The situation
Teams was growing faster than almost any product Microsoft had shipped. Search was still built like a keyword lookup over messages and files. As organizations moved their whole working day into Teams, “find that thing someone sent me” became one of the most frequent, and most frustrating, things people did.
What I owned
- Product for Teams search: the results experience, relevance, autosuggest and personalization.
- Bootstrapping a Teams ML platform so ranking and personalization weren’t rebuilt feature by feature.
- Leading a team of 12 PMs across Search and Mobile, partnering with 60+ engineers and designers.
Key decisions
Set the charter: Teams search as the front door to M365 knowledge. Rather than build a Teams-only index, I re-platformed Teams search onto the M365 index and substrate and aligned the Teams and M365 Search organizations behind it. That made Teams the place people searched across all their work, not just their chats.
Relevance over recency. We moved search from chronological, keyword-matched results to a modern results page ranked by relevance. This was a bet that users would trust a ranked list more than a strict timeline, and we measured it on session success rather than clicks. It raised search session success by 1.7%, which at Teams’ scale is a very large number of people finding what they needed.
Build signals once, use them everywhere. Rather than each feature inventing its own notion of “who matters to this user,” I bootstrapped a shared ML platform with affinity scores: user-to-user and user-to-channel. Those signals fixed autosuggest, improved personalization in search, and became a foundation other teams could build on.
Leading mobile
For my last two years I also led the PMs for Teams mobile, making it a first-class productivity app across work, school and home, including experiences such as calendar and feeds. The job shifted from making the best call myself to making sure a team made good calls without me: setting the bar, coaching on trade-offs, and keeping strategy aligned across engineering, design and the wider Teams org.
What I took from it
Early-stage products run on customer observation, not dashboards. And at hypergrowth scale the highest-leverage product work is often infrastructure for other teams: a shared signal or platform beats ten clever features.