Delivering High Quality Analytics at Netflix

Netflix is a data-driven entertainment company, where analytics are extensively used to make informed decisions on every aspect of the business. As such, when something goes wrong - and, inevitably, it will - it’s imperative that it’s quickly caught, communicated, and fixed.
In this talk, Michelle Ufford will share how the Data Engineering & Analytics team at Netflix does exactly that. She'll describe their analytics environment and how data is used across a variety of roles, then delve into how Netflix is tackling common analytics challenges affecting data quality.

[@1infiniteloop244]

For those who are not aware, user event is a nice word for tracking data

[@severtone263]

Impressive presentation! I was all in from start to finish.

[@rohitkumarz]

Was that a movie i just watched or a tech talk ... Loved it... I couldn't look away

[@jonassteinberg3779]

Exceedingly high level

[@AmarMTV]

Very good insights in terms of data infrastructure and strategies for checks on "unfortunate" data. But is Netflix working on getting rid of the "unfortunate" movies on the Netflix?

[@cantwaittowatch]

Great presentation, but I'm curious on how upstream analysis or what decision is taken with incomplete or incorrect data if unfortunate data is not streamlined into DW? That is, at frame 20:02, I hear the unfortunate data is not shown in reports.

[@jasonzhang6534]

Good talks, thanks for sharing

[@bernardstachman3507]

All this work and suggestions never surprise me with something good :(

[@forrestlinshi94]

Awesome presentation ! Thanks ! For the question in the speech, I think that Tiktok from ByteDance shall be a very good example of data driven. Such data driven growth shall be sort of basic or DNA of ByteDance. Am I right ? thanks

[@DavidMoreau]

Curious about the claim that 80% watch based on recommendations. Maybe it wasn't the case in 2017, but the home screen on Netflix if dominated by recommendations. I typically already know what I want to watch, but it is often an original being recommended. The recommendation has zero impact. What percentage of clicked recommendations are actually people being influenced by recommendations, and what percentage are merely people getting where they wanted to go with less clicks and scrolling?

[@nnemakhavhani-up1986]

Either something is wrong in the pipeline or we are the remaining 20%

[@jamespaz4333]

Which portal are using for data visualization?

[@xhinker]

Great talk!

[@ParasDoshiBlog]

Great talk, thanks Michelle!

[@brothermalcolm]

this is a goldmine

[@460enrique]

very good i dont undestand nothing by good keep it up (with the kardashians)

[@sahithkrishna5504]

Any source for the latest pipeline at Netflix?

[@karthikr6506]

Wonderful

[@Solivagus10]

Great talk - am curious how device is defined. One hundred million users on four thousand devices seems skewed.

[@tbgr]

4 000 different devices mean that you can find a Netflix App on 4 000 different devices, not only 4 000 devices total. A PS4 is a device, but you can have 1 000 000 users using their own PS4 to access Netflix. So 100 000 000 users, I think, is the number of user profiles (a family account can have 4 profiles for instance) that access Netflix. And those 100 000 000 users can connect to 1, 2 or maybe 3 devices during the same day : Phone in the subway, Computer at work during lunch break and then their PS4 in the evening.
@Netflix Data : maybe you can confirm such things ? :)

[@vlad1slavs]

Data is new oil!

[@jamespaz4333]

Nice intro!

[@Vincent-wd9fo]

2:00 pity for China

[@ak-gi3eu]

Lol

[@DanielPortugalRevilla]

[@fucking_ball]

I love woman that she made a cake 🍰😂

[@tdreamgmail]

Impressive but the net effect is still nothing good to watch on Netflix... lmao

[@jayeshmv]

Lol

[@tanyasaxena7968]

I dont like recommendations system , It will not allow anyone to watch other then what netflix are suggesting. like eating pizza everyday though in the market burgers , pastas are available.
Bullshit