[Experiment] “Category Suggestions” Panel in Stream Summary
Deciding what categories to stream in can be challenging, especially for smaller creators trying to reach new viewers. We’re experimenting with a new panel on Stream Summary to show categories with a favorable viewer to channel ratio that may increase your chance of being discovered. Creators randomly selected to be in the experiment will see this panel on their Stream Summary page in the Creator Dashboard.
Suggestions are generated by accumulating hour watched and hour broadcast data for the previous week for each category, and removing outliers like large partners who throw off the distribution. We’ll continue to make changes and refine our suggestions over time as we learn more about how this helps creators grow. We look forward to hearing your feedback about how we can improve this feature! Learn more here: https://help.twitch.tv/s/article/stream-summary#experiment
Only suggestion I have so far is that the "most recent stream" is not actually reflective of the most recent stream and only of whatever the currently selected category is in the stream manager. It doesn't actually reflect what was streamed last. It would be more accurate if it based it off the last category used for the specific stream date being viewed in the summary.
I.e. if I set Warzone as my category before the next stream and my last one was Halo, the most recent stream will say I stream Warzone last when I haven't yet and should read that Halo was the last stream.
My idea was combined with this one -- so I want to be sure it doesn't get lost, as this experiment is about being discovered in high-potential categories, and less so about trending categories for any viewer type and, most importantly, an adjustment to the speed of What Categories Do My Viewers Watch.
This experiment for Category Suggestions says it "should refresh daily". I hope it's possible to apply more of this urgency to the "What Categories..." data and beyond. Again, looking at the last 24 hours (or sooner) is important, or likewise having any way to catch potential trends.
We need at least a few solutions to understand trending categories and games on Twitch. In the broader view, we need to see categories that suddenly gain "trending" or "large increase in engagement" statuses. An example would be if a new Beta, DLC, or release was drawing sudden, big interest. Right now, all we can do is try to look at the top-viewed categories on Twitch but this doesn't provide any information on trends.
More specifically, to the idea subject here -- for streamers trying to understand their current followers and viewers, we need quicker data. Right now it seems to be on a 7-day cycle. That means, for example, if Among Us released and blew up with your audience, you wouldn't know for a week if it was a good idea to be streaming it. Another example, if there were drops for a game like Apex Legends and your audience was suddenly consuming a lot more of the category, you wouldn't know before it's too late.
We need at least 24-hour refreshes to this information. But even better, any information we could get on trending alerts or behavior for our audience. That could include more real-time type alerts, as well as showing more detail in the importance of what your audience watches. For example, maybe a portion of your audience watches a certain category for 2x as long as they normally do other categories.
In sum, we just need more info to jump onto audience behavior. That includes audience we don't have yet, as well as the followers, lapsed viewers, and active viewers.
This is a great idea and I am always very happy to see discoverability tips and anything relating to discoverability coming to twitch.
I do think that an "average viewership per streamer" statistic, which excludes streamers above a certain viewer threshhold, could also be a helpful stat for determining category viability. As it stands right now, the viewer to streamer ratio is affected very strongly by streamers who are dedicated to each game. For example right now the Super Metroid Category has a high viewer to streamer ratio. So for example Super Metroid has a pretty decent viewer to streamer ratio, despite most small streamer not getting much engagement. This is because it is a popular game to speedrun and as such these speedrunners dedicating their time to the game hold most of the viewers and skew the numbers despite not being partnered in a lot of cases.
So far I love the concept. I'd suggest fine-tuning it to show categories that have a large enough audience to grow. Currently a surprisingly large portion of suggestions seem to have less than 100 viewers with the majority of those viewers only being there to watch one streamer. It doesn't feel like there's a lot of discoverability.
It's a good idea, but should probably be focused on streamers that stream a lot of different games, and not streamers like me, that only streams one game.