Building a more trustworthy plant care app through community connection — a UX case study
Context
Tasked with exploring and identifying a current product, service, or mobile application that could best benefit from utilizing Augmented Reality (AR) in a way that makes it more helpful for its users, I set out to investigate whether garden planning apps could profit from an AR boost — specifically in regards to spatial planning and maintenance of garden beds.
However, once I started conducting user interviews, it became clear that AR garden planning was not a product my interviewees were clamoring for. As a result, I pivoted to exploring how computer vision (CV) could offer a more fitting solution.
Goals
- Develop a concept that brings value in a unique or effective way.
- Create an iterated, clickable prototype based on feedback within 7 days.
Design Process
Research
I interviewed 6 people across the continental US, ranging in age from 27–66. All were actively caring for plants and identified somewhere on the spectrum of novice to expert gardener.
Questions were aimed at assessing gardening interest, experience, and process as well as comfort with — and current reliance — on technology.
Synthesis
As it happened, the subjects I spoke with were not concerned with planning their gardens efficiently, and while some thought AR could be a useful tool in planning outdoor gardening spaces, most confided that they tended to put more energy, effort, and care into their indoor plants than their outdoor plants.
I sorted through the feedback I received using an affinity map and pulled out some additional key insights:
- All of the users I spoke with care for indoor plants year round and expressed frustration at how difficult it can be to accurately assess a plant’s health and decide how to care for it.
- Users rely on Google, other internet resources, and people in their network for plant care guidance.
- Users have a significant emotional attachment to and investment in their plants.
- 3/6 users expressed being distrustful of new technology.
Identifying the Problem
Home gardeners of all skill and interest levels need a better way to receive on-demand, informed, and personalized care advice for their plants because relying on Google and crowdsourced information yields questionable and oftentimes frustrating results.
To address this problem, how might we leverage computer vision via a mobile app to act as a pocket garden guru–offering personalized plant diagnostics, care tips, and progress tracking?
More Research
After identifying the problem, I needed to see if there are any apps currently in the market claiming to offer a solution — and, not suprisingly, there are! However, most focus on plant identification only or on helping users keep up with a watering schedule, requiring users to download multiple apps and toggle between them. One user mentioned this frustration in our interview and also complained about frequently getting incorrect identification results.
Between their identification, diagnostic, and maintanence features, Planta comes closest to meeting my users’ needs. However, none of my users ever felt compelled to download or use it.
I learned in our interviews that my users are hesitant to trust tech as their only source of information so it was important to me that computer vision not be the only reliable source of information in this app.
Building a social component into the app that enables users to easily find friends and other plant parents experiencing similar issues with their plants allows users to validate the CV diagnoses and suggestions amongst themselves or find alternative care options if they are experiencing an issue the CV cannot yet solve.
Sketches & Iteration
I started with some rough sketches of key features I wanted to explore and generated user flows tracking how a user would move through those features. The user flows below map out how a user would take a picture of their plant, identify it, diagnose any health issues it may have, and save it to their user database, or “garden”.
From there I drew up a paper prototype and conducted a low-fi usability test. The user was asked to sign in, take a picture of their plant, ask the app for a diagnosis on their plant’s health, and add their plant to their garden.
I received feedback asking for:
- A search feature in the user’s Garden page to easily and quickly find a plant they are looking for.
- Less reliance on text and more reliance on iconography to convey diagnoses and care information.
- An explanation of how the technology works.
Low Fidelity Clickable Prototype
I incorporated the feedback from the usability test into a clickable low-fi prototype.
Ideal Next Steps
- Conduct usability testing on the clickable prototype.
- Develop more robust in-app community connection opportunities to supplement & validate app-generated diagnoses and advice.
- Design and test a feature that allows user to share their plant care instructions and routines with plant sitters.
Final Takeaways
The users I spoke to want help keeping their plants alive and thriving but often don’t know where to turn to for trustworthy advice. Despite the many plant care apps in the market, my sample subjects were not compelled to rely on any of them for consistent support. Computer vision will not be able to reach its full potential so long as users refuse to engage with it because they feel it is untrustworthy, suspicious, or inaccessible. One way to build up trust in computer vision and artificial intelligence is to allow its findings to be fact-checkable by humans in real-time. In the case of this plant care app, whether the CV returns accurate information or not, our users can get the answers they are looking for — one way or another.
