Choosing Content for Netflix: How Data Leads The Way Paul Ellwood VP, Data Engineering & Analytics
Choosing Content for Netflix:
How Data Leads The WayPaul Ellwood
VP, Data Engineering & Analytics
Today’s TalkEvolution of NetflixData@NetflixChoosing Content for Netflix
Evolution of NetflixA Brief History Lesson
DVD Service1999 DVD Service2007 Streaming Service2012 Netflix Originals 2016 Netflix Global
Streaming Service1999 DVD Service2007 Streaming Service2012 Netflix Originals 2016 Netflix Global
1999 DVD Service2007 Streaming Service2012 Netflix Originals 2016 Netflix Global
Netflix Originals1999 DVD Service2007 Streaming Service2012 Netflix Originals 2016 Netflix Global
2012 2013 2014 2015 2016
+ 3 more + 2 more + 20 more + 30 or so more
Netflix Originals1999 DVD Service2007 Streaming Service2012 Netflix Originals2016 Netflix Global
Netflix Studios1999 DVD Service2007 Streaming Service2012 Netflix Originals 2016 Netflix Global
Local Originals1999 DVD Service2007 Streaming Service2012 Netflix Originals 2016 Netflix Global
#netflixeverywhere1999 DVD Service2007 Streaming Service2012 Netflix Originals 2016 Netflix Global
Netflix in Hollywood1999 DVD Service2007 Streaming Service2012 Netflix Originals 2016 Netflix Global
Data@Netflix
Big Data
>40PB700
B
Experimentation, Not Opinion
2013
2015
User Experience
BillboardEvidenceImageryRow / Content SelectionSearch Results… and more
Recommender Systems
The Netflix Recommender System: Algorithms, Business Value, and Innovation
http://techblog.netflix.com/2016/05/selecting-best-artwork-for-videos.html
Cover Art Optimization
https://media.netflix.com/en/company-blog/how-netflix-works-with-isps-around-the-globe-to-deliver-a-great-viewing-experience
Content Delivery
http://techblog.netflix.com/2015/12/optimizing-content-quality-control-at-netflix-predictive-modeling.html
Quality Control
But What About Choosing Content?
Can a computer write a script that would win a competition? [benjamin.wtf]
Can Data Science Create Content?
Not So Well (Yet)...
Is Netflix Doing This?
No. It’s the opposite.
We give creative freedom to the creatives.
But Can Data Help in Choosing Content?
Yes.All decisions are made by experienced
creatives, but analytic products can help.
Value Attribution On a Subscription Service
Valuing Content on Linear TV
How Should Netflix Value Content?
Many possibilities…● Hours viewed?● Completed watching?● New subscribers?● Award nominations? Wins?● Buzz?Or All Of The Above?
Content Efficiency = value / cost
Applying Netflix’s Notion of Value
Value PredictionFor Licensed Content
Predict Value For Licensed Content
Available Titles
DemandFeature
s
DemandPredictiv
eModel
Adjust If efficient, license
Predicting Value For Licensed Content
E.g.● Past performance on Netflix (if previously
licensed)● Past performance of similar titles on Netflix● Broadcast ratings● Box Office● Talent involved● Critic & User Reviews● Awards
Available Titles
DemandFeature
s
DemandPredictiv
eModel
Adjust If efficient, license
Predicting Value For Licensed Content
Available Titles
DemandFeature
s
DemandPredictiv
eModel
Adjust If efficient, license
Predicting Value For Licensed Content
E.g.● Buyer judgements● Deal term adjustments
Available Titles
DemandFeature
s
DemandPredictiv
eModel
Adjust If efficient, license
Predicting Value For Licensed Content
Available Titles
DemandFeature
s
DemandPredictiv
eModel
Adjust If efficient, license
Value PredictionFor Original Content
Predicting Value for Original Content
Originals is a more difficult problem
● Less data - no box office or reviews
● Moving target - ideas and scripts can evolve
● Fungible - execution varies with talent and budget
‘House of Cards’ script = X
Those members also watch:
Finding Comparable Titles (that we know more about)
Catalog Optimization
Programming Linear TVDemographic Based
Programming Netflix
Something for Everyone
Global Complexity
What’s Next?
Upcoming Challenges
•Choose the best mix of content globally•Identify valuable content earlier•Better analytic tools for buyers & creatives•Understand marketplace dynamics