Our Work
Twitch's Science team is a small group within Twitch charged with helping the company make effective product decisions driven by rigorous research.
Our work powers decision-making at all levels of the company. Check out some of our recent projects!
-
In Silicon Valley, everyone makes bets. Here, predicting the future is not a theoretical superpower, it’s part of the job. -
Using noisy signals from chat to help us understand what captivates viewers. -
A case study in how we approach tough causal questions about our product. -
A key function of data science at Twitch is using behavioral data to build data products. -
How Twitch performs A/B testing when using Google Play's staged rollout feature. -
The evidence behind our culture of forecasting. -
TwitchCon 2016 Talk: Ever wonder how Twitch uses Science to make decisions? Drew, Steve and Ian explain! -
We attribute 25% of Punch Club and The Culling’s sales to Twitch. And we attribute increases in week over week player retention as high as 15% to Twitch -
Melissa interviews Drew about how Data Science works at Twitch, including onboarding, mentorship, the role of qualitative work, and collaboration. -
What is data science? How is it different from adjacent disciplines like statistics, product management, and engineering? -
What makes a data team effective? Drew proposes 14 simple questions that help identify if a team is set up for success and able to focus on high leverage data science work. -
Recently watched channels get more viewership than all our existing discovery mechanisms before it was ever surfaced! -
20% Fewer unsubscribes from our communications, no downside. Inboxes throughout the land rejoice! -
What PhDs do wrong and right when applying for Data Science jobs - our advice to people making the transition into a Data Science role from academia or elsewhere. -
A conversation with Mixpanel about how our team thinks about having an impact at Twitch and crowd-pleasing topics like documentation, trusting your research team, and marketing your results. -
Finding something great to watch on Twitch isn't always easy. We identify some major hurdles in content discovery and share our mental model for how people discover and move between streams. -
Host mode is a powerful tool for broadcasters to support each other when they're not online. Who do they choose to host - channels less popular than them, about the same popularity, or channels that are more popular? -
Twitch has a diverse community that spans games, languages, and continents. In this project we explore how channels are connected through their shared audiences.
