Tired of wading through clickbait? Filter.ly, launching today at Disrupt Battlefield, wants to make sure that website visitors only see the content they are most interested in. Many recommendation engines put the most weight on keywords or topics, but Filter.ly’s B2B service delivers more refined, personalized results by looking at a plethora of signals, including how much time a reader spends in an article and their social media activity.
Filter.ly has created a news reading app as a preview for its recommendation technology, which uses natural language processing, machine learning, and customized algorithms, but that is not the startup’s main product. Instead, Filter.ly’s goal is to make its relevance platform available to any online business, including content publishers and e-commerce businesses, that want to improve the user experience on their website by serving up the best articles, links, and products for each visitor.
Founded by Adrian Bethune, Angel Prado, Diego Val, Jake Adams, and Ismael Faro, Filter.ly is a customizable solution for companies that don’t have the resources to develop their own recommendation engines. The startup wants to serve a wide range of customers, starting in size from small sites like local newspapers with a few hundred thousand unique visitors per month. The company plans to charge fees based on the number of users and recommendation requests, so it remains affordable for small publishers as they scale up.
Many publishers currently use content discovery platforms like Outbrain, Taboola and Gravity, but Filter.ly seeks to differentiate by giving better personalized recommendations, which boosts return traffic.
The startup offers several different services, including a widget and an API that integrates with content management systems and personalizes a site’s content for each visitor by putting relevant links at the top.
Anyone who has been on a site with an Outbrain or Taboola widget knows that recommendations can often seem random and don’t change very frequently, which doesn’t keeep readers engaged. Furthermore, since Outbrain and Taboola also serve as ad networks, that means visitors can hide them by installing ad blocking software, which impacts the bottom line of publishers.
To create a better experience, Filter.ly’s recommendation engine uses two modes of learning. The first is designed for people who visit a site regularly and have created an account. If that account is linked to a Facebook or Twitter profile, Filter.ly analyzes posting history and other public data to create a personal relevancy graph. The second mode looks at what types of articles a visitor spends the most time reading, then combines that with the most popular content on the site to suggest links.
One of Filter.ly’s key advantages is that it learns users’ interests and then combines them, instead of serving up links based on keywords alone. For example, veterinary science and tech startups might seem unrelated, but if someone spends a lot of time reading about both subjects, Filter.ly will start looking for links where the two overlap, such as veterinary startups.
Filter.ly is designed to serve not only a site’s regulars but also casual browsers who have only visited it once or twice. Even though there is less information about them, Filter.ly can look at data like what articles they look at, how much time they spend scrolling through it, or if they interact with it by clicking on links or reading comments. Filter.ly then delivers personalized recommendations to them each time they return to the site, with the goal of turning them into loyal readers.