DESIGNING MACHINE LEARNING · Tiny fraction of movies available to stream. Netflix circa 2007. not...
Transcript of DESIGNING MACHINE LEARNING · Tiny fraction of movies available to stream. Netflix circa 2007. not...
![Page 1: DESIGNING MACHINE LEARNING · Tiny fraction of movies available to stream. Netflix circa 2007. not enough data early product adoption curve changing user habits technical risk cannibal](https://reader036.fdocuments.net/reader036/viewer/2022071102/5fdb2a249a8bad2a844311ea/html5/thumbnails/1.jpg)
DESIGNING MACHINE LEARNINGA Multi-Disciplinary Approach
![Page 2: DESIGNING MACHINE LEARNING · Tiny fraction of movies available to stream. Netflix circa 2007. not enough data early product adoption curve changing user habits technical risk cannibal](https://reader036.fdocuments.net/reader036/viewer/2022071102/5fdb2a249a8bad2a844311ea/html5/thumbnails/2.jpg)
How might we improve Netflix’s user experience to
suit streaming services better?’
“Netflix Prize”guided investigation
![Page 3: DESIGNING MACHINE LEARNING · Tiny fraction of movies available to stream. Netflix circa 2007. not enough data early product adoption curve changing user habits technical risk cannibal](https://reader036.fdocuments.net/reader036/viewer/2022071102/5fdb2a249a8bad2a844311ea/html5/thumbnails/3.jpg)
Netflix circa 2007
![Page 4: DESIGNING MACHINE LEARNING · Tiny fraction of movies available to stream. Netflix circa 2007. not enough data early product adoption curve changing user habits technical risk cannibal](https://reader036.fdocuments.net/reader036/viewer/2022071102/5fdb2a249a8bad2a844311ea/html5/thumbnails/4.jpg)
Netflix circa 2007
1. A booming DVD business with predictable yet slowing growth.
2. Signals that internet content delivery will grow exponentially, though initially tiny market.
3. Data about users’ movie preferences. 4. Tiny fraction of movies available to stream.
![Page 5: DESIGNING MACHINE LEARNING · Tiny fraction of movies available to stream. Netflix circa 2007. not enough data early product adoption curve changing user habits technical risk cannibal](https://reader036.fdocuments.net/reader036/viewer/2022071102/5fdb2a249a8bad2a844311ea/html5/thumbnails/5.jpg)
Netflix circa 2007
![Page 6: DESIGNING MACHINE LEARNING · Tiny fraction of movies available to stream. Netflix circa 2007. not enough data early product adoption curve changing user habits technical risk cannibal](https://reader036.fdocuments.net/reader036/viewer/2022071102/5fdb2a249a8bad2a844311ea/html5/thumbnails/6.jpg)
not enough data
early product
adoption curve
changing user habits
technical risk
cannibal business
![Page 7: DESIGNING MACHINE LEARNING · Tiny fraction of movies available to stream. Netflix circa 2007. not enough data early product adoption curve changing user habits technical risk cannibal](https://reader036.fdocuments.net/reader036/viewer/2022071102/5fdb2a249a8bad2a844311ea/html5/thumbnails/7.jpg)
Measuring?
Delivering?
Process??
Argue!
Goals?
Opportunities?
![Page 8: DESIGNING MACHINE LEARNING · Tiny fraction of movies available to stream. Netflix circa 2007. not enough data early product adoption curve changing user habits technical risk cannibal](https://reader036.fdocuments.net/reader036/viewer/2022071102/5fdb2a249a8bad2a844311ea/html5/thumbnails/8.jpg)
An intersting twist…
![Page 9: DESIGNING MACHINE LEARNING · Tiny fraction of movies available to stream. Netflix circa 2007. not enough data early product adoption curve changing user habits technical risk cannibal](https://reader036.fdocuments.net/reader036/viewer/2022071102/5fdb2a249a8bad2a844311ea/html5/thumbnails/9.jpg)
An intersting twist…
![Page 10: DESIGNING MACHINE LEARNING · Tiny fraction of movies available to stream. Netflix circa 2007. not enough data early product adoption curve changing user habits technical risk cannibal](https://reader036.fdocuments.net/reader036/viewer/2022071102/5fdb2a249a8bad2a844311ea/html5/thumbnails/10.jpg)
Using insights from stakeholder interviews, data, and team research,
design a strategy for Netflix’s transition to a streaming movie platform.
Your Second Guided Investivation:
![Page 11: DESIGNING MACHINE LEARNING · Tiny fraction of movies available to stream. Netflix circa 2007. not enough data early product adoption curve changing user habits technical risk cannibal](https://reader036.fdocuments.net/reader036/viewer/2022071102/5fdb2a249a8bad2a844311ea/html5/thumbnails/11.jpg)
You are not limited to 2007 technology.
What we Expect
![Page 12: DESIGNING MACHINE LEARNING · Tiny fraction of movies available to stream. Netflix circa 2007. not enough data early product adoption curve changing user habits technical risk cannibal](https://reader036.fdocuments.net/reader036/viewer/2022071102/5fdb2a249a8bad2a844311ea/html5/thumbnails/12.jpg)
You are not limited to 2007 technology.
We want you to think highly creatively.
What we Expect
![Page 13: DESIGNING MACHINE LEARNING · Tiny fraction of movies available to stream. Netflix circa 2007. not enough data early product adoption curve changing user habits technical risk cannibal](https://reader036.fdocuments.net/reader036/viewer/2022071102/5fdb2a249a8bad2a844311ea/html5/thumbnails/13.jpg)
You are not limited to 2007 technology.
We want you to think highly creatively.
You may focus on any aspect of the problem: the user experience, the interface,
the business strategy, etc.
What we Expect
![Page 14: DESIGNING MACHINE LEARNING · Tiny fraction of movies available to stream. Netflix circa 2007. not enough data early product adoption curve changing user habits technical risk cannibal](https://reader036.fdocuments.net/reader036/viewer/2022071102/5fdb2a249a8bad2a844311ea/html5/thumbnails/14.jpg)
You can ‘cheat’ and look at what Netflix did to their product in the transition to streaming (they did some pretty creative things!), but the streaming platform
was still noticably incomplete.
What we Expect
![Page 15: DESIGNING MACHINE LEARNING · Tiny fraction of movies available to stream. Netflix circa 2007. not enough data early product adoption curve changing user habits technical risk cannibal](https://reader036.fdocuments.net/reader036/viewer/2022071102/5fdb2a249a8bad2a844311ea/html5/thumbnails/15.jpg)
Many platforms share similarities with Netflix — how do they create good UX?
Starting Points
![Page 16: DESIGNING MACHINE LEARNING · Tiny fraction of movies available to stream. Netflix circa 2007. not enough data early product adoption curve changing user habits technical risk cannibal](https://reader036.fdocuments.net/reader036/viewer/2022071102/5fdb2a249a8bad2a844311ea/html5/thumbnails/16.jpg)
Many platforms share similarities with Netflix — how do they create good UX?
What are Netflix’s product objectives? Do they actually benefit from higher engagement?
Starting Points
![Page 17: DESIGNING MACHINE LEARNING · Tiny fraction of movies available to stream. Netflix circa 2007. not enough data early product adoption curve changing user habits technical risk cannibal](https://reader036.fdocuments.net/reader036/viewer/2022071102/5fdb2a249a8bad2a844311ea/html5/thumbnails/17.jpg)
Many platforms share similarities with Netflix — how do they create good UX?
What are Netflix’s product objectives? Do they actually benefit from higher engagement?
Recommendations were a huge differentiator for Netflix, but also posed a big risk. Why?
Starting Points
![Page 18: DESIGNING MACHINE LEARNING · Tiny fraction of movies available to stream. Netflix circa 2007. not enough data early product adoption curve changing user habits technical risk cannibal](https://reader036.fdocuments.net/reader036/viewer/2022071102/5fdb2a249a8bad2a844311ea/html5/thumbnails/18.jpg)
Starting Points
![Page 19: DESIGNING MACHINE LEARNING · Tiny fraction of movies available to stream. Netflix circa 2007. not enough data early product adoption curve changing user habits technical risk cannibal](https://reader036.fdocuments.net/reader036/viewer/2022071102/5fdb2a249a8bad2a844311ea/html5/thumbnails/19.jpg)
Starting Points
![Page 20: DESIGNING MACHINE LEARNING · Tiny fraction of movies available to stream. Netflix circa 2007. not enough data early product adoption curve changing user habits technical risk cannibal](https://reader036.fdocuments.net/reader036/viewer/2022071102/5fdb2a249a8bad2a844311ea/html5/thumbnails/20.jpg)
If last week was about the value of ethnography, this week is about the value of lo-fi prototyping.
Make Lo-Fi Prototypes!
![Page 21: DESIGNING MACHINE LEARNING · Tiny fraction of movies available to stream. Netflix circa 2007. not enough data early product adoption curve changing user habits technical risk cannibal](https://reader036.fdocuments.net/reader036/viewer/2022071102/5fdb2a249a8bad2a844311ea/html5/thumbnails/21.jpg)
Team 1: Katia Teran, Tyler Su, Vikram Jindal
Team 2: Zoe Weinberg, Andrew Huang, Farid Soroush
Team 3: Abhishek Garg, King Alandy Dy, Erica Pincus
Team 4: Barr Yaron, Daniel Levine, Marcy Regaldo
Team 5: Tianxing Ma, James Liu, Ian Taylor
Team 6: Angelica Willis, Gal Ron, Alexander Maschmedt
Project Groups