Disclosure: Stream, the API for scalable feeds, has previously sponsored Hacker Noon.
Today, we’re going to catch up with their CEO, Thierry Schellenbach about Facebook news feed, and the future of the news feed at large.
For previously sections of this interview, please visit:
David: So Facebook’s news feed has been under a lot of scrutiny and pressure. For example, stories shared on the platform impacted the election and have altered our democracy. And now, the removal and the devaluation of publisher posts from news feeds have left the Buzzfeeds of the world in a tough spot and under-reporting on their earnings.
So the news feed is very, very powerful. If you could have a conversation with Mark Zuckerberg about how to fix or improve the Facebook news feed, what would you say to him?
Thierry Schellenbach: Oh, that’s a tricky one! I think it’s actually a combination of problems that they struggle with. First, they have fake accounts, which is a problem that not only Facebook has encountered, but Twitter even more so.
Yeah. I mean on Twitter, you’re allowed to have fake accounts or pseudonyms.
I think that’s definitely part of the problem. Doing additional vetting on their advertisers would be beneficial since advertising played a huge role in how a few individuals managed to control the conversation on platforms like Facebook. In short, they should take a closer look at fake accounts and advertising and start to tackle those issues.
Another thing to mention is that personalization (machine learning) for Facebook is crucial since there are so many updates that need to be prioritized. I mean realistically, you can’t see ALL the information your friends are posting, it would just be too much. I believe the average Facebook timeline gets 1,500 new posts a day, which is impossible to review manually. They had to create some sort of prioritization methodology. However, while I do believe using this technology is necessary, it’s important to not get caught in a filter bubble. In addition to seeing views and ideologies similar to theirs, users should also see opposing views.
I would advise Facebook to optimize their machine learning to not only show what users are most likely to engage with but to also, every now and then, encourage users to check out new topics and things that outside of their regular interests. Quora is platform that does this well…or at least does a much better job than Facebook.
Didn’t the Quora founder used to be the CTO of Facebook?
But in terms of picking things outside your interests, that’s a very dangerous territory for the platform to be in. It’s content that you haven’t done something to opt for. Is that what you mean?
So, Facebook’s personalization methodology is done in a different way than most other platforms. Quora will show you content outside of your feed. Instagram mainly uses their personalization for their Explore section — you have your regular feed AND you have the explore section where you can discover new content. Facebook primarily uses personalization to reorder content within your feed. By doing it this way, they also create a really strong filter bubble.
To get out of that filter bubble, Facebook should, every now and then, show content that is outside of your predefined interests. For example, if you don’t engage with posts about a certain topic they should occasionally show you that topic again to see how your preferences have changed. What’s interesting to an individual may drastically change day-to-day, let alone year-to-year.
That’s really well articulated about how the filter bubble works. Personally, I’m just not convinced Facebook wants to get out of the filter bubble, and I think the filter bubble is very profitable. And this self-fulfilling prophecy is what they want and wouldn’t tell you. Just a thought…
Yeah, that may also be part of it. Their machine learning is clearly set to optimize for engagement. While they obviously have to optimize for engagement, there’s an important difference between long-term and short-term engagement. I think finding a greater variety of things people are interested in, and showing more of that, could potentially be better for them in the long run. I’m not convinced that showing you only predefined interests and never going outside of that bubble is actually good for long-term Facebook engagement.
Yeah. I think some of the marketing they’ve put out recently is confirming that, i.e. ‘we want you to spend less time and have more meaningful engagement.’ So how closely do you monitor all these updates in the news feed of other people’s tech? Are you just always monitoring the Instagram, Twitter, Pinterest, etc. news feed changes and learning from it?
We definitely look at the market and see what the new approaches are. The latest iteration of our tech is actually very similar to LinkedIn’s feed technology. LinkedIn has a large team working on their feed. They’ve publish papers and blog posts about how they’re developing it. (Linkedin & RocksDB, Linkedin’s AI)
Another trend that we’re seeing now is aggregation. So if someone posts 100 updates, aggregation works its magic to show those as 1 update. This greatly reduces the noise in the feed. Bandsintown is a good example of an app that leverages aggregation.
We’re also starting to see feed-based personalization for more business-practical use cases, in addition to traditional social networks use case. Feeds are now being used in real estate apps, job sites and even e-commerce. Perhaps you could say they’re actually more useful in other industries than for social networks.
Are there any other industries where you think the news feed is really going to expand and become a more prominent feature in technology for that industry?
Based on trends we’re seeing today, I believe news feeds will continue to become increasing popular as the discovery feature on websites. When you know what you’re looking for, it’s quite simple to just search for it on any given site. On the other hand, at times you go to a site without knowing exactly what you’re searching for or want to find and can use some assistance.
Let’s use the real estate use case. All real estate sites, whose goal is to help you find a house, want to help you discover house postings that meet the user’s unique needs — # of bedrooms, bathrooms, garage, neighborhood, etc. As these types of sites continue to improve, feeds will play a much larger role.
A great example of a company that uses advanced feed technology for their shopping experience is Etsy. Etsy wrote a few papers on how they approach machine learning. With Etsy, whenever you click on or search for an item, their machine learning essentially ‘learns’ your interests and continues to improve and optimizes for e-commerce conversion.
Yes, that’s cool.
As feed technology becomes easier to use and machine learning becomes more powerful, it will power content discovery in more and more apps making content more meaningful.
We’re gonna get to the point where every single home page is personalized. Like visiting Coke.com and Pepsi.com you probably surface a personalized experience based on your activity everywhere on the internet.
Well not every site needs to be personalized. Some communities and sites are very uniform — only appealing to certain audiences. Let’s use Dribbble for instance. If you go on Dribbble.com, you’ll find a community for designers. The site has a very niche and narrow focus. So I think these types of focused/niche sites will never personalize the user experience.
However, most sites appeal to a more diverse audience. You’ll see them start more and more of them starting to implement personalization to have a more tailored experience for their users.
Let’s get serious for a moment. What are the biggest threats to your company?
I think the biggest danger to us, and for any other company in this space, are the big players like Microsoft, Google, and Amazon — and the chance that these big hosting providers will come out with dedicated products.I know that SendGrid, another Techstars company, ended up getting competition from Amazon.
I’d say that’s my biggest concern that keeps me awake at night. If the folks from Amazon or Google decide to enter our market, it would give us some tough competition. Other than that, similar to so many other startups, it’s mainly about your own execution.
It’s way more common for startups to run into trouble because of their own choices, or their own execution than external factors.