Agripredict — Case Study | UX Design, UI Design
Background
Agripredict uses machine learning to provide Zambian farmers with access to information that will drastically help to identify diseases and predict pest infestations and weather conditions.
My Role
User Research, Prototype, User Experience Design, User Interface Design
What were the frustrations with the farmers’ day-to-day activities?
To provide the farmer’s best user experience, it was essential to clearly understand the current farmers’ problems while using the app. We began with research feedback from the pilot program.
- Our target audience is less tech-savvy
- Poor connectivity, low-speed internet
- Illustrations were misinterpreted
- Farmers wanted to locate agro-dealers
Design
To address the challenges discovered during research, we utilised the human-centred approach process to gather feedback on existing and potential app features, test prototypes and craft various app components.
- Discover: Remote stakeholder meetings to understand the business needs and identify pain points of the farmers’ ecosystem
- Define: Farmer’s experience map through the pilot program, user testing of the current app
- Design: New user flow high fidelity prototypes, interactive design prototypes
- Prototype: User testing of clickable design prototypes, focus groups to test new UI and language
- Plan & Implement: Detailed user flow, complete UI kit design system, exportable assets for developers
Based on user insight gathered, it was clear that our users struggled to understand the illustrations’ concepts. Also, most of the farmers are not tech-savvy. These findings made it clear that our designs should only feature the necessary elements.
Onboarding
Therefore, we used pictures for the onboarding flow to give a quick overview of the app. This is because our target users were more inclined towards visuals.
We ensured the process and design assets are simple and to the point for the app’s primary use case. The farmers can quickly scan their crops, receive a diagnosis (plus accuracy percentage) and suggested treatment. Also, farmers get a list of agro-dealers that can provide materials for the treatment.
While the app was designed to help detect crop diseases, pest and suggest treatments, farmers’ discussions revealed other challenges in their daily activities. This led to additional features such as weather forecast and advice and farming learning content.
Conclusion
For any ordinary person, anything related to machine learning seems abstract and cryptic. This project provided an insight into how smallholders interact with mobile apps and emerging technologies.
The subsequent changes to the user experience and interface will make the app more engaging, easier to navigate, understand and adapt to — consequently increasing the Agripredict app usage. The intuitive user experience will also reduce the time spent on training the farmers.
