Your Ops Guy
018 - 13. Processing the Processing Power
Part III-A: The Algorithm
In the previous 12 posts of this series, we've discussed the Facebook platform from 2 perspectives...the user and the business. This final part will dive into the backend of Facebook and how it leverages input variables from both perspectives to monetize and accumulate endless streams of data. Does anyone ever ask where this information goes? Are there endless streams of frigid warehouses filled with servers? How can all of this data become available to any user anywhere in the world at any given time? ...hang on, we'll get there.
First, it is important to understand how the Facebook machine, also known as the algorithm, processes input data and churns it into sellable assets. While we covered many of these points in the previous parts to this series, it is worth mentioning again from a slightly different vantage point. As discussed in part I, any interaction with the Facebook network (including external sites linked via Pixel or partner sharing platforms) add input variables that help define a user. We think of ourselves as people, who have feelings and emotions...but the algorithm treats us all as a series of zeros and ones...nothing but numbers and statistical data sets. As these data sets grow over time, the user is segmented into various categories designated by the algorithm. For example, if a user has behaviors of engaging with football related content, the algorithm will add that user to categories involving football, NFL and NCAA. But what is more impressive on the learning capabilities of the AI is the ability to predict user behaviors and interests based on that same input. Using the same example, the algorithm will run it's own A/B test on the user and see if baseball and basketball also fit within the user's interest perimeters. If correct, that user now becomes a part of that segmentation category as well. These A/B tests are constantly running within the Facebook algorithm and become more and more defined with each user interaction.
You may be asking yourself at this point...how does Facebook realize which content fits that segmented category? The answer, once again, is input. When a user posts a photo and writes a blurb of copy, tags some friends or throws in some hashtags, the algorithm defines that bit of content into a category. When a business does the same, it is even more defined due to the business input data sets during the onboarding phase of getting the business on Facebook at the beginning. If all of this is sounding very technical...that's because it is. Facebook is a very complex platform! Technical language will be used throughout this part of the series, but don't feel overwhelmed. It will all come together in the end.
Before we get too far ahead of ourselves, I want to revert back to an important question asked at the beginning of this post. Where does all of this information get stored? In short, a lot of places! Just like the integrate setup of a complex server network, Facebook communicates with itself using several server locations scattered around the world. To date, there are roughly 15-16 massive complexes set up at various points of the globe. (See map)
These areas are not the only places that house the endless streams of data, but they are the primary hubs that process them. Lets put the size of these complexes into context. As of March 2020, there was approximately 7.8 billion people on the planet. ~35% of that population uses Facebook on a monthly basis. ~23% of the population uses Facebook daily.
When you determine the magnitude of input and A/B tests and output that is required to accommodate all of those users on demand, the processing power is very intense. So intense, in fact, that the security measures taken at these massive complexes resemble that of government buildings...literally.
Facebook is in the process of finalizing a massive complex in Los Lunas, New Mexico. While doing other business in the area, We came across the site where this complex was being constructed. At the time, it was very early in the build stage, but the framework of the complex was clear. There was a massive warehouse secured by several acres of vacant land. Around the perimeter of the land...a wall. Not just any wall...this is the same wall that secures the White House. Cast iron gates secured with six feet of concrete pillars that stretch 8-10ft below the surface. The moon can crash into this fence and it won't budge. The estimated cost of this security gate is ~$100 per foot. With 43,560ft surrounding the circumference of an acre, the final question is...how many acres is the complex? Answer: 465. I'll save you the calculation...Facebook spent about $2,000,000 on a security wall. That should tell you how valuable your data is to them!
On the next few posts we will take a literal trip through the Facebook machine as a single piece of data. It's one hellova roller coaster ride, so be sure to check back in!