Reimagining scalable feed control on X
Independently built and shipped two features from exploration and ideation to execution.
Overview
In this solo project, I explored how X (formerly Twitter) could enhance user agency by giving users more control over their content interests and reducing feed clutter. To address this, I led the effort to improve the agency-driven product experience on X by addressing the user problems embedded in current X users.
the problem
Where does current social media apps fall short in meeting user's needs?
The more an algorithm knows and decides, the less a user needs to, and the platforms that chase engagement metrics most aggressively tend to be the ones that ask users the fewest questions about what they actually want.
user research
5 in-person interviews later, the initial findings are…
Poor algorithmic recommendations and lacking interest segmentation are the universal pain points among the participants.
Ghost interests lingering their feed with irrelevant content also produced their product experience unproductive.
🦻 Affinity mapping
Synthesizing the interview results revealed that….
Users generally crave control but also a clear split in user behavior suggests that distinct mindsets require different targeted solutions. Any new feature needs to serve both without imposing friction on either
"Intentional" open X with a specific goal and prioritize efficiency, they find irrelevant content actively frustrating;
"Exploratory" users enjoy serendipity and algorithmic surprise, and are less motivated to manually curate their feeds
users crave control, but it has to be effortless
Context Collision Is a Core Pain Point
Users consistently preferred action over explanation
🙋 hmv questions
How might we design an in-context feed control experience on X that lets any user - casual or professional - effortlessly switch between modes and shape their feed intensity in real-time without disrupting their browsing flow?
🖌️ from research to ideation
With the problem statement in mind, design ideas were formed
translating insights into actionable design concepts, from feature set, sitemap, to task flows
feature set
must haves
calendar view
dashboard
basic period logging
pattern spotter
pattern insights
visualization & short analysis
predictions
smart / customization reminders
nice to haves
symptom severity tracking
free-text notes
goal settings
sitemap
task flows
task flow 1: editing historical data
Validate that the editing process is intuitive and users can easily navigate to historical data without confusion
Identify confusion points and users can easily distinguish between current and historical data while making accurate corrections.
task flow 2: logging a customized symptom
Identify friction points that might cause abandonment while ensuring complex tasks don't overwhelm new users.
Ensure the customization process is discoverable and straightforward without overwhelming users with complexity.
Based on current insights and to make the MVP lived in real-life experiences, both of these two core user tasks are related to the key actions that users might take to note down her their body variation during menstruation.
📒 iterations
Multiple Iterations later, from wireframes to
higher fidelity
My iterations started with low-fi wireframes to quickly explore layouts, and then moved forward to focus on content hierarchy, navigation and interaction patterns to flesh out.
low-fi wireframes
mid-fi wireframes
During the mid-fi phase, for the first task, I overhauled the homepage interface. The mid‑fi introduces clearer hierarchy and actionable controls, showing from card blocks to a structured card layouts. Overall it feels more scannable, and task‑oriented than the low‑fi one, which was more generic and blocky in structure.
The UI now supports the period‑editing task with explicit flows, with labeled states, and tappable affordances that reduce ambiguity and cognitive load.
For the symptom logging flow, the mid-fi one now feels more customizable, and confirmable than the low‑fi, with guided entry points and editable bubbles. Explicit save states that reduce ambiguity and increase confidence logging sensitive data.
It also aligns with a supportive tone and daily check‑ins by making customization first‑class without overwhelming the base categories.
🔍 High fidelity & prototyping
Applying the visual language to bring the wireframes to life
With the brand identity and design system in place, I put everything together into high fidelity wireframes by applying the visual language, which enhances accessibility and usability, thus bringing the design to life.
Final Prototype
view prototype
🎙️ usability testing
Making sure that the product works not just in theory, but also in users' hands
The usability testing mainly focused on out two task flows, and we also asked users to comment on our dashboard, calendar view and symptom screens based on their understandings, and if the information conveyed on each screen is clear, visible and easy to navigate.
Key Findings
Overall, users responded positively to the overall flow and layout but flagged a few points of confusion in language and interaction. Based on their feedback,
Participants wanted a stronger visual feedback that reflects their action states after completing each task.
Participants preferred a more intuitive and personalized experience, such as a customizable symptom screen to prioritize desired categories.
Low icon affordance caused frequent misclicks and confusion during testings.
🧠 takeaways
Key Learnings Recap
Using systemic design thinking to design for distinct user behaviors rather than an abstract "average" user on the surface level that serve no practical use to real-world people.
By turning research insights into a core design principle, I was able to create a feature that offers meaningful agency. This project solidified my ability to integrate new features into a complex, existing ecosystem while maintaining strict UX consistency and usability.