Next Project >
Class Discovery Redesign at Skillshare
Helping learners choose the right class with clearer cards, previews, and CTAs.
Context
As a Product Designer at Skillshare, I worked on improving the class discovery experience, a key step in learner activation and retention. We focused on making class cards more informative and usable so learners could confidently decide which class to take.
Team
Jayden Kang
1 PM
3 Engineers
Timeline
4 Weeks for Research + Design
Live on Skillshare Member Experience
Key Outcomes
+47% Class Saves
in A/B Testing, resulting in full rollout
+5% Improved Click Through Rate
in A/B Testing, resulting in full rollout
Survey Design & Analysis
Created pre-design survey to identify information hierarchy.
Usability Testing
Designed and ran Maze test with 183 participants.
SUmmary to save you your time
tl;dr
Here’s a high level overview of the research and design process.
Please reach out to jaydehnajun@gmail.com for full case study.
Problem
Users make layered decisions when choosing a class, but the current experience only supports a shallow flow: card → full class. Key info is buried, leading to drop-off, misclicks, and low start rates after saving.
User Interviews
Across past and ongoing interviews, we heard a consistent mental model: users scan, skim for fit, then evaluate before committing.
User Survey
To validate and prioritize needs, I ran a survey (n=2277) asking users to rank class details. Results confirmed what interviews suggested, and this was helpful in informational hieracrhcy
Design
I redesigned the class discovery flow to better mirror how users think. Instead of forcing a jump from card to full class, We introduced a progressive model:
‍Card → Hover → Optional Modal → Class
‍Each step reveals deeper information aligned with what users need at that point in their decision.
Usability Test and Final Changes
We tested the new flow using an unmoderated Maze study (n=183). Based on the results, I refined interaction timing, added Save/Bookmark functionality, and clarified CTA language.
Experiment -> Rolled out to 100% of users
After a successful A/B experiment, it is now rolled out to all users on the platform.
Challenge
Users make layered decisions when choosing a class, but the current experience only supports a shallow flow: card → full class.
Identifying the user's mental model
What were the users saying?
In our continuous discovery interviews, we consistently heard that users make layered decisions when choosing a class
‟
“I immediately scan the thumbnails. Then I check the class duration. Time is always my biggest deciding factor.”
‟
“My checklist is: editor picks, student count, and then I test the first couple videos.”
‟
“I look for reviews and whether people actually finished the class. Then I watch the intro to see if it grabs me.”
Mental Model: 
Based on the interviews, I formed a comprehensive mental model that reflects the layers of decisions users make.
What does the existing experience support?
Auditting the Existing Experience
The current design forces users to commit to a full class page to find the context.
We evaluated the existing class card design against the user's mental model. Users scan first to see if the class is relevant and time-feasible. While the title, student count, and duration support that, key “Skim for Fit” questions go unanswered:
Existing Class Card
Commit to Class Page for reviews, projects, and skills taught.
User survey
Identifying the Priority of Information
User Survey
As part of a user research survey (n = 527), we asked learners to rank which class details were most vs. least important when deciding whether to start a class. The results helped us validate the informational hierarchy and guided decisions around progressive disclosure in the redesigned flow.
Design
Progressive Information Disclosure
Redesigning the Card
We redesigned the class discovery flow to better mirror how users think. Instead of forcing a jump from card to full class, we introduced a progressive model: Card → Hover → Optional Modal → Class

‍Each step reveals deeper information aligned with what users need at that point in their decision.
Forming Information Architecture by Mapping out Information to Journey Step
With the prioritization of information learned through the user survey, I mapped out the information to the steps of the journey map.
Default Card
1. Title
2. Materials/Skills ★
3. Level of Class ★
4. # Students Enrolled
5. Class Length
6. Rating ★

★ is new information
Hovered
1. Autoplay intro video
2. All Materials/Skills
Design
Usability Test and Iterating
Modal for decision making
I designed a modal that surfaces a clear class overview, highlighting key information users care about—such as the class trailer, project outcomes, reviews, skill level, and time commitment.
Usability Test (n = 183)
We tested the new flow using an unmoderated study. It was a 78.5% task success, mostly from unclear CTAs. Users loved hover previews and trailers, with the option to select the modal or to start the class right away, but wanted better CTA labels and a Save option.
Card for Usability Test
- copy for modal is "view more"
- no save button
- no level, student count, and class length
Updated Card based on Results
- copy for modal is "learn more"
- save button
- level, student count, and class length
Interaction Logic
I worked with developers to form the logic for the hover card, to reduce flicker and autoplay jump-scare. I also added a hover buffer so that the users don't accidentally close the cards.
Outcome
Results
A/B Test
The redesigned experience was A/B tested and showed clear improvements in engagement. Based on positive results, it has since been rolled out to 100% of users.
Looking back...
Reflection
Thinking about user decision making
This project reinforced that great UX isn't just about presenting information—it's about aligning with how people actually make decisions. By mapping the mental model and layering the experience around it, we created space for users to explore without pressure.
Collaborative Effort
It was also a deeply collaborative effort. I worked closely with our PM to translate research into product scope, partnered with engineering to fine-tune hover behavior and interaction timing, and synced with research to test early assumptions. Every design decision was grounded in shared understanding—and that made the process just as satisfying as the outcome.