Founded by Brian Whitman and a team of ex-Spotify employees, Canopy has developed a recommendation engine that uses on-device machine learning and differential privacy to deliver personalized content recommendations without collecting or storing any personal data. Through our research-led process, we collaborated closely with Canopy's product team to concept and design the first app built on this technology — launching, testing, and iterating with a growing group of Beta users until the app was released publicly, as Tonic.
Before we could design Canopy's first app, we needed to define the problem that it would solve. Through our research process, we conducted a combination of interviews and surveys and formed a clear picture of our target audience — their behaviors and attitudes towards content recommendations, algorithms, and data privacy — that would inform how to productize this revolutionary technology.
Most of the people we talked to during our research process get their content recommendations through social media. But the endless glut of content is taking its toll, and many are beginning to question its effects on themselves, their health, and society.
Plagued with predictable suggestions, opaque controls, and clickbaity titles, algorithmically recommended content can range from "terrible" to just "pretty good." Even good recommendations start to suffer when there is no element of surprise or unpredictability.
What we heard was not only a growing discontent over the amount of data that is collected, but also a sense of powerlessness over how that data is being used. For some, this has led to a sense of apathy for some, and for others, concern.
Whether concerned or apathetic, we learned that most people are now taking some steps to protect their data. But, outside of a few design patterns established the big players like Google and Facebook, there is no clear template for what data privacy looks like. Most of the people we surveyed confessed that they don’t actually know how to protect their private data, and feel that nothing they can do will make much of a difference anyway.
Using what we learned from our research, we began ideating on the product design. With Canopy's engineering team focusing on how the technology would work on the back-end, our job was to imagine all the ways a private recommendation engine could work in the user’s hands. We tested rough concepts to gather reactions and validate our ideas.
Although people were initially skeptical of the technology, the concept for a bite-sized, thematically-curated, daily mix of content resonated with those for whom the endless scroll was becoming a daily grind, so we focused on simplicity and putting the content first, forgoing the technical explanations and radical transparency that we'd initially thought would be our key differentiators.
Once our concept had been established, we began architecting the user experience of the app in its entirety. Using wireframes, we defined the app’s information architecture, navigation, content, and interactivity.
Designing a product that forgoes the current value exchange gave us the unique opportunity to rethink some established design patterns. With an underlying theme of "designing a better internet", we leaned into Canopy's warm and friendly aesthetic. While powerful technology was driving this new recommendation engine, the experience needed to highlight the human side of connecting people with great content.
Our design team worked as an extension of Canopy’s product team. Throughout the Beta program, we continued to concept, evaluate, prototype, and test new ideas and solutions as part of an iterative product release cycle until we found the right product-market fit.
Our onboarding process was designed to emphasize content over preferences, relieving users of the mental exercise of having to choose which topics they like and allowing them instead to be immediately immersed in interesting articles, building their own first playlist of content, which would then inform future recommendations.
Our Beta users loved the fleeting selection of content they received each day, but they didn’t want to miss out. One of the first features we added was a notification setting that allowed them to set a daily reminder at the time of their choosing. Whether they were a morning rush-hour reader or a late-night podcast listener, they wouldn’t miss out on their daily picks.
Although our early iterations taught us not to lean too heavily on explaining the technology, we still wanted to offer some transparency into how the recommendation engine works and give users a sense of control that went beyond the usual, ineffective mechanisms. Through ideating, prototyping, and testing we explored a variety of concepts that lets users see how their past activity informs their future recommendations, and gives them the ability to adjust it.
Our Beta users provided invaluable feedback through surveys and user tests, helping us to design and refine the product the world now knows as Tonic. With our iterative research-lead design process, we were able to help Canopy launch a product that resonates with users, while confidentially deviating from common product patterns for the sake of a delightful privacy-focused experience.
The Tonic app now continues to gain enthusiastic users and traction, as Canopy continues to develop a privacy architecture for the web.
Read more about how we went from Zero to Beta using a research-led product design process, or download the Tonic app for iOS from the Apple App store.