Content Recommendation Engine Review: Outbrain


For the past two weeks, I've been testing Outbrain recommendation engine at this site. The screen shots on this post are from the dashboard documenting the period this site was included in the catalog so I could review part of the product.

Used by almost 1,000 brand publishers in the U.S. and Europe, New York-based Outbrain does links a bit differently than the rest of them.

Instead of basing its algorithm on similarity of topics, serving up stories that are "like" those you are reading, it looks at how likely it is that you're going to stick around on the site. That is the real goal of publishers.

When it comes to sticking on sites, it turns out context doesn't perform as well as other variables like behavior and popularity. This is consistent with my experience at this site. Whenever I add resources on the same topic at the bottom of posts, they don't get a high click through rate.

Outbrain's algorithm measures reader engagement with a site after the click through. How many pages they view from the original page. Then, ranks its catalog of hundreds of thousands of stories based upon the results.

The company classifies its content recommendation algorithms into four major buckets:

  • Popularity: that is trending up in popularity on the site
  • Contextual: related to the page the person is currently on
  • Behavioral: based on audience dynamics. E.g., what people with similar reading habits have read, that is not mainstream popular, and that the person has not read before
  • Personal: within broad categories the person reads and not necessarily related to the page they are on

The metrics they evaluate against are click through rate (CTR) or how frequently people click on links based on the algorithmic approach, and page views after click or how many pieces of content the person visits on the site after clicking on the link.


This is consistent with the observation that when it comes to information, we prefer to discover it, even when it's news, than being pitched with it. Discovery connects with our points of interest and curiosity and is one more reason why PR is about public relationships.

It looks like AI agents as discovery channels is here.

There is a second piece to the tool, a widget, which I didn't get around to installing here (why I said part of the product), which recommends stories on the site in the left column, and stories from Outbrain's catalog in the right column.

Outbrain clients pay to promote their content as sponsored recommendations throughout the network of brands and companies that buy the service. The recommendation engine are drives which to pull and post based on the algorithms.

[Note: FTC regulations apply to product endorsements, like the affiliate links you see on the sidebar here. If you buy those guides, I receive a monetary compensation for recommending them.]

How is sending people to other sites going to help them stick around?

Write and link to helpful content that answers questions potential customers have. 

Any and all marketing can be designed to provide an engaging and useful experience, even as each part helps you address a different stage in the buyer's information needs.

When developing content assets for digital media, you should think more about distribution — how it is publicly shared, built upon, linked to, quoted, etc.

Also remember that although we'd like to think we make individual choices all the time, the reality is we tend to go with what others are saying and doing.

Which is how having widgets on many sites works. Twitter feeds, RSS widgets are other examples of creating a distributed presence. In turn, the widgets send back data that in aggregate provides a fuller picture of what happens to the content across the Web.

Combine access to a source of useful content with discovery and recommendations based on the preferences of others, and you have a compelling reason to go back to that site.

It starts with the content, then picking the right tool.


Last December, the company closed a $35 million round of funding, bringing its total financing to $64 million.

How does the company make money? Its model is based on a cost-per-click for sites to be listed and included in the catalog. Host sites, or the brand publishers, get a recommendation engine that serves up the content on their site most likely to keep readers engaged.

According to Outbrain, sites get an average 5-10% lift in views. They also get a revenue stream when readers click over to other sites through the widget.


To submit a question or a program use this link.


For a proposal delivered with confidence, contact me today.

Leave a Reply

Your email address will not be published. Required fields are marked *