Data Science at Twitch - CEO Perspective: Emmett Shear Interview
We recently caught up with Emmett Shear, CEO of Twitch. We were keen to learn more about his background, how data and Data Science have influenced Twitch's growth to this point, and what role they have to play going forward...
Hi Emmett, firstly thank you for the interview. Let's start with your background and how you became interested in data and the early days of Twitch/Justin.tv...
Q - What is your 30 second bio?
A - I've been doing startups my whole life. I started a now-forgotten calendar startup called Kiko Calendar right out of school with my friend Justin Kan, which we sold on eBay a year later. Then we started Justin.tv as a 24-7 experiment as a new kind of reality TV that we called "lifecasting", along with our two new co-founders Michael Seibel and Kyle Vogt. Justin.tv grew to be pretty big, but eventually our growth plateaued and we had to figure out what to do next.
I had the idea to pivot the company entirely and go after live video game streaming and became the CEO of Twitch, which we built out of Justin.tv gaming. I still run Twitch today, and also spend some time as a Y-Combinator part-time partner mentoring startups.
Q - How did you get interested in data?
A - I've always been interested in data. Even before starting companies, I loved "personal analytics" where I'd graph things happening in my own life. The hard part isn't having fun looking through data for patterns – it’s figuring out how to find data that can actually drive decisions.
Q - Can you talk about how data was used in the early days at Twitch/Justin.tv?
A - When we worked on Kiko and Justin.tv, we mostly just shot from the hip. While we occasionally referenced data in our decisions they were driven primarily by a product vision.
With Twitch, I did things very differently. From day 1 we've focused on data to guide our decision-making process. Some of the time, this means classic data science and crunching the numbers on user behavior in aggregate across millions of actions. Sometimes it means interviewing 10 key example broadcasters to understand their views of the world. Sometimes it means market research on how competitor's features are working.
This was a magical, transformative experience for me. Instead of guessing what users wanted, I actually knew the answer. We tended to jump to actual solutions for their problems instead of adding features that no one wanted or used.
Q - Was there a specific "aha" moment when you realized the power of data within Twitch/Justin.tv?
A - When we were first getting started working on gaming, I wanted to figure out what would be responsible for driving growth. I had the idea to go and see which streamers on the site were pulling in the most new users -- getting the word out about us. When we focused down on delivering value for those streamers, it caused a huge bump in our growth. And it was thanks to having real data and understanding that we knew who to talk to.
Q - What role did data have in deciding to create Twitch from Justin.tv?
A - Honestly, less than you'd think. I started Twitch because I was passionate about video game content and I thought there would be a big market for it, based on how many gamers were out looking for interesting things about video games on IGN and YouTube / Machinima. It wasn't a particularly data driven decision.
Very interesting background and context for where both you and Twitch are at today - thanks for sharing! Let's talk in more detail about the role Data Science can play at Twitch going forward...
Q - Which companies do you think Twitch should try to emulate with respect to their use of Data Science? Why?
A - I think mobile gaming companies are some of the sharpest in the world in their Data Science usage. I always learn something when I talk to a product manager or Data Scientist from one of those companies.
Q - Where can Data Science create most value in Online Gaming?
A - I'm not really an expert on gaming per-se. I know a lot about video and a little about gaming!
Q - Fair enough! Let's focus on Twitch ... How is Data Science used at Twitch now? Is this different from Justin.tv?
A - It's very different from Justin.tv. We've invested a huge amount more into Data Science now than we ever had before. The team at Twitch is 5 full-time members now and growing. We had zero full-time on Justin.tv.
It's also much more useful because we have dashboards internally which allow everyone in the company to get insights without having to do all the scripting and analysis themselves. That's been a huge shift.
Q - What are the biggest areas of opportunity/questions you want Data Science to tackle at Twitch?
A - It's easy to calculate things like Lifetime Value (LTV) in a retrospective way. Figuring out your predictive LTV for new users as they come in and understanding how changes impact that LTV is the holy grail.
Q - What projects has the Data Science team been working on this year, and why/how are they interesting?
A - We've started digging in on much more sophisticated questions recently. The newest one is working on clustering viewers to understand the different types of viewing habits in aggregate. Hopefully we'll have fun results to share on that front eventually!
Q - What has been the most surprising insight/development they have found so far?
A - I was surprised to learn how common it was for people to subscribe to a single channel one month, and then jump to another channel the next month. That accounts for a huge number of our subscriptions!
Q - Interesting! So, how does the Data Science team work with the rest of the organization? How would you describe their role?
A - The Data Science team owns and builds the backends and dashboards and notifications that allow the entire organization to understand the impact of their actions and make good predictions.
Q - And where/how are they integrated into decision making?
A - They're brought in on almost every product decision that gets made, because you can't make good decisions without understanding the data behind it. Usually very early in the process, as we're figuring out if we need more instrumentation and analysis, and how we're going to test for success.
Q - Which other groups/teams do they work with most closely? Why?
A - Product management for sure. That's where the data gets integrated with the rest of the business realities.
Q - Makes sense. You are currently hiring/building out the Data Science team at Twitch... What are you most looking for in candidates?
A - We need great engineers to build analytics systems, and we need great data scientists to make use of them. Both pieces are crucial.
Q - What is the most compelling reason why they should join Twitch?
A - It's incredibly impactful and challenging work. We deal with tens of billions of events per month, and the output of that analysis directly drives all our most important decisions.
Emmett, thanks so much for all the insights and details behind Twitch - sounds like a great time to be part of the Data Science team! Finally, one quick question on your role at YC...
Q - You've advised several batches of YC startup companies - how has their use of data/Data Science been evolving? What excites you most about these developments? What advice do you give in general and related to data?
A - More and more startups are actually actively tackling these problems as a service. I've seen several companies starting to produce automated tools to attack problems we've had to solve one-off at Twitch, which is really exciting for new companies that don't have the advantage of their own in-house Data Science team.
Emmett - Thank you so much for your time! Really enjoyed learning more about your background, how data and Data Science have influenced Twitch's growth to this point, and what role they have to play going forward. Twitch can be found online at http://www.twitch.tv and on twitter at @twitch.
Readers, thanks for joining us!
P.S.If you enjoyed this interview and want to learn more about
- what it takes to become a data scientist
- what skills do I need
- what type of work is currently being done in the field
then check out Data Scientists at Work - a collection of 16 interviews with some the world's most influential and innovative data scientists, who each address all the above and more! :)