Advenno built EduSpark, an adaptive learning platform with AI-personalized learning paths, intelligent tutoring, and social learning features. For BrightPath Academy's 45,000 learners, completion rates jumped from 34% to 81% and NPS reached 72.
The Challenge
BrightPath Academy had established itself as a mid-market online education provider with strong course content created by industry practitioners. However, the platform's delivery mechanism was fundamentally limited: every learner watched the same sequence of pre-recorded video lectures, completed the same assignments in the same order, and moved at the same pace regardless of their background or ability. A software engineer with 5 years of experience enrolled in an advanced Python course sat through 4 hours of basics they already knew before reaching new material — if they hadn't already abandoned the course out of frustration. Conversely, a career-changer with no programming background hit a wall at module 3 with no support, remediation, or alternative explanation available. The data told the story clearly: 34% of enrolled learners completed their courses, and drop-off analysis showed two distinct peaks — at the 15% mark (bored advanced learners) and the 45% mark (overwhelmed beginners). With a customer acquisition cost of $45 and subscription pricing of $29/month, BrightPath needed learners to stay engaged for at least 4 months to achieve positive unit economics. The 34% completion rate meant two-thirds of learners churned before reaching that threshold, creating a fundamental business model crisis. Competitors like Coursera and Udemy were investing heavily in personalization, and BrightPath's leadership recognized that their differentiation through practitioner-created content would erode quickly without a platform that could deliver that content intelligently.
- 34% course completion rate with drop-off peaks at 15% (bored advanced learners) and 45% (overwhelmed beginners)
- Zero personalization — all 45,000 learners experienced identical linear video sequences regardless of background
- Effective cost per successful outcome was $132 versus $45 CAC, creating unsustainable unit economics
- No adaptive difficulty, alternative content formats, or intelligent support for struggling learners
- Competitors investing heavily in AI personalization threatened BrightPath's practitioner-content differentiation
- Monthly churn averaged 18% as learners disengaged from one-size-fits-all courses
Our Solution
Advenno designed EduSpark as a complete reimagining of BrightPath's learning delivery infrastructure. The platform begins each learner's journey with a diagnostic assessment — an adaptive quiz that efficiently maps existing knowledge across the course's concept graph in under 10 minutes. Based on results, the AI generates a personalized learning path that skips already-mastered concepts, sequences topics in the optimal order for the individual learner, and calibrates initial difficulty to their demonstrated level. As the learner progresses, real-time performance analysis adjusts difficulty dynamically — if quiz scores exceed 90% consistently, the system increases complexity; if they drop below 60%, it introduces supplementary explanations and practice exercises before allowing progression. Content is available in four formats — video, interactive text with embedded exercises, audio for commuters, and hands-on lab environments for technical courses — and the system learns each learner's format preferences over time. An AI tutoring assistant powered by GPT-4 answers learner questions in the context of the specific lesson they're studying, providing explanations, analogies, and worked examples without simply giving away answers. A spaced repetition engine schedules review sessions at scientifically optimal intervals to maximize long-term retention. Social features including matched study groups of 4-6 learners at similar levels, peer mentoring connections, and weekly accountability check-ins create the community support structures that research shows are the strongest predictor of course completion.
- Diagnostic assessment that maps existing knowledge and generates personalized learning paths
- Dynamic difficulty adjustment based on real-time quiz performance and engagement patterns
- Four content formats — video, interactive text, audio, and hands-on labs — with learned preferences
- AI tutoring assistant providing contextual explanations without giving away answers
- Spaced repetition engine scheduling reviews at scientifically optimal intervals for retention
- Algorithm-matched study groups and peer mentoring for social accountability
- Progress dashboard with estimated completion timeline and mastery visualization
Our Approach
Learning Science Research
Partnered with BrightPath's instructional designers and an external learning science advisor to establish the pedagogical framework. Reviewed 40+ research papers on adaptive learning, spaced repetition, and social learning theory. These principles directly informed every feature decision in the platform design.
Content Knowledge Graph
Worked with course creators to decompose all 120 courses into concept-level knowledge graphs — mapping prerequisite relationships, difficulty levels, and assessment criteria for 8,400 individual learning concepts. This graph became the foundation for adaptive path generation and diagnostic assessments.
Adaptive Engine Development
Built the personalization engine using a combination of Item Response Theory for assessment calibration and a reinforcement learning model for path optimization. The system was trained on historical learner interaction data from 180,000 course enrollments, learning to predict which content sequence maximizes completion probability for each learner profile.
Multi-Format Content Pipeline
Created tooling and workflows for BrightPath's content team to produce each lesson in four formats efficiently. Developed an AI-assisted transcription and adaptation pipeline that converts video lectures into interactive text and audio formats with 80% automation, requiring only editorial review from subject matter experts.
Controlled Rollout with A/B Testing
Launched EduSpark to 5,000 learners in an A/B test against the existing platform over 8 weeks. The adaptive group showed a 71% completion rate versus 34% for the control — more than double. Social features added during weeks 5-8 pushed the adaptive group to 78%, validating the combined approach before full migration.
The Results
EduSpark fundamentally transformed BrightPath Academy's educational outcomes and business trajectory. Course completion rates climbed from 34% to 81% within six months — the most dramatic improvement in the company's history and well above the 60% target that leadership had set. The Net Promoter Score surged from 23 to 72, moving BrightPath from the 'passives' category into the 'excellent' range by industry standards. Learner retention improved 64% year over year, with average subscription duration extending from 3.1 months to 5.8 months — comfortably past the 4-month break-even threshold. Average time to course completion decreased by 28% as the adaptive engine eliminated redundant content for learners with existing knowledge, delivering the same outcomes in less time. The AI tutoring assistant handled 34,000 learner questions in its first quarter, with 91% rated as helpful — effectively providing one-on-one support that would have required 50+ human tutors. Study groups became the most beloved feature, with 78% of active learners participating and completion rates 23% higher among study group members versus solo learners. BrightPath's organic search traffic increased 45% as course review scores improved and word-of-mouth referrals accelerated, reducing customer acquisition cost from $45 to $31. The improved unit economics enabled BrightPath to raise a $12M Series A round, with investors specifically citing the adaptive learning technology as the key differentiator.
Return on Investment
Technologies Used
Integrations
EduSpark is the most significant investment we've ever made in our learners' success. Going from 34% to 81% completion isn't just a metric improvement — it means thousands more people are actually gaining the skills they came to us for. The adaptive engine feels like having a personal tutor for every single student.
Project Gallery
Lessons Learned
- Building the concept-level knowledge graph for 120 courses was the most labor-intensive but most critical investment — adaptive learning is only as good as the underlying content structure
- Social features — especially study groups — had more impact on completion than the adaptive engine alone, validating that learning is fundamentally a social activity
- The AI tutoring assistant needed careful prompt engineering to avoid giving away answers while still being genuinely helpful — the sweet spot required 3 iterations
- Multi-format content production was initially seen as 4x the work, but the AI-assisted pipeline reduced it to 1.4x while dramatically improving accessibility
Summary
Advenno built EduSpark, an adaptive learning management platform for BrightPath Academy's 45,000 learners. The AI-powered system personalizes difficulty, pacing, and content format per learner while providing intelligent tutoring and social learning features. Completion rates jumped from 34% to 81%, NPS reached 72, and learner retention improved 64%.
Key Takeaways
- Diagnostic assessments that map existing knowledge in 10 minutes enabled personalized paths that skip mastered content
- Dynamic difficulty adjustment eliminated both the boredom of advanced learners and the overwhelm of beginners
- Study groups emerged as the strongest completion predictor — 78% participation with 23% higher completion rates
- AI tutoring handled 34,000 questions in Q1 with 91% helpfulness rating, replacing the need for 50+ human tutors
- A/B testing showed the adaptive engine more than doubled completion rates before full rollout commitment
Frequently Asked Questions
Key Terms
- Adaptive Learning
- An educational approach using AI and algorithms to adjust content difficulty, sequence, and format in real time based on individual learner performance and behavior.
- Item Response Theory (IRT)
- A psychometric framework for designing and scoring assessments that estimates both the learner's ability and each question's difficulty, enabling efficient diagnostic testing.
- Spaced Repetition
- A learning technique where review sessions are scheduled at increasing intervals based on memory decay curves, scientifically proven to optimize long-term knowledge retention.
Facts & Statistics
Sources & Citations
- National Training Lab: Learning Pyramid
- EdSurge: The State of Adaptive Learning 2025
Facing a similar challenge?
Every project has its own constraints. If this case study resonated with a problem you are working through, we are happy to share what shaped the decisions we made here.
Tell Us What You Are Working On
