How Ruangguru Reached 22M+ Students via Tech

 Published: February 2026 | Reading Time: 15 minutes

Key Takeaways

• Ruangguru scaled 1M → 28M students through a three-phase partnership: stabilization (99.5% uptime in 30 days), monolith-to-microservices migration (40x concurrent growth), live learning/AI features.
• COVID-19 emergency scaling (Mar 2020): 10x DAU (200K→2M), 8x concurrent (50K→400K) in 2 weeks—zero downtime via 72-hour capacity sprints.
• Microservices migration: Auth/Content/Video/Payment/Analytics/Live Class services with dedicated DBs—independent scaling, reduced risk for 2M+ exam peak users.
• Video optimization: Multi-CDN (Akamai/Cloudflare/local), edge caching, adaptive bitrate—cut start time 8s → 0.8s (10x) across 2G-5G.
• Embedded team model: AgileSoftLabs engineers + Ruangguru staff—evolved from 2019 project delivery to 2022+ strategic advisory (internal team now leads development).

The Challenge: Scaling Education Technology

When Ruangguru first engaged AgileSoftLabs in 2019, they faced a challenge common to rapidly growing startups: technology that worked at 100,000 users was failing at 1 million users — and they were targeting 10 million. Peak loads during exam preparation periods crashed the platform. Video buffering frustrated students. Slow page loads drove users to competitors.

The urgency was acute. Indonesia's education market was digitalizing rapidly, competitors were emerging, and Ruangguru's window to establish market leadership was measured in months, not years.

Initial State (2019)

• Users: ~1 million registered students
• Content: 100,000+ learning videos and practice questions
• Peak load: 50,000 concurrent users
• Critical issues: 6-second page load times, 8-second video start times, system crashes during exam periods, 94% uptime

The Growth Trajectory

User growth:

• 2019: 1M registered users
• 2020: 6M users (COVID-19 acceleration)
• 2021: 15M users
• 2022: 22M users
• 2024: 28M+ users

Concurrent user peaks:

• Normal day: 200,000 concurrent
• Exam prep season: 800,000 concurrent
• National exam day: 2M+ concurrent
• Challenge: 40x variation in load requiring elastic infrastructure

This trajectory demanded architecture that could scale horizontally, isolate failures, and optimize for Indonesia's unique infrastructure challenges — users on everything from 5G in Jakarta to 2G in remote islands.

Technical Partnership Approach

Our engagement evolved through three phases as Ruangguru's needs matured.

Phase 1: Stabilization (3 Months)

Before scaling, we had to stabilize. The initial assessment identified five critical bottlenecks.

Issues identified:

• Database bottleneck: Single PostgreSQL instance serving all queries
• Video delivery: Origin server handling all video requests directly
• Session management: In-memory sessions lost during restarts
• No auto-scaling: Manual capacity management requiring hours of lead time
• Limited monitoring: Reactive incident response, no proactive alerting

Immediate actions:

• Database read replicas with connection pooling (reducing read latency 60%)
• CDN implementation for video content (offloading 95% of video traffic from origin)
• Distributed session management via Redis cluster
• Auto-scaling configuration based on CPU and request metrics
• Comprehensive monitoring with Prometheus, Grafana, and alerting

Results achieved in 30 days:

• Uptime: 99.5% (from 94%, representing 6x fewer outages)
• Page load time: 6s → 2.1s (3x improvement)
• Video start time: 8s → 1.5s (5x improvement)
• Zero exam-period outages during stabilization phase

This foundation — reliability before features — proved essential. You cannot scale what you cannot reliably operate. Our cloud development services emphasize this principle across all engagements.

Phase 2: Architecture Evolution (6 Months)

With stability achieved, we rebuilt for long-term scale through microservices migration.

Key Technical Decisions

The migration was incremental, not big-bang. We moved one service at a time — authentication first (lowest risk), then content delivery, then payment processing, finally analytics. Each migration was validated under production load before proceeding to the next. This minimized business disruption while maximizing learning at each step.

Our custom software development methodology prioritizes exactly this pattern: deliver value continuously, learn from production, adapt based on real data.

Phase 3: Feature Development (Ongoing)

Beyond infrastructure, we built new capabilities that differentiated Ruangguru in the market.

Live Learning Platform:

• Real-time video streaming supporting 100K+ concurrent viewers per session
• Interactive Q&A allowing students to ask questions during live classes
• Whiteboard collaboration enabling visual teaching
• Recording and playback for students who miss live sessions
• Bandwidth adaptation delivering quality appropriate to each student's connection

Adaptive Learning Engine:

• Student performance tracking across all assessments and practice sessions
• Personalized content recommendations based on learning history
• Difficulty adjustment adapting to individual student progress
• Weakness identification highlighting topics needing additional practice
• Learning path optimization guiding students through optimal content sequences

Assessment System:

• Large-scale exam delivery handling 500K+ simultaneous test-takers
• Anti-cheating measures including randomized question order, time limits, and browser lockdown
• Instant grading and feedback for objective questions
• Performance analytics dashboards for teachers and administrators
• Question bank management with tagging, difficulty calibration, and usage tracking

These features, built using our AI and machine learning solutions, transformed Ruangguru from a content library into an intelligent learning platform. The adaptive engine alone increased student engagement by 35% and completion rates by 28%.

For organizations building similar education management platforms, these capabilities — live instruction, personalization, assessment at scale — represent the competitive differentiators that drive market leadership.

Results and Impact

Technical Metrics

Business Impact

Growth metrics:

• User base: 1M → 28M (28x growth over partnership period)
• Content library: 100K → 1M+ videos, practice questions, assessments
• Live classes delivered: 10K/month → 500K/month
• Revenue growth: 15x over partnership period
• Market position: #1 EdTech platform in Indonesia

Student outcomes:

• 10M+ students prepared for national exams using the platform
• 85% of active users report improved academic performance
• 2M+ scholarship assessments processed
• 500K+ hours of live instruction delivered monthly

These outcomes demonstrate that technical excellence translates directly to business and social impact. Faster load times increase engagement. Higher availability builds trust. Scalable infrastructure enables market expansion. Our case studies across industries show this pattern consistently.

COVID-19 Emergency Scaling

When Indonesian schools closed in March 2020, Ruangguru faced an unprecedented challenge: scale capacity 10x in two weeks with zero downtime.

Before (February 2020):

• 200K daily active users
• 50K peak concurrent users

After (April 2020):

• 2M daily active users (10x increase)
• 400K peak concurrent users (8x increase)
• Required timeline: 2 weeks maximum

Our 72-hour emergency response:

• Hour 0-48: Emergency capacity planning, load forecasting, infrastructure sizing
• Hour 48-72: Infrastructure provisioning across compute, database, caching, CDN
• Week 1: Performance optimization sprint targeting highest-impact bottlenecks
• Week 2: Free tier launch enabling all Indonesian students to access platform
• Result: Zero downtime during transition, 99.9% uptime maintained throughout crisis

The ability to execute this scaling event successfully validated every architectural decision from the previous year. Kubernetes auto-scaled seamlessly. Microservices isolated failures. Multi-CDN provided redundancy. Event-driven architecture absorbed traffic spikes.

Organizations building platforms where demand can spike unpredictably — whether education during pandemics, e-commerce during flash sales, or media during viral events — benefit from similar architecture. Our web application development engagements now incorporate elastic scaling as a first-class requirement, not an afterthought.

Lessons from the Partnership

I. What Worked

1. Embedded team model: Our engineers worked alongside Ruangguru's team in their Jakarta office, building internal capability through daily collaboration rather than delivering black-box solutions and leaving.

2. Incremental migration: We moved to microservices piece by piece — authentication, then content, then payments — not big-bang cutover. Each service was validated under production load before migrating to the next.

3. Load testing obsession: We tested at 3x expected peak load before every major event — national exams, product launches, marketing campaigns. Finding breaking points in test environments is cheap; finding them in production is catastrophic.

4. Local optimization: Indonesia-specific choices mattered — local CDN providers for better peering, edge caching in regional data centers, mobile-first design for bandwidth-constrained connections.

5. Knowledge transfer: We documented architecture decisions, ran internal training sessions, and conducted code reviews focused on teaching, not gatekeeping. The goal was always for Ruangguru's team to own the platform.

II. Challenges Overcome

1. Infrastructure variety: Indonesian users span 5G in urban Jakarta to 2G in remote islands. Adaptive streaming, aggressive caching, and offline-first mobile apps addressed this disparity.

2. Peak load unpredictability: Viral content or breaking news could 10x traffic in hours. Horizontal auto-scaling and over-provisioned headroom (always 2x peak capacity ready) prevented outages.

3. Regulatory compliance: Indonesian data localization requirements, content moderation rules, and education ministry standards required legal and technical controls.

4. Rapid feature demands: The business moved faster than typical enterprise development cycles. Two-week sprints, continuous deployment, and feature flags enabled velocity without sacrificing stability.

Similar challenges appear in school management systems, attendance tracking platforms, and curriculum management tools — the architectural patterns transfer across EdTech applications.

Technology Stack

This stack balances maturity (Kubernetes, PostgreSQL), performance (Go, ClickHouse), and developer productivity (Node.js, managed services). For organizations evaluating mobile app development alongside web platforms, the backend architecture remains consistent — React Native apps consume the same APIs as the web frontend.

Partnership Evolution

2019: Project-based engagement

• Specific deliverables (stabilization, then microservices migration)
• Fixed team size (4-6 engineers)
• Clear handoff milestones

2020-2021: Embedded team model

• Engineers working in Ruangguru's Jakarta office daily
• Participating in Ruangguru's sprint planning and retrospectives
• Knowledge transfer as primary objective alongside delivery

2022-Present: Strategic advisory

• Architecture reviews for new product initiatives
• Scaling guidance as Ruangguru expands across Southeast Asia
• Ongoing support relationship for complex technical challenges
• Ruangguru's internal team handles day-to-day development independently

This evolution — from doing the work, to doing it together, to advising while they do it — represents partnership success. The goal was never vendor dependency. The goal was building lasting capability.

Conclusion

Ruangguru's journey from 1 million to 28 million students demonstrates what becomes possible when technology scales with business ambition. The technical principles — incremental migration, obsessive load testing, embedded collaboration, local optimization — are transferable across industries and geographies.

Today, Ruangguru's internal engineering team handles most development, exactly as we planned from day one. Our ongoing role is supporting their continued growth and tackling new technical challenges as they expand beyond Indonesia into Thailand, Vietnam, and the Philippines.

The education technology market is growing globally. The patterns we validated with Ruangguru — elastic infrastructure, adaptive learning, large-scale assessments, live instruction at scale — apply whether you are serving 100,000 students or 100 million.

Building an EdTech platform or scaling an existing one? Contact us to discuss how we can help. Review additional transformation stories through our case studies, explore our education technology products, or follow the AgileSoftLabs blog for ongoing insights on platform scaling, architecture evolution, and technology partnership models.

Frequently Asked Questions (FAQs)

1. How did Ruangguru reach 22M+ students?

Marketplace model connecting 500K+ teachers/students, Google Cloud migration (2017), free pandemic access (10M students), Ruangkelas LMS for 46K schools/480 districts.

2. What technology stack powers Ruangguru?

React/Node.js frontend, gRPC microservices, PostgreSQL, WebRTC live classes, Google Cloud infrastructure—80+ engineers maintain 99.9% uptime for 22M users.

3. Who were Ruangguru's key technology partners?

AgileSoftLabs (React/Node platform for 15M students), Google Cloud (scaled from 2.5M to 22M), Kissflow (workflow automation), East Ventures (funding).

4. What made Ruangguru's scaling successful?

10x cheaper than offline tutoring, personalized learning paths, 40K+ animated videos/1M+ quizzes, 70% rural users, 3x higher university acceptance (69% vs 23% national).

5. How does Ruangguru handle 22M concurrent users?

Microservices (gRPC), horizontal scaling on Google Cloud, CDN caching, real-time WebRTC, data-driven content personalization for K12/professionals.

6. What role did Google Cloud play in scaling?

Migrated 2017: Handled 2.5M students → 22M+; auto-scaling compute, BigQuery analytics, Cloud Storage for 40K videos—cost 30% lower than on-prem.

7. How does Ruangguru monetize 22M students?

Subscription videos ($1-5/mo), private tutor marketplace (300K teachers), corporate training (100K+ employees), offline centers (15 locations).

8. What LMS does Ruangguru use for schools?

Ruangkelas: Free for 46K schools/480 districts, 250+ teacher modules, 200K+ teachers trained—powers national exam prep (7-22% score uplift).

9. Ruangguru's Indonesia edtech competition?

Zenius, Quipper; Ruangguru leads with 22M users vs competitors' 2-5M; 450% revenue growth in 2020, first profitable edtech unicorn.

10. How does Ruangguru personalize for 22M users?

Data-driven learning journeys, AI homework solver (Ruangguru), concept mastery tracking, adaptive quizzes—90% users improve grades.

11. What metrics prove Ruangguru's 22M impact?

69% state uni acceptance (3x national), 90% grade improvement, 500K scholarships, 20M rural students reached, $55M Tiger Global funding.