What follows is a description of how I believe Facebook’s newsfeed algorithm works in general. This is not based on any special or intimate knowledge of Facebook. Although I do work with products that integrate with Facebook’s advertising platform, I don’t have the particulars about Facebook. What I do have is a special set of skills and I will find you and… wait. No. I’m just saying I work in ad tech around big data and data management tools and thus have an inflated sense of my own ability to comment on how Facebook’s newsfeed algorithm works.
I think the first thing to understand about Facebook is that their primary objective is to provide something(s) that users want to use over and over again. In the parlance, they want to create a platform that is engaging for end users. That may seem obvious, but many people skip over this point and right to the idea about Facebook selling ads and using data to target ads and all that. But selling ads is a result of maintaining an engaging platform that users come to a lot and enjoy.
Some people will argue with me about this, I know. They’ll say that Facebook just wants to make money selling ads and that’s why they make Facebook so attractive to us users. This, friends, is Adam Smith’s Invisible Hand, so I am fine with that argument so long as we also recognize that the exact same thing is true for the extreme majority of content providers we interact with every day. Television shows? Yep. Money-grubbers. Magazines? Total greedballs. Newspapers? Radio? Yup and yup. All are selling ads. And in order for anyone to be willing to pay for those ads, they have to be able to attract audiences. And in order to attract audiences, they have to provide content of interest.
Before I look at how Facebook tries to determine what you will find interesting, let’s talk about some of the ways they can measure engagement.
“Engagement” is a very common term in online ad tech. In fact, if it weren’t such an important concept, the rate at which it is abused would turn it instantly into a buzzword. But it is very important. Simply put, “engagement” simply refers to users interacting with some bit of content or website in some way. Oftentimes, you’ll heard the word “reach” used along with engagement, but reach tends to refer to a more coarse number of visitors or eyeballs that content reaches. Engagement attempts to assess how appealing that content was to the eyeballs it reached.
So, how can you measure engagement?
You can look at the number of repeat visitors. Obviously, if people keep coming back, then you’re doing something right! But that’s pretty high-level and the internet is a wonder of technology that allows us to do a lot more. Here I’m going to start referring more and more specifically to Facebook.
Facebook can tell what you click on. If you click on something, then it caught your eye, right? You’re engaged with that content!
You can click “like” on something. That literally means you like it… whatever liking literally means. (And, let’s face it: that Like button doesn’t always carry a lot of meaning behind it. I just clicked “like” on someone correcting a typo in an earlier comment to a post, for instance. What? Will I get content related to dictionaries, spellcheckers, and style guides now?)
And there are Likes for pages and Likes for posts. And there is something sort of like a Like when you say you Like something in your profile.
You can hide things to show a negative engagement with this. Facebook will even ask you to provide some level of information about why as well.
You can stop and stare at things. This is a little fuzzy and some people are creeped out by it. But in ad tech we have a way of knowing how long a piece of content (usually an ad) is on the screen and how much of it is on the screen. So, if you stop and stare at something, it suggests there’s something about it that interests you.
All of these are more or less direct means of measuring engagement.
But Facebook CAN go a step farther in trying to figure out what else you might like because they can just look at what your friends like. Birds of a feather and all. So, even if you didn’t click or Like anything, they could just take a look at what is trending among your friends and make a pretty solid guess that it might interest you as well.
Another thing Facebook can look at in order to figure out what interests you that people sometimes forget about is the social sign on. You know how you can use Facebook to log into a news site or an app or whatever? That’s social sign on. And Facebook sees when you grant those permissions and they can make inferences about what you might want to see in your feed.
So, there are lots and lots of ways that Facebook can measure your level of interest in a particular bit of content and they can use any, all, or subset of them to make good guesses at what else might interest you.
Why am I talking about this? Well, because one of my Facebook friends asked about it in the wake of these articles:
My friend wanted to know how Liking or not Liking would REALLY affect her newsfeed. She’s been toying with the not-Liking strategy.
I just know there’s a report somewhere that shows how much traffic these two pieces are getting from Facebook relative to other sources and I would bet dollars to doughnuts FB is a leading source of referral traffic for both. I’d like to say it’s because I have amazing intuition about what the internet is doing, but I happen to already know that Facebook is a leader in social referral traffic everywhere. Oh, And FB has been coming after the Google, too, as a top referrer to news sites. So, basically, I’m a cheater at gambling.
Does Liking or Not Liking affect your newsfeed? If so, how?
Well, you can read those two articles above and see how it affected two people, but that is an embarrassingly small sample set. So, let me add my speculation to this embarrassment.
To start, let’s be clear: We won’t ever really know what Facebook’s algorithm is doing because it’s their secret sauce and because they’re always changing the recipe. I also think it’s reasonable to suspect that there is a learning component to the algorithm, so it should — in theory — evolve to get a more nuanced grasp of the things that interest us.
The basic idea, though, is if you have demonstrated engagement in a particular topic, brand, activity, event, or whatever then they want to show you more of that.
I don’t think Facebook wants to be a perfect mirror of your past behavior, though. Interests change. Trends shift. And a certain amount of diversity in our newsfeeds just makes good sense in terms of making it more interesting to us.
So, what if you remove one of the most heavily weighted measures of engagement, the Like, from the formula? Well, the writer for Medium felt that it evened things out and made her feed “better.” Better for her, mind you. It also forced her to adopt a different behavior in order to engage with the content she was seeing. But as mentioned, that doesn’t mean she was invisible to Facebook. Not by a long shot. (More on Facebook advertising in a moment.) I suspect her experience is merely a reflection of the fact that the Like is heavily weighted in the newsfeed algorithm’s calculations of what may interest you. As a result, the calculations had to rely on “weaker” signals for engagement and respond accordingly.
But what if you overwhelm the system with lots of noise by Liking everything? Again, the writer at Wired describes a result that would suggest that the Like is a significant factor in determining what you see in your newsfeed. He saw increasingly schizophrenic and polarized content, exactly the result we would expect.
Taking these two results together, my understanding and suspicions about how the newsfeed algorithm work are confirmed: you see more of what you engage with. So, it’s a bit frustrating to me when I see people complaining about what they see in their newsfeed and it’s even more annoying when they acknowledge their own role in creating it. If you don’t want to see a particular type of content, don’t click on it! Hide it! Send a signal that you don’t like it!
One of the things that I think is kind of strange in all of this is the notion that one might monkey with their Liking patters because in the words of the Medium writer, “I no longer wanted to be as active a participant in teaching Facebook how to advertise to me.”
My first reaction to that sentiment was, “Wut?” But then I had to remind myself that I work in ad tech and the way normal humans perceive online advertising tends to bounce between thinking the robots are some kind of omnipotent, omniscient gods at manipulating people into buying things by showing them a banner ad at the exact right time and that maybe the robots are actually just really slow monkeys. REALLY slow monkeys. Maybe actually moles. But I digress.
Here’s the thing: your newsfeed activity is not likely to be a strong source of data for targeting ads. Crazy, right?
Facebook is spending all this time and money to make your newsfeed something you really, really, really want to see, but at the same time they’re not relying on it (heavily) to target ads to you.
The reason is because first of all your newsfeed is entirely too specific. In online advertising, you buy ad impressions on the scale of millions. Advertisers do not have time to worry about whose favorite My Little Pony is Rainbow Dash and whose is Pinky Pie. Honestly, most media planners aren’t even looking at whether you like My Little Pony in particular when an affinity for animated television will do.
But let me shock you even more: Facebook has actually been cutting down on the number of ads they show you. And they’ve been raising their prices. AND they have learned that all this granular targeting doesn’t work as well as basic broadcast demographic targeting!!
That’s right! Rather than targeting some tiny group of people with a rare, shared interest, they’ve lately been learning that the most efficient targeting on Facebook is the same sort of targeting (Age, gender, income) that has been in use for a kajillion years!
I’m not saying no one is doing finely tuned targeting on Facebook. They are. And I’m not saying Facebook doesn’t offer targeting options that are derived from affinities observed through newsfeed engagement. They do.
But there are lots of ways not at all related to what you Like that are used to target ads every day.
It’s not clear to me why anyone would want to be shown irrelevant ads. If you’re in Facebook and seeing ads, why not do whatever you can to make sure they’re ads that might interest you? That’s a topic for another day.
The point is that stopping use of the Like button does not remove yourself from being an “active participant in teaching Facebook how to advertise to me.” It’s just not. Advertisers don’t rely solely on Facebook to guide their ad targeting decisions and Facebook doesn’t rely solely on Likes to determine targeting options.
It’s a much, much, much bigger ecosystem than that. I’m astonished that people don’t get that. Do folks think all the billions of dollars spent on online advertising and technology each year are just being shoveled into a boiler somewhere to keep the monkeys (moles?) warm?
Anyway, if you change your own behavior in order to change what you see in your newsfeed, I think that is fine and passingly interesting. But let’s not delude ourselves here. You aren’t denying Facebook anything… because you’re still using Facebook. And unless you’re like that lady at Medium who clearly clicks on a lot of things she doesn’t actually like very much, you’re probably not going to see an improved newsfeed.
My recommendation is that if you want to improve your newsfeed, then you should be more judicious about what you Like. You should look at all the different ways you can engage with the content you see and use them all just as judiciously. It’s not about teaching Facebook how to advertiser to you. It’s about teaching Facebook how to create a newsfeed that is meaningful to you.