The problem
A search team fails in two opposite ways. The lead becomes the only person who can change relevance, or the lead steps so far back that nobody owns the answer the user gets.
The approach
I came up as a Solr developer and a staff engineer, and I now lead engineers on a search and matching platform. The published book is the sequence I want a new manager to have: listen first, earn one visible win, then build a team that runs without the hero.
The conclusion
Your output becomes the output of the team. That is also how you keep a search platform honest: more than one person can explain why the top result is the top result.