Outbrain CEO Yaron Galai on content recommendations for publishers

Yaron Galai is CEO of Outbrain, which provides content recommendations for publishers including The Guardian, The Sun and CNN. 

I recently included irrelevant related links/further reading in my list of dodgy ads and UX on news sites, and Yaron responded on Twitter, so I thought it would be interesting to find out more. 

For the uninitiated, can you explain what Outbrain does? 

Outbrain is a content recommendation platform. For publishers, Outbrain powers the recommended links that you'll typically see at the end of articles (sometimes called "You Might Like" or "From Around The Web", etc).

For content marketers, Outbrain enables to attract a highly interested audience, at scale, back to their article or video content. 

 

Why should should publishers use your service? What do they gain from this? 

Publishers gain in three ways when they install Outbrain:

1. About half of the links we typically serve point back to the publisher's own content. Those links are a free part of our service, and just by having them on the page we'll typically lift the publisher's pageviews by five to 10%.Those new PV's can be monetised by the publisher, and again, they were generated for free.
2. For bigger publishers that fit our criteria, we share the revenues we collect on all the links pointing outside the publisher's site. For many publishers, this part becomes an important revenue source which did not exist for them before Outbrain.
3. Lastly - having Outbrain installed on a site improves the overall user experience. The links we serve create a complimentary service to the publisher's content, helping the user with the question of "What should I read/watch next?" 

So for publishers it's a source of additional new PV's, of additional revenues, and we do all that while improving the user experience. 

Which publishers and advertisers are you working with? 

• Publishers: CNN, Fox News, Hearst, Sky News, The Guardian and The Sun.
• Brands: General Electric, Allstate.

How do you choose which content to promote on publishers' sites? 

Over the years we've developed about 35 proprietary algorithms that compete to select the best link for each individual visitor to a publisher's site.

Some of the algorithms try to match links to the story the user just read/watched. Some of the algorithms look at traffic patterns across the site and try to match links with a "people who read this also read that" approach.

Some of the algorithms look at all the inbound traffic from social networks like Facebook and Twitter. 

At the end, all of the algorithms try to best personalize the recommended links for each individual user based on the content they found to be interesting on the same publisher site in the past. 

How important is contextual relevancy and how can you achieve this? 

In some cases contextual relevancy is very important, and in most other cases not very much. Contextual relevancy has been extremely important in search: if a user is looking for something specific, and can express that in a keyword, then the search results ought to be very relevant to that search query. 

Outbrain is the leader in content discovery, not content search. What we do is most akin to the flipping through the pages on your favorite magazine or newspaper. When you finish reading an article in a newspaper and flip to the next page, you don't have the expectation of high contextual relevancy from one page to the other.

In fact, the more contextually relevant the next news story is, the less interesting you'll likely find it to be. When we flip through the newspaper we want to go from the most important news stories, to finance, to entertainment and to sports. 

When we build algorithms at Outbrain, we try to focus on that metaphor. What would be the next most interesting story to read if this user were flipping through their favorite print magazine or newspaper? 

How do you measure the performance of content, and encourage clients to improve quality? 

We look a lot at the overall engagement a user has with the content we send them to. For example, did the user stay on the page we sent them to for a time long enough to read the article? Or did they come in and quickly leave? Did they stay to read/watch more stories after the link we sent them to? Did they come back to the site later? 

All those metrics matter a lot for informing us on whether a recommendation we made was really good, or whether it was just a clicky title.  

The best way we encourage clients to improve the quality of their content is through the pricing mechanism. We constantly encourage clients to pay us *less* by providing us more content and better content.

If the content performs well with users, they can afford to pay us less and still attract the size of the audience they're looking for.  

What are the biggest challenges for you? 

The biggest challenge we have involves user trust, which is the fundamental currency of everything we do. At Outbrain we spend a lot of our waking hours making sure that the links we recommend to people are trustworthy, authentic, properly-disclosed content.

I strongly believe that if you care greatly for the users and their trust in these units, then over time the financial rewards will present themselves. But once users' trust is violated with bad, non-trustworthy recommendations, the whole rug is pulled from under this category's feet. 

Many of those that have cloned Outbrain lately do not share that fundamental belief with us - they'll prefer pocketing the $$'s they can get from a shady marketer, even if that means serving the user with shady, non-trustworthy links.

The biggest challenge for the whole category that Outbrain has created, is that people stop trusting content recommendations and become blind to them like they did to so many other forms of linking and advertising in the past, when those were abused by the short-term thinkers.

It takes a long time and great effort and caring to build trust in a format, but can be very quick to erode that user trust if you're not careful.