How EdTech Companies Use Data Analytics

Learner Engagement — How It Works:

1. Event tracking
   Every click, video play, quiz attempt is logged.

2. Funnel mapping
   Signup → first lesson → first quiz → completion.
   Edtech companies usually find 40-60% drop
   between first lesson and lesson 3.

3. Segmentation
   New vs. returning, mobile vs. desktop,
   paid vs. free, region, cohort.

4. Re-engagement triggers
   Users inactive for 5+ days → push notification
   or email with a personalised course suggestion.

Tools used:
- Mixpanel, Amplitude, Google Analytics 4
- Segment, RudderStack (event pipelines)
- Braze, CleverTap (engagement automation)

Metrics Notes:

Core metrics used in edtech:

DAU/MAU (stickiness)
- 20%+ is healthy for consumer edtech.
- Under 10% = engagement problem.

Time-on-task
- Not time-in-app. Time actually learning.

Completion rate
- Course completions / enrollments.
- Under 15% = serious content or UX problem.

Lesson 1 → Lesson 3 drop-off
- The single most predictive metric.
- If users hit lesson 3, they often finish.

NPS + feedback
- Qualitative signal that explains the numbers.

Lesson: measure behaviour, not just clicks.

Learning Outcomes — How It Works:

1. Skill graph
   Each lesson is tagged with skills + prerequisites.
   Mastery is tracked per skill, not per lesson.

2. Item Response Theory (IRT)
   Some edtech platforms use IRT to score learners
   on latent ability, not just raw marks.

3. Adaptive paths
   Learners who score high skip ahead;
   struggling learners get remedial content.

4. Predictive analytics
   A classifier predicts completion risk.
   Learners flagged "at risk" get coaching.

Tools used:
- LMS platforms (Canvas, Moodle, custom)
- Python + scikit-learn for models
- Looker / Power BI for instructor dashboards

Predictive Model Notes:

Common model: "Completion risk" classifier.

Features:
- Days since last login
- Average quiz score trend
- Time spent per lesson
- Number of failed attempts
- Support ticket history

Output:
- Score (0-1): probability of completing course
- Risk band: low / medium / high

Actions per band:
- Low: standard nudges
- Medium: weekly progress email
- High: instructor intervention + phone call

Realistic accuracy:
- 70-85% is typical.
- Below 65% = not worth acting on.

Revenue & Growth — How It Works:

1. Cohorts
   Group users by signup week or month.
   Track retention at Week 1, 2, 4, 12.

2. LTV / CAC
   LTV = average revenue per user × lifetime months.
   CAC = marketing spend / paying users acquired.
   Healthy: LTV/CAC > 3.

3. Attribution
   Which channel brought the paying user?
   - Last-click
   - Multi-touch
   - First-touch

4. Pricing experiments
   A/B test plans, prices, free trials.
   Watch for revenue cannibalisation.

Tools used:
- dbt + warehouse (Snowflake, BigQuery)
- Looker, Tableau, Metabase
- Amplitude for product-led cohorts

Cohort & LTV Notes:

Typical edtech retention pattern:
- D1: 30-40% return
- D7: 15-25% return
- D30: 8-15% return
- D90: 3-8% return (paid courses higher)

LTV depends heavily on pricing model:
- Subscription: LTV = ARPU / churn rate
- One-time purchase: LTV = average order value × repeat rate

What to avoid:
- Reporting blended churn (hides weak cohorts)
- Optimising CAC while ignoring LTV
- Trusting attributions blindly

Best practice:
- Cohort by acquisition channel
- Cohort by plan type
- Track 30/60/90-day retention separately

Engagement Examples:

1. Duolingo
   - Streaks, daily goals, and push notifications
     based on user behaviour.
   - "At-risk user" prediction triggers retention
     campaigns.

2. Coursera
   - Tracks time-on-task, video completion, quiz
     attempts.
   - Uses these signals to suggest "next lesson".

3. Khan Academy
   - Mastery-based dashboards for students
     and teachers.
   - Identifies students at risk of falling behind.

4. BYJU'S
   - Engagement + learning logs feed a personalised
     content recommender.

Common patterns:
- Every session produces an event.
- Events feed a segmentation pipeline.
- Segments drive personalised nudges.

Adaptive Learning Examples:

1. Knewton (pre-acquired by Wiley)
   - Adaptive learning engine that adjusts
     difficulty in real time.

2. Smart Sparrow (acquired by Pearson)
   - Authoring platform for adaptive lessons.

3. Squirrel AI (China)
   - Adaptive tutoring system using mastery
     tracking + knowledge graphs.

4. Carnegie Learning
   - Adaptive math tutor built on cognitive models.

Common patterns:
- Skill graph + prerequisite rules.
- Mastery measured per skill, not per lesson.
- Content recommendation based on knowledge gaps.
- Predictive models flag struggling learners early.

Revenue & Growth Examples:

1. Chegg
   - Cohort-based subscription analytics.
   - Retention tracking per acquisition channel.

2. Unacademy
   - Growth analytics for subscription plans
     and cohorts.
   - Attribution across paid + organic channels.

3. Vedantu
   - Cohort retention and LTV analysis
     by course type.

4. Byju's
   - LTV/CAC discipline in performance marketing.

Common patterns:
- Monthly cohort reports, not blended churn.
- Channel-level CAC and LTV.
- Pricing experiments run quarterly.
- Retention is treated as a product metric,
  not just a marketing one.