#30Days30Companies : How Facebook Uses Data Analytics
Some unpopular facts about Facebook
Every year when our birthday comes, you know only a few person remembers our birth date and wishes us on that exact day. For the most of us they are our mom, probably our father, then maybe our best friend. But one person who always, like literally always remembers our birth date and wishes us, is Facebook.
No matter which year it is, how old I am, where I am, with whom I am - Facebook will wish us; not only birthday wishes, but it also remembers where we were a year ago, or whom we were with, or what we were doing ( definitely on the info we share). It might happen that you even forgot any past event, but Facebook never!
Our all time friend, Facebook!
But, well yes, there is a BUT, which is how Facebook remembers all of these things? To be specific, it’s our data that we share with Facebook. But then (again, a but came!) how Facebook uses our data, well for that, please continue this journey...
What is Facebook?
Dude, seriously! 😑 Is this even necessary to explain what is Facebook? Bro, everybody knows that Facebook is a social networking site, which simply brings people closer together. Well, in fact their mission statement is: "Give people the power to build community and bring the world closer together."
And who haven’t yet watched ‘The Social Network’ movie, it’s the story of Mark Zuckerberg, the founder of Facebook; Facebook is the biggest social media platform connecting almost 2.6 billion peoples(monthly) around the globe.
Please tell us something new and exciting. Don’t bore us with the stuffs we already know.
Okay okay! Look, I actually like to keep my blogs in a proper manner. Moving on…
Why Facebook uses data analytics?
A company whose worldwide monthly number of user is 2.6 billion, they definitely generate an enormous amount of data every single day; to be precise, it is almost 4 petabytes of data per day! Besides, the data is not only in the form of our name, email, phone no., or any kind of text based data; but also in the form of photos and videos.
Well, to manage this amount of data, only way anyone have, is to use data analytics. Cause it’s really a large amount of data - Big Data.
I want to add some point here.
Okay, please share with us!
You know, this huge amount of data is stored into Hive, Facebook’s data warehouse; it stores almost 300 petabytes of data in 800,000 tables.
Wow, that’s an incredible amount of data! Thanks for sharing with us.
How Facebook uses data analytics to understand our posts/data and even recognize our face?
Whenever we open Facebook and see something, or maybe like some post, or comment on it, or share it, or visit any page, or join some group, or maybe search something or someone - Facebook tracks it all. No matter who we are, whom we are friends with, who we are talking to, where we spend most of our time, what we like, what we dislike - Facebook knows it all.
According to Daniel Newman in a Forbes article, “Facebook Knows Us Better Than Our Therapist.”
Well, the main problem for Facebook’s data scientists is most of the data stored is in unstructured manner. That’s because most of the data is generated is through the photos or stories or videos we share and it’s really complicated to understand them. And here comes Deep Learning into the game.
So what actually is Deep Learning ? Deep learning(DL) is an approach to AI(Artificial Intelligence); a subset of Machine Learning. It’s been long since it’s here and is really showing great results when it comes to development of some autonomous, self-teaching systems which is literally evolving the modern industries. For example, Google is using it in its voice and image recognition algorithms, Netflix and Amazon is using it to decide what you want to watch or buy next, and researchers at MIT to predict the future.
DL plays a very important role in Facebook, as it helps it to give structure to unstructured data, by quantifying it and representing it in a form from which analytical tools can derive insights. It also have the capability to recognize an image which contain cats, without specifically being told what a cat looks like. And as DL algorithms become more sophisticated, they can increasingly be applied to more data that we share, from text to pictures to videos. Some of the applications Facebook uses with the help of DL:
Textual analysis
Facial recognition
Targeted advertising
Designing AI applications
What analytics tools does Facebook use?
As the time is passing, we are inventing new and newer technologies. Everything we used the last year, probably is getting exchanged by some modern version of it. For example, landlines is now replaced by mobile; letters are replaced by emails; CRT televisions are now replaced by LCD or LED TVs. And that’s been same for Facebook either. They need to develop and upgrade their methods and algorithms to tackle this data which is increasing every single day.
The most part of any social media strategy involves monitoring metrics and measuring the performance of our posts. And to achieve those goals Facebook has different kind of tools, which we just knew in the above section - Text analysis, facial recognition, targeted advertising, different AI applications, and many more.
The most portion of data where Facebook gets the most amount of insights from, is the text based data. Definitely the videos we post are of much bigger qualities and can contain many, many lines of text, even a single photo can contain a lot of lines of text; but for answering a simple question with a yes or no, or with just a single word - one sentence is enough. Also, storing those data which doesn’t involve any insights and still be large in size is not a efficient thing to do. That’s why text based data is one of the main source of insights. For this, Facebook use a tool named DeepText, developed by itself.
What DeepText does is it analyze the posts we share and learn from them. As it is build upon DL, means there are some neural networks as they are the heart of deep learning, and these helps the tool to understand the meaning between words and how the meaning varies from word to word. This means it also understands variations in spelling, slang, and even different languages. The way we humans connect with each other, speak with each other ( maybe with a slang or using ‘bro’ or ‘brother’, anything) - it understands it all.
The facial recognition in Facebook is done by DeepFace; again built on DL. Facebook claims their method reaches an accuracy of 97.35% on the Labeled Faces. With the help of it they detect any person from our photos who’s also having a Facebook account and tell us to tag them as well.
Then comes the targeted advertising, also built on DL and neural networks. This has always been the backbone of Facebook’s business - to decide which adverts to show to which user. With the help of machine learning and DL, it divides us into different clusters in the most insightful way to serve us the best ads. Because of this huge user base, it kind of have an edge over the business compared to other companies like Google.
[To know more, visit LinkedIn]
The “Facebook” approach towards data analytics:
At Facebook, the Analytics team in Ads is made up of three different functions :
Data science
Data engineering
Product experience analytics teams
These three functions are the core analytics partners for most of the Ads teams, which is unique to Facebook. What they do is they bring the team together, as each team has their own specific leverage and responsibilities, and thus they make the product more reliable and effective for their advertisers. This collaboration also helps them ensure to prioritize and contribute in the most impactful way.
Hey, can we read this whole blog?
Yes sure, just go to Facebook Careers.
What makes Facebook different from other companies?
We just saw the Product experience analytics team is something which differs Facebook from other companies. well, that’s not all.
Beside these tools, Facebook also has its own framework, known as PyTorch. This open-source DL framework is build to be flexible and more ductile to research, with all the stability and supports for better product deployment. It provides us a python package with lots of features like tensor computation with strong GPU acceleration, and TorchScript. With the latest release of PyTorch, the framework provides graph-based execution, distributed training, mobile deployment, and quantization. It is so powerful that it can perform trillions of data based operations every single day.
Beside these, as Facebook migrated to PyTorch, means it will now be working with them closely; thus creating a big developer community around it, which is big plus nowadays.
If we compare Facebook’s PyTorch with Google’s TensorFlow, we can see the trend is more towards PyTorch than TensorFlow; mostly because of the ease of use and better performance of PyTorch.
Source : GitHub
[Read more in Analytics India Magazine]
What fascinates me about Facebook?
As an entry level data analyst, the main thing that fascinates me about Facebook is its huge community around the world. If we can utilize this properly, we can build our audience from it. Not only just growing audience, we can collaborate with different members in that group, we can learn from them (especially if you’re a beginner).
What Facebook’s mission statement is, "Give people the power to build community and bring the world closer together", and this community absolutely satisfies the mission of Facebook.
That’s it? 🥺
Yeah man, that’s it for this one.
“This one”, that means means more blogs are coming?! 😏
Yeah, absolutely! More of them are on the way.
Sounds cool, can we help you anyhow?
Yes, please. You can share this blog with your friends in social media, or just send it to whoever it needs to learn more about how different companies use data analytics.
And, You can follow us on Twitter. We share cool stuff, also we are always open for any discussions. Just hit up.
📌 “#30days30companies” is a blog project about 30 different companies, where I'm learning how these companies leverage the power of data, how do they perform data analytics, what is making them different from others, and why I am fascinated to work with them.
See you, in the next one.