Facebook is one of the largest social networking companies in the world. They are also one of the largest media companies in the world. From Instagram, to Oculus, to WhatsApp, Facebook owns a number of communication and media products intended to bolster the News Feed, the company's central product.
Facebook's News Feed is populated with content shared by users from thousands of external sources. Unlike other social streams like Twitter, the News Feed is sorted using a proprietary algorithm commonly referred to as EdgeRank. While some the EdgeRank sorting mechanisms are known to the public, the algorithm is largely a mystery to outsiders, and even most Facebook employees.
For good reason, the company argues. The algorithm is core to the Facebook consumer experience, and the company's bottom line. News Feed code costs considerable resources to develop, and is constantly adjusted by engineers at the company's Menlo Park headquarters.
As a competitive strategy, and to keep pace with an ever-changing news industry landscape, in 2015 Facebook hired over a dozen in-house employees to create original, native news content. And, like the algorithm, Facebook's original news content development process was largely a mystery.
Until now. TechRepublic interviewed a former member of Facebook's news team about how the technical and editorial product was produced. Fearing retribution from the company, TechRepublic's source asked to remain anonymous.
Explain your role at Facebook.
I was a news curator on a contract with a third-party company, and I worked on the trending news project in New York. That was how all the news curators were employed.
What were your professional qualifications prior to working at Facebook?
I was a freelance journalist. And I still am. Although, when I started at the company Facebook requested I not list the company on my CV, social media, or LinkedIn account. I'm not sure why.
Let's cut to the chase. Do you have a grudge against Facebook. What is your gut check feeling about the company, and your role in the news process?
There's been a lot of hype. I personally don't have a grudge against them. I mean, I appreciate that they gave me an opportunity. It was fine.
I don't think anyone is holding a grudge against the company. Everyone who left, left for a very good reason. We're not getting paid that much. It wasn't the greatest environment for people who want to do actual reporting. When you have a company full of inexperienced people, it's a little bit difficult to have somebody who knows how to be a leader, and that was part of the problem, administratively.
I think the way their hierarchy is set up, the people who were actually Facebook employees were pretty far removed from what we were doing. Our team leads, the managers, they were actually Accenture employees.
Your bosses weren't Facebook employees?
How much were you paid then, and now?
I made 27 dollars per hour. I make more now, obviously. Because that was a pittance. I had to freelance on the side as a necessity, in addition to preparing myself for post-Facebook life. Because, towards the end, I was always worried about losing my job.
Can you explain the job and the mechanics of your work?
We worked in 8-hour shifts, 24 hours a day, so if I was on shift from 4 p.m. to midnight on a given day, I would come in, would log into the Facebook tool. All communication was done through Facebook Messenger. There were [Messenger] Groups set up for people who were on shift at that time. That included one or two copy editors and several news curators, depending on the shift.
In the newsroom no one would speak in person. [We communicated] only over Facebook Messenger, which meant [all communication] could be recorded and monitored.
The algorithm that was running in the background would pull up about a list of 50 news topics that were trending on Facebook, in terms of what was being shared in that country vertical. They had the US, Great Britain, India, and Australia.
SEE: Seven ways to build brand awareness into your digital strategy (Tech Pro Research report)
There would be a combination of hashtags and keywords. [The trending topics] were organic within the system, like the algorithm itself identified particular keywords and then surfaced them to you as a news writer. You would get things like the word 'Syria' would trend, and we would have to go into this topic in the tool and figure out why it was trending.
The negative side of that is that things like the word 'Bear' trend in Canada, for instance. I mean, for whatever reason that word was constantly trending. Often, it would just be memes that people were sharing, or various local stories, or a keyword was being shared a lot within a country, but not some national headline.
Just to make sure I understand you correctly, it was your job to then figure out why a keyword was trending, and filter out the 'Bears' from the 'Syrias?'
Yeah. It was essentially their CMS. It was the interface to their algorithm tool. You would go in and you would have all these fields to fill out, with an image, a news story link, a headline that we would write, a description that we would write, and keywords. If somebody searched Facebook they could find this item, and then below all of that were the actual posts from users that were sharing this particular topic. It could've been individual users, or it could've been media channels, it could be TV news, or it could be celebrities like in verified accounts.
Trending topics were a combination of all that. For instance, with the 'Syria' hashtag, if we saw James Foley was killed we would go in and would see that people were sharing all these stories based on that incident, and that was grouped under the 'Syria' keyword.
So the algorithm surfaced keywords, you then added context?
Tell me about your approach to composing copy.
There was a style guide. We were told by managers to be as unbiased and neutral as possible. We were all used to it. We had all worked on news stories before then, so this wasn't something groundbreaking for us.
We had two or three managers at any given point, and they were called team leads. Usually one was on shift, except for the overnight shifts. Managers wouldn't be on overnight shifts, but they would be on call. They trained us at the beginning for a couple of weeks saying, "This is how you write the news headlines. This is how you write the descriptions. These are how many words were allowed, how many characters."
How did the style guide compare to other editorial style guides that you've seen and worked with, and other industry standards?
The style guide wasn't anything earth-shattering. [Our style] was very clipped. It was very short. You wouldn't obviously get the whole story [from Facebook], but you would get 90% of it.
Was there anything distinct about the work itself that made it good, or bad? As a former and current freelancer, was it a good gig?
At first, I think it was a good gig. There were 13-15 [curators], and they really wanted to ramp up the project at that point, so things moved really fast. [Facebook] made decisions quite quickly. Every couple of weeks there was a new task added onto our list of things to do. "Okay, now you have to fill in this keyword. Okay, now you have to click this other radio button in order to designate it in this topic vertical," versus just doing straight news.
We would write the headline, then put in the description, pick an image from Getty, then click the topic and select if a story was about video games or politics or entertainment or something like that. They refined those [topics] as we went along. We would put it in draft form, and the copy editor would go in and do all the copy editing they needed to, and then publish it, or they would edit and tell you to change things. You would get it back, change them, submit it again, and then they would do whatever they need to do to publish. That was all pretty standard.
READ: Enterprise startups: The cream of the new technology crop (Tech Pro Research story)
Writing stories took a little bit of time, and these stories are not always straightforward when they come through the algorithm. Sometimes you have to wait and see if a story was reported by a top news outlet.
The problem was for me—and I know it was for other people—was that it became a numbers game. When you didn't hit your numbers, it was always stressful. And management never managed well. At some point, it was unofficially like, "Okay, you have to reach X number of topics per shift," and it was understood that if you didn't do that, you would get called out. They had a public system of calling us out.
Can you elaborate on that system?
It was a spreadsheet with everyone's names and the number of topics published that week. (Editorial note: Facebook referred to news stories as 'topics.') One column noted the difference between the week prior and the current week, and a number of 'points' accumulated in the next. The points could be redeemed for Facebook swag, and after a certain level, some money. I don't think anyone had gotten to that level. Eventually they got rid of the system after people indicated how childish it was.
What did they use for spreadsheets?
It was Excel.
Did you have a quota?
Unofficially. It was never actually stated. But, I think unofficially it was 20 or 25 topics. Sometimes, like when it's a slow news day, it was hard to hit 25 topics. I mean, there were people that hit 30 or 40 on a regular basis, but if you looked at what they were writing...
Sometimes you see [a trending keyword], you would think it's an important story, but then you're like, "Screw it. I just need to get to my numbers."
How did you verify trending topics?
Literally, Google. The place we would go to verify all our information was Google.
At Facebook, you used Google?
We would go to Google to verify things. Like someone's name, or to get details of stories that we couldn't glean from bad local news writing, we would go to Google. We relied very heavily on Google.
I would take what people were sharing on Facebook, and I would search the terms on Google to figure out if it was actually a story.
Did your bosses know you used Google?
They absolutely knew, and they were doing the same thing. Because if we couldn't find more information, they were Googling it.
Would you ever search inside Facebook first?
The only Facebook sources we used were what was being shared by Facebook verified accounts.
Were signals other than Google and Facebook verified accounts used to check the veracity of a story?
No. You had to be a verified Facebook account, or you had to be an account that we could independently verify through Google, like a police department or something like that. We would go to their site. If they had a link to their Facebook account and we knew it was legitimate, we could publish that. For missing person's cases, the police department, if the parent was somehow verified or famous or something. For all other stuff, you had to go to Google. We relied on it.
Can you explain the editorial process, and how you selected stories? For example, let's say there's 10 trending topics, one is Britney Spears, one is Syria, one is United Nations. Can I just choose whatever I want?
Yep. At first it was a free-for-all. You just choose what you wanted, and then every once in awhile there would be something big that would come along, and the copy editor would say, "hey, if you're not working on a topic right now, or you're about to choose something else, why don't you work on this?"
But it evolved over time. Then they would say, "Oh, well we want to reach 50 topics in each vertical every shift." They started implementing subject specific verticals. I think they only covered US news. One was politics, one was entertainment, one was video games, one was tech, one was like business.
On overnight shifts you would just get to the point where you're like, "okay, I'm five and a half hours in, six hours in, I'm not meeting my numbers. Screw it, I'm just going to get all the entertainment topics out of the way, because it's easy." We all did this. We just had to pump topics out.
There are editors that would say, "hey guys, come on. Let's do the stuff at the top of the list that came up in the tool, versus what's at the bottom."
Can you explain how the news algorithm worked?
We're so removed from that, we just didn't have any input on the algorithm. They want to personalize that News Feed, and we were there to fill the News Feed with stuff.
The engineers who worked on the algorithm were in the Menlo Park headquarters.
Did the algorithm have a political bias? Would it surface more liberal or conservative stories?
No, not in my experience. I mean, we all definitely had our own personal bias, obviously, but that happens in every newsroom.
READ: Choosing the right tech for your company's future (Tech Pro Research story)
Where I think that bias is was actually in the way we were doing verification. You often had to wait to have one of the top 1,000 news outlets for us to pick it up. The thing is I don't know who designated those top outlets. Our understanding was that outlets were selected by the managing editor, and the manager below him.
Can you summarize your experience on the Facebook News team?
Accenture was an awful company to work for. Facebook didn't care. News curators were at the bottom. I mean, if this is a ladder, we're the bottom rod.
Why do you think that's the case?
I think they're a company that, like any company that hires journalists, don't really understand [news]. They seem to think it's easy and automated.
What's the right solution to the Facebook news problem? Let's say the leadership of Facebook sat down with you sincerely, and said, "help us build this the right way." What would you suggest?
I suggest they train editors on how to manage people, and how to manage journalists. I think the people who were in charge were following directions, and so they had very corporate policies. Their policy is, "okay, we have to reach X number of topics overall per shift, or advertisers won't be happy." They should come right out and say it.
Dan Patterson has nothing to disclose. He does not hold investments in the technology companies he covers.
Dan is a Senior Writer for TechRepublic. He covers cybersecurity and the intersection of technology, politics and government.