Skip to main content

A long talk with Facebook about its role in journalism

A long talk with Facebook about its role in journalism


Happy birthday, Instant Articles!

Share this story

A year ago today, Facebook introduced Instant Articles — a quick-loading news format, hosted on Facebook’s own servers, that opens inside the company’s flagship app up to 10 times faster than the mobile web. What began as a small test has now rolled out broadly to hundreds of publishers, and the format is now available for anyone to use. Consumers have responded well to faster news browsing: the company says people are 20 percent more likely to read Instant Articles, which are accompanied by a lighting bolt icon in the feed, and 30 percent more likely to share them with friends.

If consumers’ embrace of the format was easy to predict — who enjoys waiting for a website to load? — publishers’ feelings about Instant Articles were harder to gauge. The run-up to launch was met with tremendous anxiety among some publishers as they grappled with two realities: one, the majority of their audience is consuming news on Facebook; and two, allowing Facebook to host their articles directly meant giving up some control over their appearance and the ads that could run inside them. More than one publisher worried Facebook’s end game was to get publishers "hooked" on the format and then demand an ever-growing share of their ad revenue.

But so far, Facebook has taken the opposite approach. When publishers asked to be able to include more ads per article, Facebook let them. In March, it enabled video advertising; a week later, it allowed publishers to share sponsored posts as Instant Articles. In short order, Instant Articles became a model for the industry, with Apple and Google quickly stepping in to offer content hosting solutions of their own.

Facebook's growing power has raised new questions

But Facebook remains the dominant player in the game — and its growing power in the distribution of news has raised new questions. In December, I posed the idea that Facebook ought to take responsibility for making its users more informed. The company reached out to me afterward, and I suggested we discuss the subject on Instant Articles' anniversary. Facebook set me up with Will Cathcart, a six-year veteran who oversees product management of the News Feed.

The timing was fortuitous: a controversy over what news goes into Facebook's Trending Topics feature dominated headlines in the days before our interview. (If you want highlights from that, I've helpfully aggregated myself here.) Trending Topics is overseen by Tom Stocky's search team, not the News Feed team, but Cathcart was able to clarify for me how it works.

The result is a chat that finds Facebook committed to minimizing editorial oversight over news products. In some ways, this marks a retreat. The company hired a managing editor in 2012, but he left a year later, saying: "Facebook is meant to sort of fade into the background. When Facebook starts producing content, it takes you away from that mindset." At the same time, Facebook acknowledges that journalism is one of the primary reasons people return to the News Feed each day. And it turns out the company is committed, in a way, to making its users feel "informed" — though in this context, what "informed" means remains very much open to interpretation.

This interview has been lightly edited for clarity and condensed.

Casey Newton: It’s been a year since Instant Articles launched, triggering a rush to off-platform distribution for some of the biggest media companies in the world. How is it going so far, from Facebook’s perspective? What have you learned?

Will Cathcart: A year in, the thing we’re really excited about is one, it really works to improve the experience. We announced at F8 some of the stats around this: for people who see enough Instant Articles in their feed, they’re 20 percent more likely to go read an Instant Articles than an identical article hosted on the mobile web. Because they’ve learned the experience is so much faster and richer. For people who click on Instant Articles, they’re 70 percent less likely to bounce. And then I think, coolest of all, people who read an Instant Articles are 30 percent more likely to share it with their friends. So from that perspective it’s going really well.

People who read an Instant Article are 30 percent more likely to share it

The other thing we were really focused on from the beginning is, how do we do this in a way that works for publishers? We knew publishers weren’t going to take the time to adapt their content to Instant Articles unless it worked for their business. And there I think we’ve done really well — we’re seeing lots of publishers coming onto Instant Articles. We went to completely open sign-ups at F8, and have seen a ton of momentum after that. All along, we’ve focused on listening to publishers about where Instant Articles was working for them and where it wasn’t, so we’ve made a bunch of changes throughout the course of the year.

Facebook Instant Articles WordPress

Like what?

We’ve been starting tests with some publishers where we help them identify readers that have read multiple articles from them, and then upsell them to consider subscribing to an email newsletter. If you’re not someone who knows a ton about The Verge but over the course of a couple weeks, three or four of your friends share cool articles from The Verge, so you’ve actually read a few, we think it’s a great time to say, "Hey, would you consider subscribing to get more content from The Verge?" This we think could help publishers reach a new audience of people they might not otherwise have reached, and start to move them down the funnel to become subscribers.

After the launch, some publishers began pressuring you to let them insert more ads, and different ads, into the product. How do you balance giving them what they want without slowing down the experience?

I would actually just describe it as, we’ve been talking to publishers and trying to get feedback from them, which we think is a good thing. We have been very conscious to try to work through what types of ad experiences we think work for Instant Articles specifically, and also do a good job of balancing the real need that publishers have to generate revenue with creating a good user experience. So for example, we added support for autoplay video ads inside Instant Articles, but with the sound off. We added support for rich media ads, but where they stay within a fixed iFrame and don’t overlay on top of the content. So just a lot of detail-level decisions with publishers about asking them, "What do you need for your business?" But then trying to map that to all the research we have from consumers on what types of ad experiences resonate the most with them.

From a product perspective, where would you like Instant Articles to go next? Do you see the format as more or less complete, or are there things you’d like to see it add?

There’s a lot we can do to continue improving it. One whole dimension is continuing to make it work better and better for publishers. We’ll continue to get feedback from publishers and try to add more functionality. The second is the reach of the program — we just went to open signups at F8 about a month ago. We have a lot more to do to get this available — particularly all around the world. The more you get to people who are using worse network connections, or less powerful phones, the more important Instant Articles become. We talked about a couple of countries at F8 where the lift isn’t 20 percent, it’s 20 to 40 percent. That means people in those countries are reading that much more news when you get publishers on Instant Articles.

The worse the network connection, the more important Instant Articles are

And then the last is, what new types of interactivity or richness could we add to Instant Articles? I think we could explore more around, once you’ve read an article that’s really interesting, how do we help you discover other articles?

When Instant Articles launched, some publishers expressed fears that Facebook planned to tax distribution somehow, cutting into any ad revenue. But a year later, the company seems committed to keeping the format free (while taking a cut of any ads it sells). How does Facebook think about monetizing Instant Articles? Is it simply the case that Instant Articles make the News Feed better, and you monetize the News Feed?

We got into this to improve a core part of the consumer experience. People come to Facebook for news, and they want to read it. If we make that experience better for consumers, that makes News Feed better overall, which is good for us. It’s good for us to serve the needs and desires people have when they come to Facebook. We deeply believe that the only way to have publishers use Instant Articles is if this works for them. So we’ve always had, from the beginning, the approach of figuring out what publishers need and realizing this needs to be net better for publishers’ businesses. And I think that’s why, a year later, if you look at all the changes we’ve made, they’ve been changes that are in response to publisher feedback, and not in the other direction, which might have been the fear.

instant articles

One thing that I think has generally been under-discussed is how many news products Facebook has built. In addition to the News Feed, it built Paper, it built a bot platform that media companies are using, and it built its box of "trending stories." Is demand for news growing on Facebook, or are you guys just responding better to that demand?

Seeing how people use Facebook has always been one of the ways we inform what our product roadmap is as a company. When the only way you could share photos on Facebook was through the profile picture, people at Facebook noticed consumers were changing their profile picture frequently, because they wanted to showcase more pictures. And that led the company to go build photo sharing.

I think with news, we didn’t start out building news-specific products, but found that people used Facebook to share news with their friends and talk about news with their friends. And because of that we’ve said, "OK great people are doing this, how do we improve the experience?" If you look at what people come to News Feed for, I think they come for a few different, overlapping reasons. One is to connect with their friends and family. Another is to be entertained. And another is to find information that’s informative about the world. So because people come to Facebook wanting to talk to their friends about news, to learn about news, we’ve been increasingly building products that try to make that experience better.

"Fundamentally, our role is to help people get what they want."

I want to build on that and ask some higher-level questions. Facebook now has a greater reach than any individual publication in the world. How do you view the company’s role in distributing the news? And what responsibilities does it think it has around that?

Fundamentally, our role is to help the people who use News Feed get what they want. Part of what they want is to be informed on the topics and interests they care about, from the publishers they care about. Or from the friends that they want to hear from. And so we view our role as to do the best job of that — giving each person what they want. Which means for each different person, from a different range of viewpoints on a different set of topics, from a different set of publishers. We don’t view it as our role to help shape in any way what that is — we want to help people get what they want.

In particular I think we view our role — and I know you’ve been interested in this in the past — is helping people get the core of what they want, which is to be informed. In particular for news, what is most informative to people based on their personal interests? And how do we do a good job of helping people find that out?

My next question was around that particular word, "informed." The mission that you guys often talk about is making the world more open and connected. And something I’ve said is, that’s great, but you should also make the world more informed. It sounds like you’re saying that does square with the company’s values. From a product perspective, how do you measure whether people are informed?

One thing we do is we ask people. We have people who go and look at their News Feed and then for each of the stories that they saw, they answer how much did they want to see it in their News Feed? And we ask along different dimensions of what kind of value did they get from the story — did this connect them with their friends and family? Did this entertain them? Did this inform them? We ask that. It shows up in the core data we use to figure out how to improve News Feed.

"Did this inform them? We ask that."

More broadly, in terms of how we measure overall how good a job is the product doing to help inform people — one is we look at how much people are using News Feed, how often people are coming back, and asking them how happy they are with the experience and why. And the other thing we can look at is how much are people learning about the news or discovering the news from Facebook, or their friends on Facebook. There’s a lot of external data that suggests for some people it’s really, really meaningful.

Certainly I’ve seen articles that suggest that Facebook is the primary source of data for a lot of people, which is maybe the data that you’re referring to. Are there external measures of how informed those people are? It becomes really hard without actually giving them a quiz on current events, right?

Yeah, it does. I haven’t seen anything — and I think it’s really hard to, in particular, get this internationally, and across the whole spectrum of stuff people are interested in learning about.

facebook trending

"We think about what we do in a really personalized way."

So as we saw from the news this week, informing people can be tricky, because it involves questions of editorial judgment, which is subject to second-guessing and criticism in a way that algorithms usually haven’t been. How comfortable is Facebook, generally, with the idea that it can’t distribute news without exercising some kind of editorial judgment? Would the company abstract away all editorial judgment if it felt it had built algorithms that could do it? Or do you see a role for that kind of thinking around curating a news package for consumers?

We think about what we’re trying to do in a really personalized way. And so I think if you’re trying to build a product for over 1 billion people to be informed about the news that they care about, you can’t really be building a product that has judgments about particular issues, or particular publishers. Because that doesn’t match what 1 billion people around the world want.

So we’ve really thought about it as trying to build a product that helps each individual person find and talk about the news they care about, which should really range across topics, across points of view, across publishers. I think we are comfortable, where it makes the product better, relying on people. For example, the stuff you’ve heard about in Knoxville, where we get feedback from people about what content was most informative to them. But I think that’s very different from having an editorial point of view.

We do have some product points of view we’ve made. For example, we’ve chosen to have a News Feed that shows people the content that’s most meaningful for them, rather than most recent. We could have built a product that was purely chronological or timely. We’ve chosen to map not just to what people find entertaining, but also what they find informative, because we’ve seen from the data that people want that. But I think the traditional editorial choices you’ve talked about for a news publisher just don’t work if you’re trying to build a product that works for a billion people around the world. Because they all want different things.

"Traditional editorial choices don't work if you're trying to build a product that works for a billion people."

Giving people what they want and keeping them informed have some overlap, but not 100 percent overlap. Traditionally, the way editorial organizations have thought about that is, we’ll present you with a package of goods. Some will be really entertaining; some will be informative. Those organizations would be really comfortable leaning on their editorial judgment. What I’m hearing you say is, you’re not interested in Facebook figuring that out at scale. Or it seems impossible, at scale, to take that same kind of view.

If we’re trying to do something that really captures what a billion people in want in all the countries around the world, in all the different languages, it’s hard to imagine we could sit down and figure out what people should hear about. We’re not going to do a good job of that.

I do get what you’re saying about how what people want may not perfectly line up with what’s informative. But I actually think if you look at how people use Facebook, they really do want to be informed. We didn’t set out to build a bunch of products directly tied into news, but we did build products to help people share what was going on with their friends. And they started sharing the news. The reason we got excited about building things like Instant Articles is that we saw how often people saw news in their News Feed and clicked on it and read it, even despite it being a really hard product experience.

So for most Facebook users, the biggest source of news on the platform is the News Feed. But you also have a box of "Trending Topics" that lately has been the source of controversy. What is the relationship between the News Feed and Trending Topics supposed to be? What roles do they serve?

I can talk a little bit about it — the nitty-gritty details of Trending I don’t work on specifically. It’s part of our search team. But part of that is the high level of how we thought about the product. News Feed I think of as the place you go to see the most meaningful or informative or entertaining posts from the people you follow. You curate this as a user. Trending and search help you explore what’s going on in Facebook, in general. I’m really interested in the Syrian refugee crisis. If I come to News Feed I’ll see what my friends are saying about it. If I follow some news publishers I might see what they’re posting about it. If I go to search and search for it I could get anything on Facebook. And Trending fits into that, where it shows people what is most discussed or talked about on Facebook right now across the whole network, whether or not it shows up in the News Feed and whether or not it comes from their friends.

"Trending and search help you explore what's going on in Facebook."

This week, an anonymous former employee said he was prevented from inserting stories from the conservative press into the Trending box. Facebook has denied that this happened. But in newsrooms, there are always discussions over news judgment between editors and junior staffers. How developed is Facebook’s sense of news judgment, would you say? In five years, should we still expect human editors to be making news judgments on a feature like that?

We care about creating the product that people want. Whether or not we can do that entirely with automated systems, or it’s helpful to have people help, is actually just a detail. What’s more important is the product principle, which is that we want this to show you what you’re most interested in. We’re not interested in adding our point of view — we actually don’t think that works for a billion people. And some of the stuff that we’ve talked about publicly in response these questions about why we have a team is things like, if someone people are talking about Star Wars and some people are talking about "May the 4th be with you," it’s kind of a bug if the product treats those as separate stories. Or if a bunch of people start talking about lunch today, it’s kind of a bug if "lunch" showed up at the top. So that’s helpful to fix — but how we fix it is just a detail of how we get the product right.

I think the question is, what happens when everybody’s saying, "Obama was born in Kenya." Is there someone who comes in and says no, actually he was born in America?

I think you already see that happen on the platform today. It doesn’t have anything to do with us — people post a lot of this stuff and talk about it, and other people post different points of view. And the nitty-gritty of the details of how we should be involved I actually think is less important than building a platform where if people want to talk about that, it’s really easy to talk about that and find different points of view.

So you’ll make it easy for them to talk about it, but you’re not going to put up a big banner that says "by the way, this is nonsense."

I actually think people’s friends do a lot of that already, which is one of the really powerful things about Facebook. There’s been some research published on this that says if you’re connected to a lot of friends across a set of diverse points of view, you actually end up learning more about the different points of view than if you weren’t using Facebook.

There have been some calls for Facebook to offer more transparency around how it decides which stories to feature. On one hand, traditional publishers are rarely asked to explain the method of their news judgment to anyone. On the other, they don’t have the distribution power that you do. How do you think about those calls for transparency?

For the Trending stuff specifically, I’ll defer to Tom and the search team. I think for News Feed and some of the news products we’ve built overall, it’s so personalized that it’s hard to talk about what transparency we would give in general about specific stories. The general thing we’ve done over the past couple of years, in part in response to these questions and criticisms, is to try to be more transparent about how the News Feed works in general. We talk about the algorithms we use and why we use the algorithms we use, and why we change it we talk about when we change it. So doing a bunch of whiteboard sessions where we go through, here are the top things that the algorithm looks at, and doing a blog post every time we make a substantive change to the News Feed. I think we’re trying to be a lot more transparent about it, without being able to point to every specific person’s News Feed, what were all the decisions that went into it.

"It's hard to talk about what transparency we would give about specific stories."

Well those were my big questions. Are there other things you want to say? Do you want to gesture expansively toward the future of news of Facebook?

I don’t know about "gesture expansively." The main thing, given that it’s been a year [since Instant Articles], the main thing is that we’re pretty pleased with how Instant Articles have gone. When we set this out we knew from the data that people loading mobile website really slowly was a problem. If you fast forward a year, Instant Articles has demonstrated it’s working. It’s kicked off a conversation across the industry about how important it is that news be really, really fast on mobile. And I think we’re still getting feedback from publishers and adapting over time, but I think we’ve managed to do this in a way where it’s better for the end user experience but also works for a publisher’s business.