Since launching
Steam Labs, Community feedback on our initial Store discovery experiments has lead us to create a number of new features and improvements we're excited to share.
What we've learned
We've received really positive feedback since launching the
Interactive Recommender. We've heard from many of you that the Interactive Recommender is helping you find interesting games, and we also see this reflected in our early data. One way we study what’s interesting to users is to look at how frequently a visit to a store page turns into a positive action like adding the item to a wishlist, or purchasing it. That frequency varies depending on how users arrived at the store page. We also look at how frequently people choose to visit a store page via the recommender.
Our initial data show the Interactive Recommender is performing very well by those measures. We do of course take these signs with grain of salt, given the novelty and promotion of our experiment likely make for an unfair comparison. Next up, we will work to evaluate the recommender in ways that eliminate this potential bias.
Furthermore, we're especially pleased to see that users are being exposed to a broad range of titles. In fact, nearly 10,000 different games have been added to wishlists from the Interactive Recommender page so far. So yeah, initial signs indicate the Interactive Recommender experiment
is working!
Episode 1 of the
Automatic Show, a half-hour algorithmically-generated video about Steam games, was received by the Community with a more mixed response. While much of the feedback we received has indicated that the show's utility and format have promise, we did hear from many users that 30 minutes is...
a lot of minutes. If this rings true for you, we now offer
three new short variationsof the initial half-hour experience that we hope you'll enjoy.
If you're in search of your next favorite game, you can give the Interactive Recommender a try for yourself at
https://store.steampowered.com/recommender/ and watch the new short n' sweet Automatic Shows at
https://store.steampowered.com/labs/automaticshow.
What's new
The Interactive Recommender
New "exclude" feature
You can now tell the interactive recommender to
exclude some of your recently played games when generating recommendations for you. By default, the set of games that are excluded in this way is taken from your global Steam ignore list
here. If you've chosen to ignore a game via its store page, we won't use it to generate recommendations for you, but you can now interactively toggle excluded games on and off to see the effect that has on your recommendations. (Toggling the exclusion state of games via the interactive recommender in this way will not affect those Steam-wide ignore settings.) If you want to ignore a particular game across all of Steam, you can set that on the store page for that game.
Let us know how this feature works for you!
User interface improvements
Your feedback has also informed a few minor improvements to the interface. For example, hovering over a played game in the left-hand column now displays the title of the game, which is useful when the thumbnail art itself is not so legible, and clicking a game in the played list now navigates to its store page, as one might expect.
Always training
The interactive recommender model adapts in two ways. First, it adapts right away to an individual user's behavior; as you play new games, or revisit old ones, the model uses that data to give you updated recommendations. The second way the model adapts is by periodically re-training itself to take into account global changes, staying up-to-date with the latest releases and the gradually changing patterns of player behavior. This re-training process is an intensive operation that crunches billions of data points and can take a whole day to complete. We've been doing some behind-the-scenes cleanup work to make the re-training process smoother and more automated, which will enable us to use the technology in new contexts, like other Labs experiments or the Store itself.
The Automatic Show
New short shows
- Top Releases for June, covering 21 titles from our monthly roundup in a tight 2 minutes 34 seconds.
- The 3-Minute VR Show, which covers some of the latest VR titles across all genres.
- Rapid! Fire! Horror! In the first two shows, each game clip is 8 seconds long. This show experiments with 3-second clips. Can you handle it?
We hope you'll
check out our new shows, then
let us know what you think in the discussions.
What's next
We've heard your requests for more dynamic tag selection tools to help guide the Interactive Recommender's results, and we hope to build these soon. We will also continue to monitor and improve the Interactive Recommender's success connecting users with compelling content. The strong performance we're seeing so far may be due in part to the novelty of the feature, so we're continuing to monitor results and conduct additional tests to confirm our initial findings. Meanwhile, given your positive feedback, we're exploring ways to offer the Interactive Recommender's features in other parts of the Store.
As we're looking at next steps for each of our fledgling experiments, we are also embarking on a fourth, which we're excited to share with you soon.
Keep in touch
These experiments are guided in large part by the Steam community of players and game developers. We love to hear what you think of them in the Steam Labs community forum at
https://steamcommunity.com/groups/SteamLabs/discussions/
