4 Lessons from Crashing a 1700+ Person Livestream

Last week, my team at LinearB and I hosted a virtual conference for 1700+ engineering leaders. It nearly failed in the first 5 minutes.

We launched our Interact stream - and it crashed 🤦‍♂️

We had dev leaders from 95 countries in the chat and my team(particularly Adam Noble and Jackson Wells who deserve a huge shout-out) and I had spent 20+ hours that week fine-tuning the content and prepping every session. Yet the day of, the stream was skipping, crashing, and coming through incredibly grainy to the audience.

The next few minutes were hectic, urgently moving to our backup, ad-libbing on stream, and getting everything working. Here are 4 lessons I learned from the experience about hosting large live-streamed events, and crisis management when things go wrong:

1. Be adaptable
   We had a recorded intro ready to go. We'd meticulously edited our content and prepped timing, but when the stream went down, and we were searching for a solution before switching to the backup, we had to be agile - we took down the edited video, I threw off my cozy robe and jumped on camera to calm down the audience and let them know we were working to fix everything. It gave a human face to the situation and our audience was graceful about the screw-up and empathetic to our tech issues.

2. Have a backup plan (or two)
   The day before, we'd set up backups - if somehow the stream failed from our primary streamer, I had the content downloaded to my PC and a hardwired ethernet connection ready to go to upload it to the stream. If an earthquake had hit Seattle and put me out of commission, we had a team member in Chicago ready to go. If we hadn't created these backups, the problem would likely have taken much longer than 5 minutes to solve. Luckily, we were able to switch quickly to having me control the stream and share our content.

3. Always do a walk-through
   My colleague Margaret Shaeffer and I did a walk-through the day before so when things failed, we knew how to switch. But, this is also an area of improvement for us - we didn't stress test the stream experience adequately. Next time, we'll set up a copy of the event to test in advance with audience members from different locations to ensure quality, plus do more practice on making an emergency switch if needed.

4. Content is still king
   The first 5 minutes of the event were frankly, a disaster. It was 6am PT and I was wide awake from the adrenaline - and terrified that months of work were about to fall apart. But we managed the crisis, and once the content we'd worked so hard on was actually being delivered, the audience quickly forgot about the technical difficulties that had kicked off the event. Instead, they reveled in the panels with engineering leaders, laughed along with Tiff in Tech and Stereotype Breakers telling them what engineering leaders do that devs hate, and were glued to their seats for our Dev Interrupted podcast.

In the end, we delivered an incredible event - 6 hours of content with attendees staying for 3+ hours on average and giving us an overall rating of 8.9/10 💪

But it almost didn't happen.