Like This if You Like Pandas! Facebook Says Publishers Shouldn’t Fret About News Feed Changes.
In 2011, Google changed the way it ranked websites, in an effort to punish spammers and “content farms” that showed up high in search results but delivered crummy pages.
This week, Facebook announced that it was changing the way it ranked content in its all-important News Feed — the main page Facebook users see on their desktop and on their phones — in order to promote “high-quality content.” And Facebook said it would make things like “meme photos” harder to see.
The immediate reaction from several publishers I’ve talked to this week: “This is Facebook’s Panda.”
But if that’s the case, then who is Facebook trying to punish? And why does Facebook care about this, anyway — isn’t the crucial thing that people like the stuff, and not what the stuff is?
One way to get some answers is to ask a Facebook executive directly. News Feed manager Lars Backstrom got on the phone with me to explain the company’s thinking.
Peter Kafka: Your post says that you want to promote some kinds of content and demote other kinds of content. But I don’t really understand what you want to push up and down, and why.
Lars Backstrom: We don’t really think about it that much in terms of promoting and demoting certain kinds of content. The way we think about it is that we’re doing a better job of identifying value.
In the past, there were a lot of things that all fell into one bucket, and we would treat them all the same, even though they clearly weren’t. If you see a funny meme photo in your feed — sure, you get some value from that. But if you compare that to reading 1,000 words on AllThingsD, you would presumably get more value from that experience than the first one. And, in the past, we were treating them as the same.
But if I like them both, aren’t they the same? From a Facebook perspective, shouldn’t the things that people like be the things that people like?
I’m not saying that one doesn’t have value. And we’re not trying to impose our will and view on the world. But we went and asked people which of those things they get more value from. We’ve run surveys, and asked people to rate stories and things.
And they’ll say, “The cat photo was great, and I had a good chuckle, but of those two, the second one enriched my life more, and I got more value out of it.”
It’s not us trying to be more proscriptive. We’re trying to align our definition of value with that of our users.
You make a good point, which is that the surveys are not necessarily the truth. But it’s just as naive to treat every single click as having the same value.
Are you paying attention to the source of the content? Or is it solely the type of content?
Right now, it’s mostly oriented around the source. As we refine our approaches, we’ll start distinguishing more and more between different types of content. But, for right now, when we think about how we identify “high quality,” it’s mostly at the source level.
So something that comes from publisher X, you might consider high quality, and if it comes from publisher Y, it’s low quality?
In your post, you specifically call out meme photos that are hosted on other sites. Is there something about photo-hosting sites you have an issue with?
No, that was just an example. There’s no targeting of one category or another.
This seems very reminiscent of Google’s Panda changes a few years ago, which was a really big deal for search. Is that what you’re doing here for social?
I’m not totally familiar with the details of Panda. At least from the way you described it, it’s maybe not quite at that scale. But it’s kind of a step in that direction.
Whenever we make a change like this, it has the potential to break some of the strategies employed by people who get distribution on Facebook. My favorite example of this is when you have a photo, and then a very explicit call to action where you say “one like = one respect.”
So, when the text or photo has a call to action, those posts naturally do much better. And in a traditional feed ranking, where we’re evaluating just on the number of likes, those things all did very well.
So “like this if you like puppies” is the kind of thing you want to push down. What about publishers like BuzzFeed, and then Upworthy, and now a slew of Upworthy clones, that seem to have figured out how to crack the Facebook/viral code. Are you trying to rein in those guys?
I wouldn’t say we’re trying to rein in anybody. I’m not sure how the most recent changes will affect those sites.
Our goal is to provide user value. We’re trying to do that algorithmically, and if people find ways to game the algorithms that we have, then we have to adapt.
So this is not aimed at BuzzFeed?
What about Upworthy?
We don’t have any sort of specific enemies or targets.
In the past, you guys have made changes to the way the News Feed works, and it has had a big impact on people who depended on Facebook for distribution. Like Zynga. Or, more recently, SocialCam and Viddy. But those people did things that regular users didn’t like, and here regular users like the stuff …
I think you’re making this into a slightly bigger thing than what it is. We’re trying to be as proactive as possible about messaging even things that might have a 10 or 20 percent impact.
So it’s not like you’re never going to see a funny cat photo from Imgur or some photo-sharing site anymore. It’s that maybe you’ll see 10 percent less of that, and 10 percent more articles, and things like that.
We’re trying to be a little better about that.
And even a 10 percent change can be very significant to some companies, given Facebook’s power.
The main thing I want to be clear about is that we want to optimize for user value. We recognize that this is one of the harder things we have to do.
There are certainly people that come to Facebook looking to see funny cat photos. And we want to make sure that it’s a good experience for them as well as people who are looking for more serious news. Ultimately, we’re trying to make it as personalized as possible, and give people what they really want to see.
(Panda photos via Imgur)