EdTech SaaS platform
Product design, user research, product management
Timeline
2023-2025
Team
Product designer acting also as UX researcher and Product manager, Engineers
Tools
Figma, Userbrain, A/B testing
Table of Contents
Improving the folder management system
This project focused on enhancing the folder management system within the educational platform. My main goal was to improve the existing interface to boost user navigation and organization, thereby aiming to drive growth and retention.
Research
Users reported difficulty efficiently locating folders within their workspace.
Initial research provided critical context for this problem:
• 50% of paying users utilize folders.
• A significant portion of user feedback (22.92%, or 58 out of 253 comments) concerning the "My Activity" page explicitly requested color-coding for folders.
​
Shaping and Design
During the ideation phase, several solutions were considered, including allowing users to pin/mark favorite folders, adding a list view, or creating sections/groups.
However, the chosen solution, which was considered the most prominent solution based on user needs, was Color Tagging (Color-coding).
The justification for color-coding was that it helps users categorize content (e.g., by subject, level, or topic), find specific content faster, and ultimately leads to a better in-class experience. Users (mainly teachers) already are using a color-coding principle in real life with their printed materials.
Design​
I have created designs for web and mobile resolutions, along with specific hands-off documentation for developers. The design specification also included details about the evaluation plan, so that we know what would be considered a success.​​
Validation
To measure the impact of the color-coding feature, an A/B experiment was conducted.
Additionally, we have established a success criteria on user adoption of the new functionality: 10% of newly created folders have a non-default color chosen.​
​
Key Outcomes:
-
Net Subscriptions: Variant B (Color coding available) showed a significant positive change of +6.6% in net subscriptions compared to Variant A, with a high confidence level of 96%.
-
Retention: This change contributed to lowering the overall churn rate by approximately 6%.
-
User Adoption: SQL analysis revealed that approximately 50% of newly created folders were colored by users using a non-default color.
-
The most commonly chosen colors included Red, Yellow, Green, Indigo, Orange, and Grey.
Results and Conclusion
Based on the comparison of the experimental data, the recommendation was to accept this change and implement folder color-coding for all users.
​
The positive result on subscriptions indicated that even a small design change that allows users to configure their work in meaningful ways can have a powerful effect on retention
Enhancing clarity and usability in interactive templates
Wordwall has various templates serving different educational needs. We were continuously conducting usability studies around game experience to iterate on improving the functionality always validated through A/B experiments.
​
The central issue identified was that the question at Airplane game is hardly noticeable by users, especially when the activity is played on mobile devices or when an image is used as part of the question​.
Research
User feedback and usability studies strongly supported this problem:
• During a usability study, 3 out of 5 users did not notice the question at the bottom of the screen. Users reported having to search for "the little stuff at the bottom" before understanding what to do.
• Feedback confirmed that the statement below the plane needed to be bigger because the current size made it difficult to read, especially for students in initial grades.
• Users also noted that images used in the questions were too small and were "barely be seen," making them almost impossible to use effectively as questions.
• Some feedback suggested that before the plane starts moving, a big phrase should be displayed all over the screen.
​
Shaping and Design
1. Show the question on a full screen and animate it to the gameplay, putting it into its current position. (This option was selected for implementation).
2. Make the question always visible in the background. (Rejected due to being theme-based and potential rejection if implemented across all themes).
3. Show the question at the top of the gameplay. (Rejected because players tend to look at the top of the screen for things like 'Submit Answers' buttons, which could be distracting).
4. Put the question on a large tile that is shown at the start of each question.
The selected approach involved displaying a "big question" at the start. This meant showing the question on a full screen overlay and then animating its transition to the normal position (at the bottom) during the start sequence. This solution was noted to have been already implemented for the Maze Chase template. Implementation required ensuring that the overlay layout was full-width (especially on mobile) and that the font size was appropriate.
Validation
The change was evaluated using A/B testing, comparing the current question view (Variant A) against the big question layout (Variant B). The change was rejected due to experimental results. ​​
​
Results and Conclusion
It was noted that while showing the big question worked successfully for the Maze Chase template, it did not work for Airplane. The proposed explanation was that Airplane is a continuous process game, whereas Maze Chase resets the maze for every new question, suggesting the new question display might be too disruptive or distracting to the player experience in Airplane.
​
We decided to run another iteration on this problem due to strong behavioural data from the usability studies. ​
Validation
The change was evaluated using A/B testing, comparing the bottom smaller question view (Variant A) against the bigger question layout on top (Variant B). ​​
​
Results and Conclusion
This experiment was successful and the change was accepted and integrated to all users.​​
Game speed optimization
There are various game settings to enhance interactive activities for the best education results and student needs. We were continuously gathering user feedback to iterate on improving the functionality always validated through A/B experiments.
Research
The speed control game option is used quite often. However, the team frequently received feedback suggesting that the existing speed settings are not optimal for various users' ages and needs.
Initial research provided critical context for this problem:
• 6/7 templates have their relative usage value as high as in 90th percentile
• User comments like "The speed is still too fast for the kids even adjust it to "1".", "Moles are very slow on the highest speed" etc.
​​​
Shaping and Design
From the heuristic point of view, the hypothesis was that users where also stuggling because of overload phenomenon. There are 10 speed options (from 1 to 10), and when selecting from a variety of choices, it's difficult to select the one that you really want/need, and user might not be happy with.
​
The ideation phase explored several options to solve the identified problem:
1. Simplification and Naming: Simplify the choice by having only "Slow," "Medium," and "Fast" settings, or changing the slider format. This was considered because describing speed with numbers doesn't clearly indicate the actual speed level or the suitable age group. However, this option was rejected due to concerns regarding potential issues with auto-translating the speed level names.
2. Decrease the Amount of Speeds: Reducing the amount of speeds (e.g., from 10 to 5) to address the choice overload phenomenon.
3. Increase the Range: Adjusting the range so that the slowest speed becomes slower, and the fastest speed becomes faster, based on user feedback that existing speeds were either too fast or too slow. This idea was rejected because it did not solve the choice overload problem and still required users to use trial-and-error.
4. Adaptive Speed: Introducing adaptive speed that decreases speed mid-game if the player gets something wrong, or increases it if they get something correct.
5. Auto Difficulty Setting: Deducing a good speed based on the content (e.g., longer content equals lower speed) or activity tags.
​
The approach selected for implementation and testing was to decrease the amount of speeds from 10 to 5. This solution involved combining existing speed states into a reduced number of options (e.g., current 1/10 remains 1/5, current 3/10 becomes 2/5, current 5/10 becomes 3/5, etc.)
The main hypothesis tested was that combining the current speed state into a smaller amount of options should improve game options choice and increase the creation metrics for the associated templates.
Validation
The change was evaluated using A/B testing, comparing the original 10 speed options (Variant A) against the reduced 5 speed options (Variant B). The experiments were run on three templates (True or False, Maze Chase, Ballon Pop).​​
​
Results and Conclusion
Upon conclusion of the experiments, the results indicated that neither of the experiments achieved confidence, and the data was generally leaning in a negative direction.
​
The change was rejected based on the experimental results. We decided to not to run another iteration on this problem and to focus on solving other problems.
Conversion and upgrades
The core issue was that the existing subscribe-to-use-this-feature modals were generic and failed to adequately explain the feature that required a subscription. As a result, basic users were not effectively informed about the value of the paid features. The project aimed to address this lack of clear communication regarding the benefits users could gain by upgrading.
Hypothesis
By redesigning subscribe-to-use-this-feature modals modals, giving each one a unique text and image, should increase subscriptions. The goal was to improve the upgrade-to-use modal to more clearly communicate the value of the features to users.
​
Shaping and Design
The selected approach involved creating Enhanced modals (Variant B). Key aspects of the implementation included:
• Updating the subscription required modal to be exclusive for paid features.
• Implementing the change under a feature switch.
• Integrating a system to add a reference query parameter (ref) to track which specific feature triggered the modal (e.g., ?ref=subscription-required-print or subscription-required-ai for features like 'Generate content using AI').​​
Validation
The feature was tested using an A/B experiment, comparing the Original modals (Variant A) against the Enhanced modals (Variant B).
​
The change was considered a small improvement that was "worth keeping live," as there was no evidence of harm, and the positive engagement signals suggested potential for more upgrades later on.
Leveraging social proof to improve search
There is a large library of publicly shared resource named Community. User has to be subscribed in order to use it. In order to improve search functionality and user experience of using the Community, a light-weight hypothesis test was conducted.
Hypothesis
Users are generally more likely to trust and interact with content that has been validated by others. By displaying the amount of likes, this would help to prove that social proof helps to convert.
​
Shaping and Design
The plan was to show the LikeCount on several key pages:
â—¦ Community homepage
â—¦ Template-about page
â—¦ Search-results pages
• The change was implemented under a feature switch.
• The scope of the change was defined to apply for both registered and guest users.
Validation
The feature was tested using an A/B experiment, comparing the Original modals (Variant A) against the Enhanced modals (Variant B).
​
The change was considered a small improvement that was "worth keeping live," as there was no evidence of harm, and the positive engagement signals suggested potential for more upgrades later on.
Other projects
Mobile applications design
Design for iOS, Android
Producing a podcast
Side project
Want to work together?
If you like what you see and want to work together, get in touch!