Teams Live Events

Scaling enterprise broadcasting to 20 million users

Watch me present this work at Microsoft Ignite

The situation

In 2018, Microsoft was losing the enterprise broadcasting market. Skype Meeting Broadcast existed but felt like an afterthought - clunky, unreliable, and disconnected from the Teams experience that was gaining momentum. Meanwhile, competitors were making inroads with simpler solutions.

The stakes were high. Enterprise customers were asking for a way to broadcast company all-hands, training sessions, and executive communications to thousands of employees simultaneously. If we couldn't deliver, they'd find someone who could.

The complication: three separate teams across Microsoft were working on overlapping solutions. Stream, Teams, and Skype each had their own vision for what live events should look like. No one was coordinating, and customers were getting confused.

What I did

Defined the product through customer immersion

I spent the first month talking to customers - IT admins, corporate communications teams, executives who hosted town halls. I wanted to understand not just what they were asking for, but why the existing solutions weren't working.

The insight: most "live events" aren't really live. They're carefully produced broadcasts that need to feel professional. Customers didn't want a video call with a big audience - they wanted a broadcast studio that happened to live inside Teams.

Unified three competing teams

I mapped out what each team was building and found the overlaps. Instead of three competing products, I proposed a single architecture: Teams would be the interface, Stream would handle video delivery at scale, and we'd deprecate the Skype approach entirely.

Getting alignment wasn't about being right - it was about making it safe for each team to contribute their expertise without losing ownership. I created a shared roadmap where each team had clear responsibilities and could see how their work fit into the whole.

Solved the scaling constraint creatively

The technical challenge was real: how do you deliver low-latency video to 20,000 concurrent viewers without melting your infrastructure? Traditional approaches would've required massive investment in CDN capacity we didn't have.

We used a tiered architecture. Producers and presenters got real-time two-way video through Teams. The audience got a 15-20 second delayed stream through Azure Media Services and CDN. This trade-off - accepting slight delay for massive scale - let us support audiences of tens of thousands without custom infrastructure.

Built the quality and reliability muscle

Live events have zero tolerance for failure. If your CEO's quarterly all-hands goes down mid-stream, you've got a very bad day ahead. I implemented weekly "war room" reviews of every failed event, no matter how small. We built automated monitoring that could detect problems before they became visible to users.

Over time, our incident rate dropped by 80%. More importantly, when things did go wrong, we caught them faster and recovered more gracefully.

Then COVID happened

In March 2020, everything changed. Suddenly every company on earth needed to broadcast to remote employees. Our traffic spiked 10x in two weeks.

The architecture we'd built - designed for enterprise scale - held up. But the user base shifted. We weren't just serving Fortune 500 IT departments anymore. Teachers were using Live Events to broadcast lessons. Small businesses were running virtual town halls. Non-profits were hosting donor events.

I noticed that educators were the fastest-growing segment, but our product wasn't built for them. They didn't have dedicated producers or IT support. They needed something simpler. This insight led to a rapid iteration cycle focused on making the product more accessible to non-enterprise users - simplified setup flows, better defaults, and clearer documentation.

The results

By 2021, Teams Live Events was reaching 20 million monthly users. We'd gone from a struggling also-ran to the default choice for enterprise broadcasting in the Microsoft ecosystem.

Key outcomes:

• Scaled from thousands to millions of monthly active users
• Supported events with 20,000+ concurrent attendees
• Reduced incident rate by 80% through systematic quality focus
• Expanded beyond enterprise to education and small business segments
• Unified three competing product visions into a single coherent platform

What I learned

Customer immersion beats customer research. Talking to customers for a month before writing a single spec changed everything. The best insights came from watching people use (and struggle with) existing solutions, not from asking what they wanted.

Alignment is a product, not an event. Getting three teams to work together wasn't about one convincing meeting. It was about creating ongoing structures - shared roadmaps, clear ownership, regular syncs - that made collaboration the path of least resistance.

Trade-offs are features, not compromises. The 15-20 second delay we accepted for scale wasn't a limitation - it was a deliberate choice that unlocked capabilities we couldn't have had otherwise. Good product thinking means being explicit about what you're trading and why.

Build for the crisis you can't predict. We didn't know COVID was coming. But because we'd built for scale and reliability, we were ready when the moment came. The best product decisions are the ones that hold up under conditions you didn't anticipate.

See all work · Read about Snapchat · Read about Segment · Email me