DATA @ NFLX 
Building a Culture of Analytics Everywhere 
Tableau Customer Conference 
2014.09.09 
Blake Irvine 
Manager, Device Analytics 
Data Science & Engineering 
birvine@netflix.com 
Albert Wong 
Manager, Reporting Platforms 
Cloud & Platform Engineering 
albwong@netflix.com
Netflix and data in the news... 
“Giving Viewers What They 
Want” --New York Times 
“The Science Behind the 
Netflix Algorithms That 
Decide What You’ll Watch 
Next” --Wired 
Data-Mining Boosts Netflix's 
Subscriber Base, Showbiz 
Clout 
--AdAge
BIG 
DATA
Big Data at Netflix 
Size 
● 50+ million members 
● 1000’s of devices 
● 100’s of systems 
● >300B data pipeline events daily 
● >10B row tables daily 
Ubiquitous 
● Data is everywhere 
● Many complex systems 
● Many engineering teams 
producing and consuming 
● Non-streaming teams produce 
and consume data 
● Culturally data driven
How do we innovate 
with Big Data?
TOOLS 
CULTURE
Tools 
DATA STORAGE DATA PROCESSORS DATABASE REPORTING 
Sting
Team Structure 
Data Science and Engineering 
Marketing 
data engineering 
reporting 
analyst 
Finance Product Engineering ... 
Business Functions
Team Structure 
Data Science and Engineering 
WE DO 
NOT WHAT Marketing 
data engineering 
reporting 
analyst 
Finance Product Engineering ...
Netflix Team Structure 
Data Science and Engineering 
data engineering 
reporting 
analyst 
Marketing 
data engineering 
reporting 
analyst 
data engineering 
reporting 
analyst 
data engineering 
reporting 
analyst 
Finance Product Engineering ... 
Business Functions
Highly Aligned, Loosely Coupled 
data engineering 
reporting 
analyst 
Finance 
data engineering 
data architect 
analyst/reporting 
Marketing
Highly Aligned, Loosely Coupled 
data engineering 
reporting 
analyst 
Finance 
data engineering 
data architect 
analyst/reporting 
Marketing
Freedom & Responsibility 
Freedom 
Don’t limit access 
Don’t limit choices 
Reduce constraints 
Responsibility 
Trust 
Don’t allow chaos 
Reduce accidents
Protected access 
CENTRAL DATA
Protected access 
CENTRAL DATA NOT WHAT WE DO
Unlocked access 
CENTRAL DATA
Don’t limit choice 
CENTRAL DATA OPERATIONAL DATA 
LOCAL DATA
OPERATIONAL DATA 
Don’t limit choice 
Sting 
CENTRAL DATA LOCAL DATA
Why is Tableau a good choice? 
Quick 
Intuitive 
Rich Visual Analysis 
Storytelling 
Emailed Reporting 
Reusability
Who uses Tableau?
How do we use Tableau?
Examples 
● Application build testing 
● Certification tracking 
● Operational Excellence
Application Build Testing (1/4) 
● Team: Product Engineering 
● Context 
○ Application automatically tested at every code 
checkin 
○ Several dozen performance tests run to measure 
change and avoid regression 
● Problem 
○ Limited graphing tool built into test tool 
○ Difficult / no customization
Application Build Testing (2/4) 
{ 
"metadata" : { 
"TestCaseName" : "Trunk.Rendering.Effects_Mask2", 
"MarkerSetId" : 2472165, 
"ESN" : "DCQA01", 
"UIBuild" : null, 
"Build" : "2689", 
"JenkinsJob" : "http://builds.netflix.com/job/208/", 
"Label" : "#2689 / 208", 
"BuildTimestamp" : null, 
"Changelist" : "2177893" 
}, 
"results" : 
[ 
{"MeanFps" : 60.284862537264004}, 
{"MeanFps" : 60.264900662251655}, 
{"MeanFps" : 60.234541577825162} 
] 
} 
Build / Test
Application Build Testing (3/4)
Application Build Testing (4/4)
Certification Tracking (1/3) 
Team: Certification Operations 
Context 
● We certify the Netflix implementation on many new 
consumer electronics devices 
Problem 
● Time consuming to generate insights across multiple 
disconnected systems
Certification Tracking (2/3) 
NTS 
Certification 
Process
Certification Tracking (3/3)
Operational Excellence (1/3) 
Team: Data Science & Engineering 
Context 
● Ensure continuous development does not negatively 
impact availability and resilience 
Problem 
● Multiple programs and data sources 
● Need to link source data patterns to engineering tools
Operational Excellence (2/3)
Operational Excellence (3/3)
Where are we with 
Tableau?
DATA @ NFLX 
● Netflix is known for being data driven 
● Big data is available everywhere 
● Our culture enables analysis everywhere 
● Tableau complements our culture 
● We have organic growth throughout Netflix 
● Growing part of our reporting platform
What can we answer? 
Blake Irvine 
Manager, Device Analytics 
Data Science & Engineering 
birvine@netflix.com 
Albert Wong 
Manager, Reporting Platforms 
Cloud & Platform Engineering 
albwong@netflix.com

DATA @ NFLX (Tableau Conference 2014 Presentation)

  • 1.
    DATA @ NFLX Building a Culture of Analytics Everywhere Tableau Customer Conference 2014.09.09 Blake Irvine Manager, Device Analytics Data Science & Engineering birvine@netflix.com Albert Wong Manager, Reporting Platforms Cloud & Platform Engineering albwong@netflix.com
  • 5.
    Netflix and datain the news... “Giving Viewers What They Want” --New York Times “The Science Behind the Netflix Algorithms That Decide What You’ll Watch Next” --Wired Data-Mining Boosts Netflix's Subscriber Base, Showbiz Clout --AdAge
  • 7.
  • 8.
    Big Data atNetflix Size ● 50+ million members ● 1000’s of devices ● 100’s of systems ● >300B data pipeline events daily ● >10B row tables daily Ubiquitous ● Data is everywhere ● Many complex systems ● Many engineering teams producing and consuming ● Non-streaming teams produce and consume data ● Culturally data driven
  • 9.
    How do weinnovate with Big Data?
  • 10.
  • 11.
    Tools DATA STORAGEDATA PROCESSORS DATABASE REPORTING Sting
  • 12.
    Team Structure DataScience and Engineering Marketing data engineering reporting analyst Finance Product Engineering ... Business Functions
  • 13.
    Team Structure DataScience and Engineering WE DO NOT WHAT Marketing data engineering reporting analyst Finance Product Engineering ...
  • 14.
    Netflix Team Structure Data Science and Engineering data engineering reporting analyst Marketing data engineering reporting analyst data engineering reporting analyst data engineering reporting analyst Finance Product Engineering ... Business Functions
  • 15.
    Highly Aligned, LooselyCoupled data engineering reporting analyst Finance data engineering data architect analyst/reporting Marketing
  • 16.
    Highly Aligned, LooselyCoupled data engineering reporting analyst Finance data engineering data architect analyst/reporting Marketing
  • 17.
    Freedom & Responsibility Freedom Don’t limit access Don’t limit choices Reduce constraints Responsibility Trust Don’t allow chaos Reduce accidents
  • 18.
  • 19.
    Protected access CENTRALDATA NOT WHAT WE DO
  • 20.
  • 21.
    Don’t limit choice CENTRAL DATA OPERATIONAL DATA LOCAL DATA
  • 22.
    OPERATIONAL DATA Don’tlimit choice Sting CENTRAL DATA LOCAL DATA
  • 23.
    Why is Tableaua good choice? Quick Intuitive Rich Visual Analysis Storytelling Emailed Reporting Reusability
  • 24.
  • 25.
    How do weuse Tableau?
  • 26.
    Examples ● Applicationbuild testing ● Certification tracking ● Operational Excellence
  • 27.
    Application Build Testing(1/4) ● Team: Product Engineering ● Context ○ Application automatically tested at every code checkin ○ Several dozen performance tests run to measure change and avoid regression ● Problem ○ Limited graphing tool built into test tool ○ Difficult / no customization
  • 28.
    Application Build Testing(2/4) { "metadata" : { "TestCaseName" : "Trunk.Rendering.Effects_Mask2", "MarkerSetId" : 2472165, "ESN" : "DCQA01", "UIBuild" : null, "Build" : "2689", "JenkinsJob" : "http://builds.netflix.com/job/208/", "Label" : "#2689 / 208", "BuildTimestamp" : null, "Changelist" : "2177893" }, "results" : [ {"MeanFps" : 60.284862537264004}, {"MeanFps" : 60.264900662251655}, {"MeanFps" : 60.234541577825162} ] } Build / Test
  • 29.
  • 30.
  • 31.
    Certification Tracking (1/3) Team: Certification Operations Context ● We certify the Netflix implementation on many new consumer electronics devices Problem ● Time consuming to generate insights across multiple disconnected systems
  • 32.
    Certification Tracking (2/3) NTS Certification Process
  • 33.
  • 34.
    Operational Excellence (1/3) Team: Data Science & Engineering Context ● Ensure continuous development does not negatively impact availability and resilience Problem ● Multiple programs and data sources ● Need to link source data patterns to engineering tools
  • 35.
  • 36.
  • 37.
    Where are wewith Tableau?
  • 38.
    DATA @ NFLX ● Netflix is known for being data driven ● Big data is available everywhere ● Our culture enables analysis everywhere ● Tableau complements our culture ● We have organic growth throughout Netflix ● Growing part of our reporting platform
  • 39.
    What can weanswer? Blake Irvine Manager, Device Analytics Data Science & Engineering birvine@netflix.com Albert Wong Manager, Reporting Platforms Cloud & Platform Engineering albwong@netflix.com