Today we released the Steam Interactive Recommender, a way for users to harness the power of machine learning to discover personalized, interactive recommendations, based on your patterns of play. Along with powerful tag-based filters, you can tailor your results on the fly, selecting your own balance of popular or niche, and recent or classic titles, to find just the right games you're in the mood to play.

Recommendations generated by the system will appear on your store homepage. The Explore and Customize button leads to the full Interactive Recommender, where you can adjust parameters and save settings. Any customizations you make will also be used on the homepage.

Originally Steam Labs Experiment 002: Interactive Recommender, this new feature is now available to all users in the Steam store.



You can also access the Interactive Recommender by visiting https://store.steampowered.com/recommender, or find it in the Your Store menu.



The Interactive Recommender uses a machine learning model that is trained based on the playtime histories of millions of Steam users. It's not directly affected by tags or reviews—it instead learns about the games on Steam by looking at what users actually play. The basic idea is that if there are other players with similar play habits to you, who also play a game that you haven't tried yet, then that game is likely to be one you'll enjoy too.

We're also starting to apply the underlying model in other parts of the Steam store, where we think it can help players see the most relevant content or make more informed choices. For example, when viewing the page for a particular game, you may sometimes see "Players like you love this game" shown as a reason why the game is relevant to you, alongside other factors.

The Interactive Recommender isn't a replacement for our existing content discovery systems, but rather an addition to the variety of ways Steam recommends games to players. Although it's a powerful tool, there are some things it can't do. For example, it can't recommend new releases that nobody has played yet, while the Discovery Queue is designed to do just that. That said, we are starting to use the technology underlying the Interactive Recommender to power other features on Steam such as Steam Labs Experiment 008: Play Next, which recommends games you've already purchased but for whatever reason have not yet played. The result is a Steam experience that is more effective at connecting customers to games they'll love across a variety of scenarios.

The Steam Interactive Recommender was first released as part of our Steam Labs initiative. Feedback from visitors to the Labs helps us evaluate and iterate on potential new Steam features like this one. While in development, your feedback led us to add tag filtering and saved settings for even more powerfully-guided recommendations. We also looked at quantitative data, measuring clickthrough rates, and conversion rates to wishlist and purchases of games from both the home page capsule, and the full Interactive Recommender page. The data compared favorably with other Steam features, not just in the few weeks after introduction, but consistently over the following months, giving us confidence that this tool is providing long term value in helping users find games they enjoy. It's also pleasing to note that games found this way covered a large portion of our catalog, not just the most popular hits, with well over 10,000 different games purchased as a result of visits to the Interactive Recommender page.

Thanks to everyone who helped us refine the Interactive Recommender for release. Check out other ongoing experiments at https://store.steampowered.com/labs.

Try the Interactive Recommender today!