Recommendations are a huge part of how people make decisions online. Who to connect with on social media sites; which products to buy on an e-commerce site; or what content to read next on an online publication all depend on recommendation engines. New startup Algo Anywhere is providing an API-based recommendation engine companies can implement in order to provide relevant, targeted recommendations to their audience to help drive engagement and, ultimately, revenue.
Algo Anywhere founder Zach Aysan and co-founder Adam Gravitishad the original idea after realizing too many data libraries were stuck in languages that people weren’t using for their products. The original idea was to wrap those libraries in APIs, but they quickly realized that while they liked the idea of solving people’s problems with APIs, they didn’t like offshoring the solution to preexisting libraries. “I was talking about how AI is the future, but machine learning guys like Adam and I were constantly getting stuck working for companies like eBay or Amazon. There needed to be a company that just productized AI,” Aysan said in an interview. Gravitis suggested a recommendation engine, as both he and Aysan had built them before internally for other companies.
The company’s flagship platform is the Recommender Algo, which delivers high-volume personalized recommendations in real-time. The hosted SaaS solution is available on major cloud services on a subscription basis. Algo Anywhere’s target audiences are publishers, e-commerce operators, and social media platforms. Aysan says those verticals make sense because a site needs to have enough data for the algorithm to work most effectively, and because a site should primarily make its money off of what they recommend. “Their bread and butter is getting the right thing – article, product, or person – to the right person.” The practical advantages of the Recommender are straightforward, unlike the technology behind the platform. Publishers can improve their content recommendations to readers, so that rather than just recommending articles by the same author or with similar keywords it understands they types of content an individual reader should be served. E-commerce companies can use the Recommender to maximize profits by suggesting targeted items to shoppers and providing other ways to encourage additional purchases. And the Recommender suggests new connections to users on social media sites, while also promoting user-generated content.
For publishers who are already using in-house recommendation tools, Aysan says their solution is better than anything in-house developers could put together. But he admits that there are existing solutions that might be more attractive to publishers from a revenue perspective. One of those solutions is Outbrain, a content discovery engine that allows publishers to make money from the content recommendations on their site. “Some other solutions out there may be a better fit for publishers, it depends on their cashflow and objectives,” he said. “Outbrain is doing great because their proposition is so easy: one in five of the recommendations are ads and they’ll share the revenue with the publisher.” The Algo Anywhere revenue model isn’t based on ad revenue, and Aysan says he doesn’t think it’s the right solution in the long term. “We think that a recommendation engine that aligns against an organization’s KPIs will be much more useful in the medium to long term than a quick fix. Obviously every company selling recommendation engines is going to cite that their algorithms are better, but ours really are good.”
There’s no set pricing structure for companies looking to implement Algo Anywhere. Ideally Aysan said they’d like to take a share of the increased profit, but they recognize that won’t work for companies that rely on predictable costs. They’re currently working with a group of early beta testers to iron out any issues before locking in a pricing structure. “To test the recommendation engine we take their data sets, rewind time, and see where items that have been liked, favorited, or purchased would have fallen on the ordered recommendation list,” Aysan says, and reports that the system is currently 20 times more effective than random recommendations.
Algo Anywhere isn’t the first company to try to capitalize on recommendations. Recommendation tool Hunch launched in 2009 and was acquired by eBay in November 2011. It aimed to “personalize the Internet” through its Taste Graph, which provided recommendations to users on the company’s site, and through use on third-party sites and apps. At the time of their acquisition the team said they would help eBay develop predictive merchandising tools. Aysan says he feels Hunch had its flaws, and the two companies target different audiences. “We sell to business trying to optimize their profit or other KPIs, Hunch was a consumer web play that ended in a talent acquisition.” But there are other companies aiming their sights on improved content recommendations for businesses, including Rummble Labs’ Social Predictions tool. For publishers, e-commerce companies and social media platforms looking to up their engagement and profits, the more companies competing for their business the better.
