefer small experiments over big rewrites
  • Holistic search quality — end-to-end thinking across indexing, ranking, UX

People