At Antenna Labs, we created an analytics tool for Twitch broadcasters to deliver insights and personalized recommendations for their content, so that they can make informed decisions about their businesses instead of relying on guesswork. We began with user research and prototype testing in the spring/summer 2019, leading up to our MVP launch in August to an initial cohort of broadcasters. Through subsequent user testing and feedback, we iterated on the existing feature set and added new visualizations, leading to a retention rate increase of +32% for our second cohort.
Twitch provides their own channel analytics to broadcasters, but focuses primarily on raw numbers and often shallow metrics. Even the stats expected to be reported on, such as "average viewers" do not give broadcasters the full story - of those viewers, how many are new to their channel? How many can be relied upon to return stream after stream? How many are fully engaged members of their community? Answers to these questions can inform the broadcaster's business strategy.
Our product targets serious broadcasters who approach streaming as a full-time or part-time business. They want to grow their stream to a point such that it becomes a reliable source of recurring income, but the available tools that deliver their stream data leaves them in the dark as to how to approach this vision.
Twitch channel analytics
Using Twitch’s API, our approach turns raw numbers into data visualizations that reflect their stream in ways they're already thinking about their business. - growth (the direction viewership is moving), engagement (viewers active in the community chat), and conversion (subscribers). We also provide a detailed breakdown of how the games they impact their stream.
We launched this MVP in August 2019 with this set of visualizations, with personalized recommendations to follow. Keep reading for a walkthrough of the product design process.
Antenna product screen
Research & Prototyping
Our MVP objective was to release a product quickly that we could then learn from and iterate on. To begin, we conducted user interviews and ran two rounds of prototyping over the course of a month. These prototypes were tested with broadcasters who would become a part of our initial cohort, with their real user data.
Our primary prototyping goals was to test the presentation of data in order to learn which types of insights broadcasters found most valuable and actionable. We didn't spend much time worrying about layout or visual design problems just yet.
Prototype 1 - screenshot with live user data
The results of the first prototype test led to a 2nd prototype iteration, again tested with real user data.
Prototype 2 - wireframe
Prototype 2 - screenshot from prototype test recording, using live user data
After the prototyping phase, the team and I worked together to define requirements and scope for our MVP.
Layout and Feature Definition
Visual development in progress
Since dark mode UI is heavily preferred by our target market, we designed a dark mode from the start. This also takes into account situational context - if they are using Antenna while broadcasting, it would be distracting to have a light UI blast their eyes among the dark UI every other streaming tool uses, and may also interfere with their lighting setup.
Visual development in progress
After testing several color combinations, this set became our primary colors for the business metric visualizations. Colors were tested for accessibility and contrast levels to ensure readability.
Style library sample
Style library sample
Throughout the MVP development, base styles and UI components were defined and set up as resusable styles.
The Insights dashboard launched as our MVP in August 2019. We built this in about 1 1/2 months.
Post-Launch Feature Development
Shortly after launch, we added 2 new features that didn't make it in time for the MVP, but that we had hypothesized would be valuable based on our user research.
A timeline view of each individual stream could be navigated to by drilling into any stream day on the bottom bar chart.
This feature excited broadcasters from the get-go, such that they kept asking for more data about their streams. So shortly thereafter, we enhanced this view with even more detail (and rechristened it the "Stream Drilldown" view since it was no longer only about a timeline).
This view quickly became one of the highest valued areas in our product, but it was difficult for users to remember how to get to, so we elevated a portion of it to the main dashboard.
"What game should I play?" is the question most asked by game broadcasters. So we added personalized game recommendations, tuned to the individual broadcaster based on genres they often play and what their audience is interested in.
Recommendations are categorized based on games their viewers also watch in other channels, games that are performing above average for their channel (which means these are games they should continue to invest in), and new releases that are similar to games they already play.
Based on feedback from the initial Game Recs release, we iterated on the design to add more information about each game. Boadcasters can review the game description, genre, and platform here, so that users don't have to google for it separately if they are unfamiliar with a title, and so they can evaluate whether a game fits their current play style. (This iteration, however, was not able to be implemented.)
The MVP launched to an initial cohort of streamers in August 2019, netting out at 24% retention - our key metric at this stage. After adding subsequent features (the stream drilldown, personalized recommendations, as well as light modifications to Insights page) over the next few months, we then onboarded a second cohort of streamers which resulted in 56% retention (a +32% improvement).
These designs have been prototyped, tested, and (mostly) finalized, but were unable to be implemented. Based on the MVP feedback, we designed a brand new Leaderboard feature, as well as overhauled the main Insights dashboard.
Broadcasters are competitive! One of the unique things Antenna does is segment streamers according to games they play, size of community, frequency of streaming, and other attributes. This is a unique approach compared to publicly available trackers, such as twitchmetrics.com, that rank all Twitch streamers against each other. This is because a broadcaster with a small but loyal community who streams a couple days a week shouldn’t regularly compare themselves to a broadcaster with a huge viewer base who streams every day for 8+ hours. Looking up to other broadcasters is fine and great for creating aspirational goals (though not all streamers want to grow huge), but is not practical on a day to day basis to set achievable targets for themselves. Their respective strategies would necessarily be very different at those two scales. Instead, our approach is to identify and compare how our users are doing with respect to other broadcasters in their peer groups.
MVP - Rank section
We wanted to expand upon this valued feature to add the clarity and detail users were asking for. As we developed the next iteration and went through several rounds of prototype testing, this evolved into a leaderboard. The leaderboard would display how broadcasters stack up against others in their segment in concrete ways, which was easier to comprehend than the more abstract notion we started out with in the MVP.
Leaderboard wireframe sample that was tested as one of several prototypes
Final design for a v1
By now, we’ve also gathered enough data to overhaul the main Insights page beyond the MVP version.
Updated Insights Dashboard
This revised design was reported to be less dense and easier to digest compared to the current product, while at the same time increasing the amount of detail and data available (something our users consistently asked for!)
© Kim Kiser Ramirez 2020