Redesigning the Tutor
Operations System for Online
Education
In the dual-teacher model, one instructor broadcasts to thousands of students while Teaching Assistants (TAs) manage localized pods. As Senior Product Designer at TAL, I led the redesign of Class Mentor — the operational dashboard these TAs run their classrooms on.
Background
TAs monitor engagement, manage real-time attendance, and run interactive exercises. The model's success depends on their ability to spot and support struggling or distracted students mid-broadcast.
As the user base grew, that operational overhead became a business bottleneck — the fragmented, multi-screen workflow had to become a single dashboard.
Problem
TAs ran every live class across 10 disconnected tools: Excel for attendance, WeChat for parents, the internal 4S platform for data, plus separate video and chat dashboards.
Problem to Solve
TAs were forced to manage their classrooms across 10 different tools simultaneously — creating an unsustainable cognitive load during high-pressure live classes.
Research
To ground the redesign in reality, we combined qualitative shadowing with quantitative validation across the TA population.
User Observation
User Interviews
Quantitative Surveys
Surveys confirmed what we observed: the most demanding monitoring tasks were the least supported by the interface. Nearly every TA depended on continuous student video and uninterrupted access to the instructor's feed just to keep context.
Research Insights
TAs didn’t need more features — they needed one ecosystem. Context switching was crowding out the mentoring that was their actual job.
Consolidate the windows.
Minimise the switching.
Streamline attendance
Monitor engagement live
Route parent feedback
What has to sit on one screen
Design Objective
Three goals anchored the redesign: consolidate the scattered tools into one source of truth, automate routine workflows like attendance, and cut the visual clutter driving cognitive overload.
Unify
Consolidate fragmented tools into one workspace
Automate
Streamline repetitive operational workflows
Simplify
Reduce cognitive load with a clearer interface
Information Architecture & Layout Exploration
We evaluated six dashboard configurations (1:4, 1:1, 3:2, 1:3:1, 1:1:3, 1:2:2) against information density, efficiency, and visual hierarchy.
Overly symmetrical layouts (e.g., 1:1) lacked a clear focal point, while margin-heavy designs increased mouse travel and slowed response to critical alerts.
1:1 (Symmetrical)
Low
Split Focus
2 / 5
1:4 (Side-heavy)
High
Margin heavy
3 / 5
1:3:1 (Center Focus)
Optimal
Centralized
5 / 5
Low-Attention Content Summary-Level Data
+
Semi-Automatic Monitoring
Low-Attention Content Course Progress Awareness
Medium-Priority Content & ActionsLarger Workspace
Primary Focus AreaPositioned at the Visual Center
High-Priority Content & ActionsPreserves Familiar Three-Panel Workflow
Attention Hierarchy
Low AttentionOverview & Awareness
Medium AttentionReview & Operate
High AttentionCore Focus & Actions
Pre-Class: Streamlining Attendance Workflows
The 15 minutes before a broadcast were a scramble: cross-reference an Excel roster, check 4S for live logins, then open WeChat to ping missing students one by one — error-prone, and stressful at exactly the wrong moment.
The redesigned dashboard surfaces real-time attendance and interaction readiness at a glance.
Serves as the starting point of the visual scanning pattern, establishing a primary visual hierarchy.
Verify data metrics based on questionnaire results
Chasing one absence used to take five steps across five separate tools. It is now a single click, without leaving the dashboard.
Live Session Monitoring
OA/4S
Wechat / 4S
Excel
Internal App
Status
Based on questionnaire data, the system consolidates the abnormal student states teachers care about most — Absent / Idle / Offline — into dynamic alert prompts.
- 15 mins pre-class: alerts for students who haven't shown up.
- In-class: alerts for offline / idle students.
Actions
One-click attendance / idle / offline reminders — send WeCom (Enterprise WeChat) notifications to parents.
In-Class: Student Engagement Monitoring
Student Analytics
Track learning metrics
Video Monitoring
Monitor student engagement
Student learning data
The learning analytics module provides teachers with real-time visibility into students' online status and classroom engagement.
After class, the system automatically generates learning reports and sends them to parents, ensuring clear and continuous tracking of student progress.
AI-Assisted Video Monitoring
Monitoring dozens of student video feeds is the TA's most demanding task. In the old system, small thumbnails and manual scrolling made it easy to miss distracted or struggling students.
The new 1:3:1 layout introduces an AI-powered video grid that automatically detects and highlights abnormal behaviors (e.g., absence, poor lighting, disengagement) and prioritizes these feeds for quick review — shifting the TA's task from passive scanning to active mentoring.
In-Class: Driving Engagement & Recommendations
TAs are also expected to actively drive participation — but with the old tools they couldn't track in real time who was engaged and who was falling behind.
So we designed a recommendation layer that suggests interventions as performance data arrives — a streak of correct answers prompts a one-click note of encouragement.
Manual Monitoring
Track learning metrics
Automated Monitoring
Monitor student engagement
@ Mention in Chat
💬 Direct Message
💬 Direct Message (Audio)
Interaction Participation
Participation rate achieved during live interactive exercises using the new platform.
Interaction Accuracy
Accuracy rate maintained by students, monitored seamlessly by TAs.
Post-Class: Automated Reporting for Parents
After class, summarizing performance for parents meant manually compiling data from several platforms.
The reporting module now aggregates attendance, participation, and behavioral data into a standardized learning report automatically.
"The automated reports save me at least an hour every day. I can finally give parents detailed feedback immediately after class finishes"
Jie Li — Lead Class Mentor
Interaction Design Guidelines
Impact & Results
The redesign consolidated 10 legacy tools into a single interface, and scored 8.3/10 in post-launch teacher satisfaction.
Teacher Satisfaction
Post-launch satisfaction score for the redesigned Class Mentor platform.
Task Completion Time
Core monitoring tasks take approximately 70% less time than the legacy multi-tool workflow.
Monitoring Efficiency
AI-assisted video monitoring improved preview and monitoring efficiency threefold.
Core monitoring tasks — attendance, engagement, parent communication — now take about 70% less time, and the 1:3:1 layout with AI assistance tripled video monitoring efficiency. The platform serves over 50,000 TAs, and that efficiency gain lays the groundwork for raising the tutor-to-student ratio: each TA supporting more students without diluting personal attention.
Reflection & Next Steps
Shadowing TAs in their actual working conditions is what made this work. Mapping their multi-tool reality first meant the solution addressed real cognitive bottlenecks rather than just looking modern.
Next: using the platform's own data for prediction and guidance, so the system anticipates what a TA needs before they go looking for it.