Insights
Institutional analytics for the SI MedEd student platform: who is using it, how they perform, whether they come back, and how faculty engage. Denominators are unique users (distinct userIds), never raw event counts — every number below says exactly what it counts and over which dates.
Loading data window…
Insights feed
How to read this: notable patterns surfaced automatically from the events data — score drops, peak days, questions to review, engagement shifts, and milestones. Each card names its denominator (n=) and links to the chart it came from. Dismiss a card to hide it on this device. First-seen cohort detection is relative to the available event history.
Analyzing…
Usage trends
How to read this: one point per day. A day with no activity plots as 0; a panel labeled "no data" means the events stream has nothing for the whole range.
Daily active users
Denominator: unique userIds with any event that day ·
Questions answered per day
Count: question_answer events per day ·
Elsie API calls per day
Count: api_usage records per day ·
Performance trends
How to read this: each week shows the mean practice-set score for sets completed that week, split by exam type. Weeks with no completed sets leave a gap (never 0%).
Average practice score by week and exam type
Denominator: mean of practice_set_complete.ctx.pct per week × exam type · last 12 full weeks + current week
| Week starting | Exam type | Sets | Unique users | Avg score |
|---|---|---|---|---|
| Loading… | ||||
Cohort comparisons
How to read this: CDU cohorts vs. external users over the last 90 days — a fixed window so cohorts stay comparable even when the range selector above changes. CDU cohorts are shown first.
Active users by cohort
Denominator: unique userIds active in the last 90 days
Average practice score by cohort
Denominator: mean practice_set_complete.ctx.pct in the last 90 days; users with no completed sets are excluded, not scored 0
Questions per user by cohort
Denominator: question_answer events ÷ unique users active in the last 90 days
| Cohort | Active users | Questions | Sets completed | Avg score | Questions / user |
|---|---|---|---|---|---|
| Loading… | |||||
Retention
How to read this: of the unique users active in week N, what share were active again in week N+1. The current (incomplete) week has no N+1 yet, so it shows "—".
Week-over-week retention
Denominator per week: unique userIds active that week
| Week starting | Active users | Returned next week | Retention |
|---|---|---|---|
| Loading… | |||
Growth
How to read this: new users are counted in the month of their first seen event. Note: first-seen is relative to the available event history (starting —), so the earliest month may undercount users who arrived before logging began.
New users per month
Denominator: unique userIds whose first event in the available history falls in that month
Cumulative questions answered
Count: running total of question_answer events by month
| Month | New users | Questions answered | Cumulative questions |
|---|---|---|---|
| Loading… | |||
Faculty & SME engagement
How to read this: actions logged to activity_log (reviews, edits, approvals, sign-ins) per week, plus the most active humans. LC (automated) actions are excluded from the contributor table.
Faculty actions per week
Count: activity_log entries per week (human + LC) · last 12 weeks
| Contributor | Actions | Sign-ins | Last active |
|---|---|---|---|
| Loading… | |||
Data sources: Firestore events (student activity stream), api_usage (Elsie calls), activity_log (faculty/SME actions). Charts render in the browser from the fetched records; no records are altered. Where a collection is empty, panels say "no data yet" instead of showing zeroes.