Amplitude for EdTech

Why Amplitude Fits EdTech Lifecycle Work

EdTech products have a retention problem that most analytics tools aren't built to surface. Users activate in bursts — around enrollment windows, semester starts, exam prep seasons — and then disappear. Standard funnel analysis misses this because it treats all users the same regardless of where they are in an academic calendar or learning journey.

Amplitude solves this when configured correctly. The behavioral cohort engine, combined with its event taxonomy and syncing capabilities, gives you visibility into exactly when learners drop off, which features correlate with completion, and which segments are worth re-engaging versus writing off.

This guide walks you through a production-ready setup for EdTech specifically.

---

Defining Your Event Taxonomy

Your event taxonomy is either your greatest asset or your biggest liability. Most EdTech companies instrument too broadly and end up with a noisy event stream that's impossible to analyze.

The Core Events to Track

Organize events into four categories:

Activation events — actions that indicate a user has reached minimum viable engagement

• `lesson_started` (first lesson in a course)
• `onboarding_completed`
• `profile_setup_finished` (learner profile, goals, level)

Engagement events — actions that signal active learning behavior

• `lesson_completed` with properties: `lesson_id`, `course_id`, `completion_percentage`, `time_spent_seconds`
• `quiz_attempted` and `quiz_passed`
• `streak_maintained` (if your product has streaks)
• `content_searched`
• `note_created` or `highlight_added`

Progress events — milestones that indicate movement through a learning path

• `module_completed`
• `course_completed`
• `certificate_earned`
• `skill_assessed`

Intent signals — behaviors that predict churn or upgrade

• `pricing_page_viewed`
• `feature_locked_hit` (when a free user hits a paywalled feature)
• `session_abandoned` (configured via Amplitude's inactive user detection)
• `cancellation_flow_started`

Event Properties That Matter

Every `lesson_completed` event should carry at minimum: `subject_category`, `difficulty_level`, `content_format` (video, text, interactive), and `device_type`. These properties are what make your cohort analysis meaningful. Without them, you know someone finished a lesson but not whether video learners complete at higher rates than text learners — which is a decision-relevant insight.

---

Segments to Build

Amplitude's behavioral cohort builder is where EdTech analytics gets specific.

Learning-Stage Segments

Build these as dynamic cohorts that auto-update:

1. New Learners (Days 0–7) — users who triggered `onboarding_completed` in the last 7 days but have not yet completed a second lesson
2. Active Learners — users who completed at least 3 lessons in the last 14 days
3. At-Risk Learners — users who were active in the previous 14-day window but have zero `lesson_completed` events in the current window
4. Completed-But-Churned — users who finished at least one course but haven't returned in 30 days

Monetization Segments

• Upgrade Candidates — free users who hit `feature_locked_hit` at least twice in 7 days
• Expansion Candidates — paid users approaching course completion who haven't browsed additional courses
• Win-Back Targets — previously paid users who cancelled and have been inactive for 30–90 days

Academic Calendar Segments

This is the segment most EdTech teams skip and then regret. Tag users at signup with `enrollment_context` (back-to-school, exam prep, new year, corporate onboarding). Build cohorts around these contexts and you'll immediately see that exam-prep users have 40–60% lower 90-day retention than goal-oriented adult learners — which completely changes your win-back strategy for each group.

---

Funnels and Charts to Build First

Don't try to build everything at once. Start with three analyses:

Activation Funnel — `account_created` → `onboarding_completed` → `lesson_started` → `lesson_completed`. Measure conversion at each step within 24 hours of signup. Most EdTech products leak heaviest between `lesson_started` and `lesson_completed` — usually because the first lesson is too long.

N-Day Retention by Content Format — Use Amplitude's retention chart, segmented by `content_format`. If interactive content shows 35% Day-14 retention versus 18% for video-only, that's a product roadmap decision, not just a marketing insight.

Feature Adoption by Cohort — Map which features correlate with users reaching `course_completed`. Use Amplitude's Pathfinder or User Paths to see what high-completers do differently in their first 48 hours.

---

Automations to Configure

Amplitude's value compounds when you connect it to your messaging stack — Braze, Iterable, Customer.io, or similar.

Sync These Cohorts to Your Messaging Tool

• At-Risk Learners → trigger a re-engagement sequence at the 48-hour inactivity mark, not 7 days. Most teams wait too long.
• Upgrade Candidates → sync daily and trigger a contextual upgrade message referencing the specific locked feature they hit
• Completed-But-Churned → trigger a new course recommendation sequence using the `subject_category` from their completed course

Amplitude Audiences Setup

Use Amplitude Audiences (formerly Amplitude Recommend) to build predictive cohorts if you're on a plan that includes it. The churn prediction model works reasonably well for EdTech with 90+ days of event history. Set your training window to match your natural engagement cycle — if your product is tied to a 12-week course, use a 12-week lookback.

---

EdTech-Specific Challenges in Amplitude

Handling Seasonal Spikes

Amplitude's retention curves will look artificially poor if you don't account for cohort timing. A cohort acquired in August for back-to-school will show strong 30-day retention and dramatic 90-day drop-off — not because your product failed, but because the school year ended. Annotate your charts with academic calendar events. Amplitude supports annotations natively. Use them.

Multi-Device and Offline Sync

Many EdTech products support offline lesson completion. If a user completes a lesson offline and it syncs later, the event timestamp may not reflect actual behavior timing. Amplitude's server-side event ingestion lets you pass a `time` property with the original completion timestamp. Implement this early — retroactive fixes are painful.

B2B2C Attribution

If your EdTech product sells to institutions (schools, employers) and end users learn through those accounts, you need a clear group analytics setup in Amplitude. Create an `institution` group type. Track institutional-level engagement alongside individual learner behavior. This matters when you're proving ROI to a district buyer whose renewal is based on student completion rates, not individual satisfaction scores.

---

Frequently Asked Questions

What events should I prioritize if I'm just getting started with Amplitude?

Start with five events: `account_created`, `onboarding_completed`, `lesson_started`, `lesson_completed`, and `subscription_started`. These four cover activation, core engagement, and monetization. Add properties to `lesson_completed` from day one — at minimum `course_id`, `completion_percentage`, and `content_format`. You can expand the taxonomy once you've built your first retention chart and identified where users are actually dropping off.

How do I handle EdTech users who are inactive for legitimate reasons (summer break, exam periods)?

Build a calendar-aware suppression list. Tag users at signup with their academic context and use that to adjust your at-risk thresholds. A K-12 student inactive in July is expected behavior. The same inactivity in March is a churn signal. Amplitude doesn't do this automatically — you manage it through cohort definitions and sync timing in your messaging tool.

Does Amplitude work well for EdTech products with both free and paid tiers?

Yes, but you need to track plan status as a user property, not just an event. Set `plan_type` as a user property that updates on upgrade or downgrade. This lets you segment every chart by monetization status and immediately see whether your freemium activation rate differs from paid activation — which it almost always does, and usually by more than founders expect.

How much historical data does Amplitude need before the behavioral cohorts become reliable?

For manual cohorts based on event criteria, you can start getting signal with 30 days of clean event data and a few hundred users. For predictive features like churn likelihood scoring, you need at minimum 90 days of data and ideally 500+ users per cohort. Plan your instrumentation timeline accordingly — if you're building toward a predictive retention model, the time to start tracking is now, not after your next funding round.